{
  "text": "歡迎收聽聽書Podcast本大由壯菜工程師製作。\n那個現在想請你在腦海中稍微想像一下一個我們日常工作極度混亂的場景好了。\n嗯,多混亂。\n這是你正身處在一個可能有八個人同時講話的那種會議室裡面。\n哇,那真的很吵耶。\n對,超級吵。\n然後有人在為了那個專案的進度激烈爭論,\n然後旁邊可能還有人在小聲碎碎唸。\n重點是大家不但中英文夾雜\n還不斷互相搶話跟疊話\n這聽起來就是\n每天辦公室會發生的噩夢啊\n沒錯沒錯\n但這個時候如果你想像一下\n如果有一個助理\n能在你剛講完話的\n呃0.16秒以內\n立刻遞給你一份完美的逐字稿\n0.16秒\n這幾乎是同步了吧\n對\n而且這份逐字稿上面\n還清清楚楚的標記了\n刚刚哪一句话是谁说的\n甚至连两个人同时开口讲话的内容\n它都分得一清二楚\n这听起来真的很像某种科幻电影里\n才有的可怕科技对吧\n真的\n一般人听起来一定觉得这不可能啦\n但其实这已经不是科幻情节了\n它已经真实发生了\n就在2026年9月1日\nMETA的超级智慧实验室\n也就是那个MSL\n哦MSL\n对\n他们正式发布了最新的\n即时音讯感知模型\n叫做Muse Voice Transcribe\nMuse Voice Transcribe\n这名字听起来蛮厉害的\n这个技术啊\n它不仅做到了你刚刚描述的那种\n在极度混乱的场景下完美记录\n它甚至在多项基准测试里面\n直接击败了Google跟OpenAI的同级产品\n哇 直接打败两大巨头耶\n所以在这场为你量身打造的\n深度探跑中\n我们绝点不会只是你知道\n很无聊的去念那些声音的规格表\n对那太无聊了\n没错\n我们会带你彻底拆解\n这个打败众多科技巨头的核心技术\n去了解这个AI\n它是如何像我们人类大脑一样\n懂得在听话的时候去犹豫\n犹豫啊\n这个词用的很好\n对吧\n对吧\n然后这项技术将如何彻底颠覆你未来\n跟科技装饰互动的方式\n比如说你的Mac\n甚至是未来的那个AR智慧眼镜\n没错,不过其实要真正理解Muse Voice Transcribe到底有多强大\n我们必须先回头看看\n过去几年的语音辨识技术到底有多么的笨拙\n笨拙,所以过去的方法很卡就对了\n超级卡,我们得先看看过去的那种拼装车系统\n拼装车\n对,因为过去如果你想要完成一份会议的逐字稿\n它其实不是一个单一的人工智慧在运作\n它是需要三套完全独立的系统来接力完成\n呃哪三套\n首先是那个ASR\n也就是自动语音辨识系统\n它只负责一件事情\n就是把听到的声音转成文字\n嗯就单纯听写而已\n没错\n接着这个文字和音讯会被丢给Diarization\n也就是说话者分离系统\n等等说话者分离\n这听起来就像是一个事后侦探也\n对对对\n它就是负责回头去分辨说\n现在这句话到底是会议室里面的哪个人说的\n然后最后还需要过一个叫做\nendpointing的端点侦测系统\n端点侦测\n这又是做什么的\n这个就是用来判断说\n这个人到底讲完话了没有\n系统该不该切断麦克风\n哇这听起来就像是一条\n非常老旧而且缺乏效率的工厂管线\n真的非常贴切的比喻\n你可以想象就是三个工人站在输送袋旁边\n第一个工人只顾着\n听声音写字\n然后把写好的纸条\n赶快传给第二个工人\n然后第二个工人\n再去对比声音\n猜这到底是谁说的\n最后再传给第三个工人\n去决定这句话\n是不是真的结束了\n没错\n就是这么繁琐\n而且只要其中一个环节出错\n整个流程就会大塞车\n真的\n我之前用过一些\n会议记录软体\n最常发生的惨剧就是\n那个\nA老板的话才刚讲到一半\n然后旁边B员工稍微咳嗽了一声\n哦对系统就切断了\n没错系统就以为A讲完了\n直接把句子无情切断\n然后还把后面半句算在B头上\n这超尴尬的\n这真的是传统架构的致命伤\n过去这种做法通常需要大量的事后处理\n也就是等你整段话都录完了\n系统才能开始跑这些复杂的对比流程\n所以根本不可能做到真正的及时\n所以Meta这次的做法是怎么解决这个问题的\n他们这次的做法可以说是破坏性的创新\n他们把三个工人的工作全部交给了一个超级大脑同时完成\n一个大脑全包了\n对\nMuse Voice Transcribe\n它其实是建构在Meta的Muse Spark家族之上的一款自回归多模态模型\n等一下先暂停一下\n那个自回归多模态模型\n这几个字听起来真的很像某种外星语言也\n如果我们要用最白话的方式来理解\n这到底是什么意思\n简单来说了\n多摩泰意味着\n这个模型它可以同时看懂文字\n也能听懂声音\n它不是只吃单一一种资料\n哦 就是眼观四面耳听八方的感觉\n没错\n那自回归的话\n就像是我们在玩文字接龙一样\n它会根据已经接收到的资讯\n不断去预测下一步是什么\n但最关键的突破在于\nMeta在同一个音讯串流里面\n直接引入了特殊的标记\n也就是TOKENS\n这完全打破了我们刚才说的\n那个三阶段管线\n标记\n在在实际运作的画面上\n到底会长什么样子\n你想象一下\n有一条连续的时间轴\n当模型在及时聆听的时候\n只要它侦测到换人说话了\n它不会去呼叫别的系统来帮忙\n那它怎么做\n它会直接在自己生成的文字流里面\n插入一个特殊的标记\n叫做Start of Turn\n也就是换人说话的标记\n然后紧接着标注\n这句话是Speaker A说的\n哇 直接在同一个时间轴上标出来\n对对对\n同样的 刚你停顿或者讲完的时候\n它也会立刻放入一个Speech Endpoint的语音结束标记\n等于说所有的文字内容\n还有身份辨识 语气停顿\n全都在同一个时间轴上\n由这个单一模型一次性全部吐出来\n没有错就是这样\n哇所以这代表着我们再也不用忍受那种很尴尬的语音输入时刻了耶\n你们一定有碰过平时对着手机讲话稍微停顿想个词\n系统就以为你讲完了\n对然后就无情的把你切断\n没错现在这个模型它能真正理解对话的节奏\n它知道你只是在脑袋里面那个卡词了而不是真的结束发言\n没错,这正是单一串流模型带来的巨大优势\n但其实,这背后隐藏着一个极度矛盾的工程难题\n怎么说?\n既然这是一个超级模型\n要及时处理所有的声音、身份和文字\n那他怎么知道什么时候该把文字吐出来?\n哦,对耶\n如果他一听到声音就立刻输出文字\n那肯定很容易听错嘛\n但如果他为了追求完美\n等听完整句话才输出\n那又失去了我们刚才说的\n及时的意义\n这真的很矛盾\n这就像我们人类大脑\n在听别人说话一样\n你想想看\n如果有人对我说苹果\n就很简单的日常词汇\n我的大脑几乎不需要思考\n马上就能写下来\n对 瞄懂\n但如果今天\n对方正在讲一段话\n但如果今天,對方正在講一個超級複雜的醫學專有名詞,或者是他講話含糊不清,甚至口音很重,我的大腦其實會稍微,呃,猶豫個零點幾秒。\n沒錯,會卡一下。\n对,就是要等它把所有的音节都发音完整了,然后我在脑海中拼错确认之后,我才敢写下来。\n这个AI模型难道也会模仿人类的这种犹豫吗?\n这个人类大脑的类比非常到位耶。\n其实META的工程师就是循着这个逻辑去设计的。\n他们在技术摆笔书里面把这个称之为自适应延迟技术。\n自适应延迟,听起来蛮高级的。\n为了做到这点,模型在接收音讯的时候,它会把声音切成极小极小的区块,精确来说是80毫秒一个区块。\n80毫秒,这也太短了吧?\n超短的,也就是说在一秒钟之内,模型会接收到12.5个音讯区块。\n而在接收到每一个80毫秒区块的当下,模型都必须做出一个决定。\n什么决动?\n就是我是要现在立刻输出文字\n还是我要输出一个特殊的Next Audio标记\n告诉自己说\n我需要再多听下一个音讯巨块\n这很酷耶\n但他怎么知道自己该不该等啊\n工程师到底是怎么教懂一个AI学会去内心的呢\n这里的关键就在于强化学习\n工程师在训练它的时候\n设计了一个非常严苛的奖励机制\n有多严苛\n他们把那个字错率的惩罚跟延迟时间的惩罚\n用惩罚结合在一起\n你可以想象这是一场极限挑战\n如果模型当个快手\n一听到声音就乱拆\n导致疯狂写错字\n他就会被扣分\n那如果他想当个腕面主义者呢\n哇\n如果他为了一个字等了两三秒才输说文字\n他一样会被狠狠扣分\n哇\n这就像是在训练一个极度严格的听打员\n你就是不能错\n但是你也不能慢\n就是这种严格的训练\n逼迫模型学会了去判断语境\n碰到简单的日常对话\n它就立刻秒翻\n碰到尖色的专有名词\n或者是模糊的发音\n它就会选择多等个几十毫秒\n收集更多声音特征之后再输出\n这蛮合理的\n对\n在工程界\n这被称回达到了Pareto Front\n也就是在速度与准确度之间\n找到了无法再被超越的最佳边界\n最终的结果是\nMuse Voice Transcribe\n从你讲完话\n到萤幕上出现正确文字\n只需要惊人的0.16秒\n等等\n我要在这里稍微挑战一下\n0.16秒\n人类大脑神经传导\n做出反应的时间\n搞不好都比这个慢\n一个AI模型要在这0.16秒内\n切分音讯\n决定要不要等待\n还要辨识是谁说话\n并伸出文字\n他怎么可能在这么短的时间内运算完\n却不吃垮电脑的效能\n这真的是个非常好的质疑\n直觉上做这么多事情\n应该会严重拖慢系统\n对啊\n听起来运算量超大的\n但实际情况恰好相反\n正因为它舍去了过去那种\n三套系统互相传递资料的管线架构\n它省去了系统之间来回沟通\n和对时间轴的庞大运算成本\n就不用大家互相等来等去\n没错\n一个单一的自回归模型\n只要硬体的记忆体频宽足够\n它处理连续串流的效率\n其实远比传统的分工合作还要快得多\n这也就是为什么\n它能压榨出0.16秒这种极限速度\n哇 原来是这样\n理论和底层机制听起来确实很惊人啦\n但我们都知道\n实验室里面的数据是一回事\n把模型丢进残酷的真实世界\n面对各种混乱的对话场景\n那又是另外一回事了嘛\n它在真实环境中的表现到底怎么样\n那我们就来看看知名的AI评测机构\nArtificial Analysis给出的硬数据吧\n在纯英文的即时串流测试里面\nMuse Voice Transcribe的字错率仅仅只有3.1%\n3.1%\n这算是很好的成绩吗\n非常好\n这是什么概面\n它直接击败了目前语音领域的王者\n也就是那个11实验室\n他们的3.6%\n哦 11 labs\n对\n它也赢过了Oper AI的GPT Live Transcribe的3.9\n甚至把Google在8月才刚发布的\nGerman Live 3.5 Transcribe Live的4.0\n都抛在脑后\n哇 全部斩压耶\n没错\n而且在讲者辨识错误率上\n它的错误率只有17.5%\n这个数字甚至赢过了市面上许多\n允许你事后慢慢处理的那种非即时模型\n不过我觉得我们必须拉回现实一下\n我们的听众最关心的\n绝对不是在一个纯英文\n大家都自证强缘的环境里面\n它有多准\n嗯 确实\n对嘛\n我们在排湾\n平时讲话可是中阴夹杂的耶\n像是那个\n我明天有个meeting\n记得remind我一下\n这种句子才是我们最正式的日常啊\nMeta其实在这方面下了很大的功夫\n这款模型发布的时候\n就已经支援了70多种语言\n其中有25种是经过完整验证的\n最厉害的是\n它原生就支援句子内部的鱼马转换\n鱼马转换\n意思是它不需要我去切换那个语言设定\n就能自动听懂吗\n是的\n在官方释出的测试来源里面\n有一段充满硬核科技术语的中文录音\n测试者说\n用Olama跑了向Meta的Muse Glamour\n直接deploy在我的NVIDIA RTX 3090上\n我这句话里面充满了英文专有名词数字\n而且中英文切换完全是随机的耶\n没错\n母亲不需要任何提示就能无缝辨识\n还能根据文法的上下文\n去自动校正评字\n说到真实世界的混乱\n我看到资料里面\n还有一个让我下巴掉下来的测试\n你说那个爬山的那个吗\n对对对\n那个是一个长达一小时的音档\n里面竟然有11个人\n在洛杉矶讨论爬山\n甚至还聊到遇到土狼的经历\n你想想看\n11个人在闲聊耶\n一定有人在大笑\n有人互相抢话\n背景绝对还有各种环境噪音\n这对AI来说\n根本是地狱级的考验\n真的\n结果系统竟然能精准的\n把那11个人的发言\n一次不漏的拆解开来\n还标上speakerA到K\n这如果叫人类的听答员来做\n听个10分钟\n绝对崩溃磁石吧\n但这在技术上\n到底是怎么办到的\n如果两个人\n在同一微秒重叠讲话呢\n这就是我们刚刚说的\n多模态标记发挥威力的地方了\n当两个人同时讲话的时候\n传统系统的音轨会整个混在一起\n变成无法辨识的噪音\n嗯就胡称一团\n对\n但Muse Voice Transcribe\n因为是从极短的80毫秒音讯区块\n去提取特征\n它可以同时触发多个讲者的token\n可以同时触发\n没错\n它会在文字流里面并行输出\nSpeaker A说了这一段\n同时Speaker B正在说那一段\n它不是单线程的处理\n而是能像解开毛线球一样\n把交织在一起的音讯分离出来\n这技术真的蛮扯的\n不过身为替台湾听众把关的酒色\n哦,什麼事實?\n就是,根據台灣知名科技媒體電腦王阿達的實際測試,這項技術目前有一個在地化的致命缺陷。\n嗯,在地化\n对,尽管你对着麦克风说的是很标准的台湾繁体中文\n它目前在荧幕上输出的文字依然全都是,呃,简体中文\n这确实蛮麻烦的\n对啊,不管设定怎么调,目前就是无法原声输出繁体字\n这对台湾的使用者来说绝对还是一个不小的痛点啊\n这的确是一个很明显的在地化缺陷\n其实这显示出虽然模型的听力没有问题\n它的神经网络完全能理解中文的语义\n听得懂但写出来不对\n没错\n但在最后的文字渲染与输出层面\nMETA的训练资料库\n可能还是过度向简体中文倾斜了\n这也反映了全球化大语言模型\n在落地的时候\n必然会面临到的文化与区域细节挑战了\n这部分就需要META后续\n透过微调来解决了\n既然提到META的策略\n我们来聊聊一个很反常的商业布局好了\n我们都知道\nMeta过去这一年来一直高举着开源的大旗\n把他们最大的AI模型免费开放给全世界\n对 开源界的老大哥\n没错\n但这一次Mail Voice Transcribe却选择了封闭权重\n也就是不开源了\n为什么\n这次Meta确实选择了一条不一样的路\n如果你是开发者\n你无法直接下载模型版身\n你必须透过Meta Model API来使用\n哦 要收费了\n不过收费其实相当低廉啦\n大约是每1000分钟的音讯收费3美元\n换算下来一个小时大概才0.18美元\n那真的蛮便宜的\n对\n另外一般使用者其实可以透过他们新推出的程式撰写工具\n叫做Muse Code\n或者是直接在Mac电脑上下载Meta AI的应用程式来使用这项技术\n所以这到底意味着什么\nMeta把这个听力超强的模型紧紧锁在自家的生态系里面\n甚至还跟Mac的键盘深度绑定\nMark Zuckerberg到底在盘算什么更大的棋局\n这不仅仅是为了赚那个0.18美元吧\n绝对不是为了这点API的零头啦\n如果我们深入观察 Zuckerberg近期的内部发言\n他不断强调未来的AI必须是个人化的超级智慧\n个人化的超级智慧\n这听起来很抽象耶\n这跟一个语音转文字的模型有什么直接关系\n关系太大了\n如果AI要成为你最贴身\n最懂你的个人助理\n它就不能只是一个\n你对着手机大喊\n嘿帮我设定闹钟\n那种死板的声控工具\nMeta的最终目标是硬体\n特别是他们正在\n大力发展的AI视觉眼镜\n哦 AI眼镜\n你试想一下\n你戴着AI眼镜的一天好了\n你走在台北吵杂的街头\n旁边有车流声\n你正在点一杯珍珠奶茶\n店员问你要几分甜\n这时你旁边的朋友\n突然插花问你\n等下要去哪里\n哇 这个场景非常真实\n也非常混乱\n对 如果你的AI眼镜\n没有完美的机器听觉\n它就只是一个\n戴在脸上的笨重摄影机\n但有了Muse Voice Transcribe\n眼睛能在0.16秒内\n分辨出哪句是电源说的\n哪句是朋友说的\n还能过滤掉车流声\n没错 甚至刚你转头\n忘记朋友刚才问什么的时候\nAI已经把答案准备好\n投射在你的镜片上了\n哇 这画面感很强耶\n这就是为什么Meta现在让你只要在Mac电脑上按住FN键\n就能在任何视窗进行痊愈的及时听写\n这不是一个单纯的打字工具\n这是Meta在训练你习惯与AI无缝对话\n为未来的穿戴式硬体生态系铺设最底层的基础设施\n今天的探讨真的把很多点都串起来了\n我们现在正在见证一个历史性的转捩点\n机器在复杂环境下的听觉处理能力\n正在以前所未有的速度超越人类\n对,真的很快\n他能毫不犹豫地在0.16秒内处理资讯\n他能同时分辨十几个人的声音\n他甚至学会了像人类一样\n在困难的词汇上稍微犹豫一下\n没错,从过去笨重的三步骤拼装车系统\n进化到现在单一串流的超级模型\n这不仅仅是技术指标上的胜利而已\n更是人机互动方式的一次彻底革命\n真的是革命\n虽然目前还像是繁体中文资源不足这样的缺点\n但这项技术为未来AI眼镜\n和无缝语音互动所打下的基础\n其实已经预示了科技的下一步走向了\n真的\n在我们结束今天的深度探讨之前\n我想留给你们一个发人深审的问题\n想象一下\n如果未来的每一台装置\n每一副眼镜\n都能在0.16秒内\n完美记录下\n20个人在会议室里\n互相强化\n甚至私下碎碎念的内容\n并精准标记\n到底是谁说的\n这对我们的社会\n到底意味着什么\n嗯\n这蛮可怕的\n对吧\n当过去那种\n听过就算了的随性闲聊\n全部变成可以被精准搜寻\n被永久记录的\n完美数位逐字稿的时候\n这会让我们的工作与生活\n变得极度高效\n还是会让我们在每次开口之前\n都变得极度焦虑多疑呢\n这是一个值得你在关掉节目后\n好好思考的问题\n谢谢收听\n欢迎订阅分享\n听书Podcast频道\n我们下次见",
  "language": "zh",
  "duration": 1312.750313,
  "segments": [
    {
      "start": 0.0,
      "end": 3.76,
      "text": "歡迎收聽聽書Podcast本大由壯菜工程師製作。",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.1484853858189867,
      "no_speech_prob": 3.484778091089602e-11,
      "compression_ratio": 1.2879746835443038
    },
    {
      "start": 3.76,
      "end": 11.84,
      "text": "那個現在想請你在腦海中稍微想像一下一個我們日常工作極度混亂的場景好了。",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.1484853858189867,
      "no_speech_prob": 3.484778091089602e-11,
      "compression_ratio": 1.2879746835443038
    },
    {
      "start": 11.84,
      "end": 13.16,
      "text": "嗯,多混亂。",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.1484853858189867,
      "no_speech_prob": 3.484778091089602e-11,
      "compression_ratio": 1.2879746835443038
    },
    {
      "start": 13.16,
      "end": 18.68,
      "text": "這是你正身處在一個可能有八個人同時講話的那種會議室裡面。",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.1484853858189867,
      "no_speech_prob": 3.484778091089602e-11,
      "compression_ratio": 1.2879746835443038
    },
    {
      "start": 18.68,
      "end": 20.08,
      "text": "哇,那真的很吵耶。",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.1484853858189867,
      "no_speech_prob": 3.484778091089602e-11,
      "compression_ratio": 1.2879746835443038
    },
    {
      "start": 20.08,
      "end": 21.52,
      "text": "對,超級吵。",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.1484853858189867,
      "no_speech_prob": 3.484778091089602e-11,
      "compression_ratio": 1.2879746835443038
    },
    {
      "start": 21.52,
      "end": 25.84,
      "text": "然後有人在為了那個專案的進度激烈爭論,",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.1484853858189867,
      "no_speech_prob": 3.484778091089602e-11,
      "compression_ratio": 1.2879746835443038
    },
    {
      "start": 25.84,
      "end": 28.88,
      "text": "然後旁邊可能還有人在小聲碎碎唸。",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.1484853858189867,
      "no_speech_prob": 3.484778091089602e-11,
      "compression_ratio": 1.2879746835443038
    },
    {
      "start": 28.88,
      "end": 31.68,
      "text": "重點是大家不但中英文夾雜",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.1757877823574091,
      "no_speech_prob": 3.4782416530321214e-11,
      "compression_ratio": 1.2669039145907472
    },
    {
      "start": 31.68,
      "end": 34.64,
      "text": "還不斷互相搶話跟疊話",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.1757877823574091,
      "no_speech_prob": 3.4782416530321214e-11,
      "compression_ratio": 1.2669039145907472
    },
    {
      "start": 34.64,
      "end": 36.04,
      "text": "這聽起來就是",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.1757877823574091,
      "no_speech_prob": 3.4782416530321214e-11,
      "compression_ratio": 1.2669039145907472
    },
    {
      "start": 36.04,
      "end": 39.18,
      "text": "每天辦公室會發生的噩夢啊",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.1757877823574091,
      "no_speech_prob": 3.4782416530321214e-11,
      "compression_ratio": 1.2669039145907472
    },
    {
      "start": 39.18,
      "end": 40.26,
      "text": "沒錯沒錯",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.1757877823574091,
      "no_speech_prob": 3.4782416530321214e-11,
      "compression_ratio": 1.2669039145907472
    },
    {
      "start": 40.26,
      "end": 42.56,
      "text": "但這個時候如果你想像一下",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.1757877823574091,
      "no_speech_prob": 3.4782416530321214e-11,
      "compression_ratio": 1.2669039145907472
    },
    {
      "start": 42.56,
      "end": 43.58,
      "text": "如果有一個助理",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.1757877823574091,
      "no_speech_prob": 3.4782416530321214e-11,
      "compression_ratio": 1.2669039145907472
    },
    {
      "start": 43.58,
      "end": 45.6,
      "text": "能在你剛講完話的",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.1757877823574091,
      "no_speech_prob": 3.4782416530321214e-11,
      "compression_ratio": 1.2669039145907472
    },
    {
      "start": 45.6,
      "end": 48.06,
      "text": "呃0.16秒以內",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.1757877823574091,
      "no_speech_prob": 3.4782416530321214e-11,
      "compression_ratio": 1.2669039145907472
    },
    {
      "start": 48.06,
      "end": 50.34,
      "text": "立刻遞給你一份完美的逐字稿",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.1757877823574091,
      "no_speech_prob": 3.4782416530321214e-11,
      "compression_ratio": 1.2669039145907472
    },
    {
      "start": 50.34,
      "end": 51.88,
      "text": "0.16秒",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.1757877823574091,
      "no_speech_prob": 3.4782416530321214e-11,
      "compression_ratio": 1.2669039145907472
    },
    {
      "start": 51.88,
      "end": 53.5,
      "text": "這幾乎是同步了吧",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.1757877823574091,
      "no_speech_prob": 3.4782416530321214e-11,
      "compression_ratio": 1.2669039145907472
    },
    {
      "start": 53.5,
      "end": 54.08,
      "text": "對",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.1757877823574091,
      "no_speech_prob": 3.4782416530321214e-11,
      "compression_ratio": 1.2669039145907472
    },
    {
      "start": 54.08,
      "end": 56.28,
      "text": "而且這份逐字稿上面",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.1757877823574091,
      "no_speech_prob": 3.4782416530321214e-11,
      "compression_ratio": 1.2669039145907472
    },
    {
      "start": 56.28,
      "end": 58.46,
      "text": "還清清楚楚的標記了",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.1757877823574091,
      "no_speech_prob": 3.4782416530321214e-11,
      "compression_ratio": 1.2669039145907472
    },
    {
      "start": 58.46,
      "end": 60.62,
      "text": "刚刚哪一句话是谁说的",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.12001450233210147,
      "no_speech_prob": 1.9917878110731202e-11,
      "compression_ratio": 1.2448275862068965
    },
    {
      "start": 60.62,
      "end": 63.92,
      "text": "甚至连两个人同时开口讲话的内容",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.12001450233210147,
      "no_speech_prob": 1.9917878110731202e-11,
      "compression_ratio": 1.2448275862068965
    },
    {
      "start": 63.92,
      "end": 65.52,
      "text": "它都分得一清二楚",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.12001450233210147,
      "no_speech_prob": 1.9917878110731202e-11,
      "compression_ratio": 1.2448275862068965
    },
    {
      "start": 65.52,
      "end": 68.96,
      "text": "这听起来真的很像某种科幻电影里",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.12001450233210147,
      "no_speech_prob": 1.9917878110731202e-11,
      "compression_ratio": 1.2448275862068965
    },
    {
      "start": 68.96,
      "end": 70.3,
      "text": "才有的可怕科技对吧",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.12001450233210147,
      "no_speech_prob": 1.9917878110731202e-11,
      "compression_ratio": 1.2448275862068965
    },
    {
      "start": 70.3,
      "end": 70.66,
      "text": "真的",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.12001450233210147,
      "no_speech_prob": 1.9917878110731202e-11,
      "compression_ratio": 1.2448275862068965
    },
    {
      "start": 70.66,
      "end": 73.4,
      "text": "一般人听起来一定觉得这不可能啦",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.12001450233210147,
      "no_speech_prob": 1.9917878110731202e-11,
      "compression_ratio": 1.2448275862068965
    },
    {
      "start": 73.4,
      "end": 75.54,
      "text": "但其实这已经不是科幻情节了",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.12001450233210147,
      "no_speech_prob": 1.9917878110731202e-11,
      "compression_ratio": 1.2448275862068965
    },
    {
      "start": 75.54,
      "end": 77.24,
      "text": "它已经真实发生了",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.12001450233210147,
      "no_speech_prob": 1.9917878110731202e-11,
      "compression_ratio": 1.2448275862068965
    },
    {
      "start": 77.24,
      "end": 79.94,
      "text": "就在2026年9月1日",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.12001450233210147,
      "no_speech_prob": 1.9917878110731202e-11,
      "compression_ratio": 1.2448275862068965
    },
    {
      "start": 79.94,
      "end": 82.28,
      "text": "META的超级智慧实验室",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.12001450233210147,
      "no_speech_prob": 1.9917878110731202e-11,
      "compression_ratio": 1.2448275862068965
    },
    {
      "start": 82.28,
      "end": 84.1,
      "text": "也就是那个MSL",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.12001450233210147,
      "no_speech_prob": 1.9917878110731202e-11,
      "compression_ratio": 1.2448275862068965
    },
    {
      "start": 84.1,
      "end": 85.44,
      "text": "哦MSL",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.12001450233210147,
      "no_speech_prob": 1.9917878110731202e-11,
      "compression_ratio": 1.2448275862068965
    },
    {
      "start": 85.44,
      "end": 85.74,
      "text": "对",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.12001450233210147,
      "no_speech_prob": 1.9917878110731202e-11,
      "compression_ratio": 1.2448275862068965
    },
    {
      "start": 85.74,
      "end": 87.92,
      "text": "他们正式发布了最新的",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.09900450040508249,
      "no_speech_prob": 2.7000093133500158e-11,
      "compression_ratio": 1.207977207977208
    },
    {
      "start": 87.92,
      "end": 89.78,
      "text": "即时音讯感知模型",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.09900450040508249,
      "no_speech_prob": 2.7000093133500158e-11,
      "compression_ratio": 1.207977207977208
    },
    {
      "start": 89.78,
      "end": 92.04,
      "text": "叫做Muse Voice Transcribe",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.09900450040508249,
      "no_speech_prob": 2.7000093133500158e-11,
      "compression_ratio": 1.207977207977208
    },
    {
      "start": 92.04,
      "end": 93.64,
      "text": "Muse Voice Transcribe",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.09900450040508249,
      "no_speech_prob": 2.7000093133500158e-11,
      "compression_ratio": 1.207977207977208
    },
    {
      "start": 93.64,
      "end": 95.34,
      "text": "这名字听起来蛮厉害的",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.09900450040508249,
      "no_speech_prob": 2.7000093133500158e-11,
      "compression_ratio": 1.207977207977208
    },
    {
      "start": 95.34,
      "end": 96.32,
      "text": "这个技术啊",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.09900450040508249,
      "no_speech_prob": 2.7000093133500158e-11,
      "compression_ratio": 1.207977207977208
    },
    {
      "start": 96.32,
      "end": 98.86,
      "text": "它不仅做到了你刚刚描述的那种",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.09900450040508249,
      "no_speech_prob": 2.7000093133500158e-11,
      "compression_ratio": 1.207977207977208
    },
    {
      "start": 98.86,
      "end": 101.56,
      "text": "在极度混乱的场景下完美记录",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.09900450040508249,
      "no_speech_prob": 2.7000093133500158e-11,
      "compression_ratio": 1.207977207977208
    },
    {
      "start": 101.56,
      "end": 104.04,
      "text": "它甚至在多项基准测试里面",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.09900450040508249,
      "no_speech_prob": 2.7000093133500158e-11,
      "compression_ratio": 1.207977207977208
    },
    {
      "start": 104.04,
      "end": 107.48,
      "text": "直接击败了Google跟OpenAI的同级产品",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.09900450040508249,
      "no_speech_prob": 2.7000093133500158e-11,
      "compression_ratio": 1.207977207977208
    },
    {
      "start": 107.48,
      "end": 110.02,
      "text": "哇 直接打败两大巨头耶",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.09900450040508249,
      "no_speech_prob": 2.7000093133500158e-11,
      "compression_ratio": 1.207977207977208
    },
    {
      "start": 110.02,
      "end": 112.28,
      "text": "所以在这场为你量身打造的",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.09900450040508249,
      "no_speech_prob": 2.7000093133500158e-11,
      "compression_ratio": 1.207977207977208
    },
    {
      "start": 112.28,
      "end": 113.08,
      "text": "深度探跑中",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.09900450040508249,
      "no_speech_prob": 2.7000093133500158e-11,
      "compression_ratio": 1.207977207977208
    },
    {
      "start": 113.08,
      "end": 114.58,
      "text": "我们绝点不会只是你知道",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.09900450040508249,
      "no_speech_prob": 2.7000093133500158e-11,
      "compression_ratio": 1.207977207977208
    },
    {
      "start": 114.58,
      "end": 116.78,
      "text": "很无聊的去念那些声音的规格表",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10403878649295364,
      "no_speech_prob": 2.0442161852152196e-11,
      "compression_ratio": 1.321311475409836
    },
    {
      "start": 116.78,
      "end": 118.18,
      "text": "对那太无聊了",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10403878649295364,
      "no_speech_prob": 2.0442161852152196e-11,
      "compression_ratio": 1.321311475409836
    },
    {
      "start": 118.18,
      "end": 118.54,
      "text": "没错",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10403878649295364,
      "no_speech_prob": 2.0442161852152196e-11,
      "compression_ratio": 1.321311475409836
    },
    {
      "start": 118.54,
      "end": 120.46,
      "text": "我们会带你彻底拆解",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10403878649295364,
      "no_speech_prob": 2.0442161852152196e-11,
      "compression_ratio": 1.321311475409836
    },
    {
      "start": 120.46,
      "end": 122.88,
      "text": "这个打败众多科技巨头的核心技术",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10403878649295364,
      "no_speech_prob": 2.0442161852152196e-11,
      "compression_ratio": 1.321311475409836
    },
    {
      "start": 122.88,
      "end": 124.46,
      "text": "去了解这个AI",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10403878649295364,
      "no_speech_prob": 2.0442161852152196e-11,
      "compression_ratio": 1.321311475409836
    },
    {
      "start": 124.46,
      "end": 126.52,
      "text": "它是如何像我们人类大脑一样",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10403878649295364,
      "no_speech_prob": 2.0442161852152196e-11,
      "compression_ratio": 1.321311475409836
    },
    {
      "start": 126.52,
      "end": 129.82,
      "text": "懂得在听话的时候去犹豫",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10403878649295364,
      "no_speech_prob": 2.0442161852152196e-11,
      "compression_ratio": 1.321311475409836
    },
    {
      "start": 129.82,
      "end": 130.74,
      "text": "犹豫啊",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10403878649295364,
      "no_speech_prob": 2.0442161852152196e-11,
      "compression_ratio": 1.321311475409836
    },
    {
      "start": 130.74,
      "end": 131.9,
      "text": "这个词用的很好",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10403878649295364,
      "no_speech_prob": 2.0442161852152196e-11,
      "compression_ratio": 1.321311475409836
    },
    {
      "start": 131.9,
      "end": 132.38,
      "text": "对吧",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10403878649295364,
      "no_speech_prob": 2.0442161852152196e-11,
      "compression_ratio": 1.321311475409836
    },
    {
      "start": 132.38,
      "end": 132.92,
      "text": "对吧",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10403878649295364,
      "no_speech_prob": 2.0442161852152196e-11,
      "compression_ratio": 1.321311475409836
    },
    {
      "start": 132.92,
      "end": 136.36,
      "text": "然后这项技术将如何彻底颠覆你未来",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10403878649295364,
      "no_speech_prob": 2.0442161852152196e-11,
      "compression_ratio": 1.321311475409836
    },
    {
      "start": 136.36,
      "end": 138.52,
      "text": "跟科技装饰互动的方式",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10403878649295364,
      "no_speech_prob": 2.0442161852152196e-11,
      "compression_ratio": 1.321311475409836
    },
    {
      "start": 138.52,
      "end": 139.98,
      "text": "比如说你的Mac",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10403878649295364,
      "no_speech_prob": 2.0442161852152196e-11,
      "compression_ratio": 1.321311475409836
    },
    {
      "start": 139.98,
      "end": 143.46,
      "text": "甚至是未来的那个AR智慧眼镜",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10403878649295364,
      "no_speech_prob": 2.0442161852152196e-11,
      "compression_ratio": 1.321311475409836
    },
    {
      "start": 143.46,
      "end": 148.76,
      "text": "没错,不过其实要真正理解Muse Voice Transcribe到底有多强大",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10428308349808836,
      "no_speech_prob": 2.0217676491296466e-11,
      "compression_ratio": 1.2467105263157894
    },
    {
      "start": 148.76,
      "end": 150.3,
      "text": "我们必须先回头看看",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10428308349808836,
      "no_speech_prob": 2.0217676491296466e-11,
      "compression_ratio": 1.2467105263157894
    },
    {
      "start": 150.3,
      "end": 155.06,
      "text": "过去几年的语音辨识技术到底有多么的笨拙",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10428308349808836,
      "no_speech_prob": 2.0217676491296466e-11,
      "compression_ratio": 1.2467105263157894
    },
    {
      "start": 155.06,
      "end": 159.02,
      "text": "笨拙,所以过去的方法很卡就对了",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10428308349808836,
      "no_speech_prob": 2.0217676491296466e-11,
      "compression_ratio": 1.2467105263157894
    },
    {
      "start": 159.02,
      "end": 163.48,
      "text": "超级卡,我们得先看看过去的那种拼装车系统",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10428308349808836,
      "no_speech_prob": 2.0217676491296466e-11,
      "compression_ratio": 1.2467105263157894
    },
    {
      "start": 163.48,
      "end": 164.22,
      "text": "拼装车",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10428308349808836,
      "no_speech_prob": 2.0217676491296466e-11,
      "compression_ratio": 1.2467105263157894
    },
    {
      "start": 164.22,
      "end": 168.16,
      "text": "对,因为过去如果你想要完成一份会议的逐字稿",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10428308349808836,
      "no_speech_prob": 2.0217676491296466e-11,
      "compression_ratio": 1.2467105263157894
    },
    {
      "start": 168.16,
      "end": 171.22,
      "text": "它其实不是一个单一的人工智慧在运作",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10428308349808836,
      "no_speech_prob": 2.0217676491296466e-11,
      "compression_ratio": 1.2467105263157894
    },
    {
      "start": 171.22,
      "end": 175.06,
      "text": "它是需要三套完全独立的系统来接力完成",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.13653233472038717,
      "no_speech_prob": 1.1998602632290467e-11,
      "compression_ratio": 1.3288135593220338
    },
    {
      "start": 175.06,
      "end": 176.98,
      "text": "呃哪三套",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.13653233472038717,
      "no_speech_prob": 1.1998602632290467e-11,
      "compression_ratio": 1.3288135593220338
    },
    {
      "start": 176.98,
      "end": 178.72,
      "text": "首先是那个ASR",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.13653233472038717,
      "no_speech_prob": 1.1998602632290467e-11,
      "compression_ratio": 1.3288135593220338
    },
    {
      "start": 178.72,
      "end": 180.88,
      "text": "也就是自动语音辨识系统",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.13653233472038717,
      "no_speech_prob": 1.1998602632290467e-11,
      "compression_ratio": 1.3288135593220338
    },
    {
      "start": 180.88,
      "end": 182.42,
      "text": "它只负责一件事情",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.13653233472038717,
      "no_speech_prob": 1.1998602632290467e-11,
      "compression_ratio": 1.3288135593220338
    },
    {
      "start": 182.42,
      "end": 185.26,
      "text": "就是把听到的声音转成文字",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.13653233472038717,
      "no_speech_prob": 1.1998602632290467e-11,
      "compression_ratio": 1.3288135593220338
    },
    {
      "start": 185.26,
      "end": 186.94,
      "text": "嗯就单纯听写而已",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.13653233472038717,
      "no_speech_prob": 1.1998602632290467e-11,
      "compression_ratio": 1.3288135593220338
    },
    {
      "start": 186.94,
      "end": 187.4,
      "text": "没错",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.13653233472038717,
      "no_speech_prob": 1.1998602632290467e-11,
      "compression_ratio": 1.3288135593220338
    },
    {
      "start": 187.4,
      "end": 191.18,
      "text": "接着这个文字和音讯会被丢给Diarization",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.13653233472038717,
      "no_speech_prob": 1.1998602632290467e-11,
      "compression_ratio": 1.3288135593220338
    },
    {
      "start": 191.18,
      "end": 193.68,
      "text": "也就是说话者分离系统",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.13653233472038717,
      "no_speech_prob": 1.1998602632290467e-11,
      "compression_ratio": 1.3288135593220338
    },
    {
      "start": 193.68,
      "end": 195.2,
      "text": "等等说话者分离",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.13653233472038717,
      "no_speech_prob": 1.1998602632290467e-11,
      "compression_ratio": 1.3288135593220338
    },
    {
      "start": 195.2,
      "end": 198.42,
      "text": "这听起来就像是一个事后侦探也",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.13653233472038717,
      "no_speech_prob": 1.1998602632290467e-11,
      "compression_ratio": 1.3288135593220338
    },
    {
      "start": 198.42,
      "end": 199.2,
      "text": "对对对",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.13653233472038717,
      "no_speech_prob": 1.1998602632290467e-11,
      "compression_ratio": 1.3288135593220338
    },
    {
      "start": 199.2,
      "end": 201.16,
      "text": "它就是负责回头去分辨说",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.13653233472038717,
      "no_speech_prob": 1.1998602632290467e-11,
      "compression_ratio": 1.3288135593220338
    },
    {
      "start": 201.16,
      "end": 204.68,
      "text": "现在这句话到底是会议室里面的哪个人说的",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.09005676981914475,
      "no_speech_prob": 1.3550109818905032e-11,
      "compression_ratio": 1.294871794871795
    },
    {
      "start": 204.68,
      "end": 207.3,
      "text": "然后最后还需要过一个叫做",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.09005676981914475,
      "no_speech_prob": 1.3550109818905032e-11,
      "compression_ratio": 1.294871794871795
    },
    {
      "start": 207.3,
      "end": 209.7,
      "text": "endpointing的端点侦测系统",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.09005676981914475,
      "no_speech_prob": 1.3550109818905032e-11,
      "compression_ratio": 1.294871794871795
    },
    {
      "start": 209.7,
      "end": 210.76,
      "text": "端点侦测",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.09005676981914475,
      "no_speech_prob": 1.3550109818905032e-11,
      "compression_ratio": 1.294871794871795
    },
    {
      "start": 210.76,
      "end": 212.18,
      "text": "这又是做什么的",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.09005676981914475,
      "no_speech_prob": 1.3550109818905032e-11,
      "compression_ratio": 1.294871794871795
    },
    {
      "start": 212.18,
      "end": 213.86,
      "text": "这个就是用来判断说",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.09005676981914475,
      "no_speech_prob": 1.3550109818905032e-11,
      "compression_ratio": 1.294871794871795
    },
    {
      "start": 213.86,
      "end": 216.08,
      "text": "这个人到底讲完话了没有",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.09005676981914475,
      "no_speech_prob": 1.3550109818905032e-11,
      "compression_ratio": 1.294871794871795
    },
    {
      "start": 216.08,
      "end": 218.28,
      "text": "系统该不该切断麦克风",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.09005676981914475,
      "no_speech_prob": 1.3550109818905032e-11,
      "compression_ratio": 1.294871794871795
    },
    {
      "start": 218.28,
      "end": 220.42,
      "text": "哇这听起来就像是一条",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.09005676981914475,
      "no_speech_prob": 1.3550109818905032e-11,
      "compression_ratio": 1.294871794871795
    },
    {
      "start": 220.42,
      "end": 223.54,
      "text": "非常老旧而且缺乏效率的工厂管线",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.09005676981914475,
      "no_speech_prob": 1.3550109818905032e-11,
      "compression_ratio": 1.294871794871795
    },
    {
      "start": 223.54,
      "end": 225.96,
      "text": "真的非常贴切的比喻",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.09005676981914475,
      "no_speech_prob": 1.3550109818905032e-11,
      "compression_ratio": 1.294871794871795
    },
    {
      "start": 225.96,
      "end": 229.0,
      "text": "你可以想象就是三个工人站在输送袋旁边",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.09005676981914475,
      "no_speech_prob": 1.3550109818905032e-11,
      "compression_ratio": 1.294871794871795
    },
    {
      "start": 229.0,
      "end": 230.72,
      "text": "第一个工人只顾着",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.0903183004474113,
      "no_speech_prob": 1.5415483126113294e-11,
      "compression_ratio": 1.398671096345515
    },
    {
      "start": 230.72,
      "end": 232.44,
      "text": "听声音写字",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.0903183004474113,
      "no_speech_prob": 1.5415483126113294e-11,
      "compression_ratio": 1.398671096345515
    },
    {
      "start": 232.44,
      "end": 233.92,
      "text": "然后把写好的纸条",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.0903183004474113,
      "no_speech_prob": 1.5415483126113294e-11,
      "compression_ratio": 1.398671096345515
    },
    {
      "start": 233.92,
      "end": 235.42,
      "text": "赶快传给第二个工人",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.0903183004474113,
      "no_speech_prob": 1.5415483126113294e-11,
      "compression_ratio": 1.398671096345515
    },
    {
      "start": 235.42,
      "end": 236.56,
      "text": "然后第二个工人",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.0903183004474113,
      "no_speech_prob": 1.5415483126113294e-11,
      "compression_ratio": 1.398671096345515
    },
    {
      "start": 236.56,
      "end": 237.84,
      "text": "再去对比声音",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.0903183004474113,
      "no_speech_prob": 1.5415483126113294e-11,
      "compression_ratio": 1.398671096345515
    },
    {
      "start": 237.84,
      "end": 239.26,
      "text": "猜这到底是谁说的",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.0903183004474113,
      "no_speech_prob": 1.5415483126113294e-11,
      "compression_ratio": 1.398671096345515
    },
    {
      "start": 239.26,
      "end": 241.28,
      "text": "最后再传给第三个工人",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.0903183004474113,
      "no_speech_prob": 1.5415483126113294e-11,
      "compression_ratio": 1.398671096345515
    },
    {
      "start": 241.28,
      "end": 242.36,
      "text": "去决定这句话",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.0903183004474113,
      "no_speech_prob": 1.5415483126113294e-11,
      "compression_ratio": 1.398671096345515
    },
    {
      "start": 242.36,
      "end": 243.58,
      "text": "是不是真的结束了",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.0903183004474113,
      "no_speech_prob": 1.5415483126113294e-11,
      "compression_ratio": 1.398671096345515
    },
    {
      "start": 243.58,
      "end": 244.06,
      "text": "没错",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.0903183004474113,
      "no_speech_prob": 1.5415483126113294e-11,
      "compression_ratio": 1.398671096345515
    },
    {
      "start": 244.06,
      "end": 245.3,
      "text": "就是这么繁琐",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.0903183004474113,
      "no_speech_prob": 1.5415483126113294e-11,
      "compression_ratio": 1.398671096345515
    },
    {
      "start": 245.3,
      "end": 247.44,
      "text": "而且只要其中一个环节出错",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.0903183004474113,
      "no_speech_prob": 1.5415483126113294e-11,
      "compression_ratio": 1.398671096345515
    },
    {
      "start": 247.44,
      "end": 249.72,
      "text": "整个流程就会大塞车",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.0903183004474113,
      "no_speech_prob": 1.5415483126113294e-11,
      "compression_ratio": 1.398671096345515
    },
    {
      "start": 249.72,
      "end": 250.14,
      "text": "真的",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.0903183004474113,
      "no_speech_prob": 1.5415483126113294e-11,
      "compression_ratio": 1.398671096345515
    },
    {
      "start": 250.14,
      "end": 251.28,
      "text": "我之前用过一些",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.0903183004474113,
      "no_speech_prob": 1.5415483126113294e-11,
      "compression_ratio": 1.398671096345515
    },
    {
      "start": 251.28,
      "end": 252.4,
      "text": "会议记录软体",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.0903183004474113,
      "no_speech_prob": 1.5415483126113294e-11,
      "compression_ratio": 1.398671096345515
    },
    {
      "start": 252.4,
      "end": 254.24,
      "text": "最常发生的惨剧就是",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.0903183004474113,
      "no_speech_prob": 1.5415483126113294e-11,
      "compression_ratio": 1.398671096345515
    },
    {
      "start": 254.24,
      "end": 255.0,
      "text": "那个",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.0903183004474113,
      "no_speech_prob": 1.5415483126113294e-11,
      "compression_ratio": 1.398671096345515
    },
    {
      "start": 255.0,
      "end": 257.68,
      "text": "A老板的话才刚讲到一半",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.0903183004474113,
      "no_speech_prob": 1.5415483126113294e-11,
      "compression_ratio": 1.398671096345515
    },
    {
      "start": 257.68,
      "end": 260.6,
      "text": "然后旁边B员工稍微咳嗽了一声",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.07808335622151692,
      "no_speech_prob": 1.2537863108841307e-11,
      "compression_ratio": 1.294314381270903
    },
    {
      "start": 260.6,
      "end": 262.58,
      "text": "哦对系统就切断了",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.07808335622151692,
      "no_speech_prob": 1.2537863108841307e-11,
      "compression_ratio": 1.294314381270903
    },
    {
      "start": 262.58,
      "end": 264.86,
      "text": "没错系统就以为A讲完了",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.07808335622151692,
      "no_speech_prob": 1.2537863108841307e-11,
      "compression_ratio": 1.294314381270903
    },
    {
      "start": 264.86,
      "end": 267.16,
      "text": "直接把句子无情切断",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.07808335622151692,
      "no_speech_prob": 1.2537863108841307e-11,
      "compression_ratio": 1.294314381270903
    },
    {
      "start": 267.16,
      "end": 269.24,
      "text": "然后还把后面半句算在B头上",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.07808335622151692,
      "no_speech_prob": 1.2537863108841307e-11,
      "compression_ratio": 1.294314381270903
    },
    {
      "start": 269.24,
      "end": 270.62,
      "text": "这超尴尬的",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.07808335622151692,
      "no_speech_prob": 1.2537863108841307e-11,
      "compression_ratio": 1.294314381270903
    },
    {
      "start": 270.62,
      "end": 272.74,
      "text": "这真的是传统架构的致命伤",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.07808335622151692,
      "no_speech_prob": 1.2537863108841307e-11,
      "compression_ratio": 1.294314381270903
    },
    {
      "start": 272.74,
      "end": 276.7,
      "text": "过去这种做法通常需要大量的事后处理",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.07808335622151692,
      "no_speech_prob": 1.2537863108841307e-11,
      "compression_ratio": 1.294314381270903
    },
    {
      "start": 276.7,
      "end": 278.7,
      "text": "也就是等你整段话都录完了",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.07808335622151692,
      "no_speech_prob": 1.2537863108841307e-11,
      "compression_ratio": 1.294314381270903
    },
    {
      "start": 278.7,
      "end": 281.64,
      "text": "系统才能开始跑这些复杂的对比流程",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.07808335622151692,
      "no_speech_prob": 1.2537863108841307e-11,
      "compression_ratio": 1.294314381270903
    },
    {
      "start": 281.64,
      "end": 284.6,
      "text": "所以根本不可能做到真正的及时",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.07808335622151692,
      "no_speech_prob": 1.2537863108841307e-11,
      "compression_ratio": 1.294314381270903
    },
    {
      "start": 284.6,
      "end": 288.78,
      "text": "所以Meta这次的做法是怎么解决这个问题的",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.08817558428820442,
      "no_speech_prob": 1.7428562065768816e-11,
      "compression_ratio": 1.2872340425531914
    },
    {
      "start": 288.78,
      "end": 291.88,
      "text": "他们这次的做法可以说是破坏性的创新",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.08817558428820442,
      "no_speech_prob": 1.7428562065768816e-11,
      "compression_ratio": 1.2872340425531914
    },
    {
      "start": 291.88,
      "end": 296.56,
      "text": "他们把三个工人的工作全部交给了一个超级大脑同时完成",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.08817558428820442,
      "no_speech_prob": 1.7428562065768816e-11,
      "compression_ratio": 1.2872340425531914
    },
    {
      "start": 296.56,
      "end": 298.44,
      "text": "一个大脑全包了",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.08817558428820442,
      "no_speech_prob": 1.7428562065768816e-11,
      "compression_ratio": 1.2872340425531914
    },
    {
      "start": 298.44,
      "end": 298.66,
      "text": "对",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.08817558428820442,
      "no_speech_prob": 1.7428562065768816e-11,
      "compression_ratio": 1.2872340425531914
    },
    {
      "start": 298.66,
      "end": 300.24,
      "text": "Muse Voice Transcribe",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.08817558428820442,
      "no_speech_prob": 1.7428562065768816e-11,
      "compression_ratio": 1.2872340425531914
    },
    {
      "start": 300.24,
      "end": 307.08,
      "text": "它其实是建构在Meta的Muse Spark家族之上的一款自回归多模态模型",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.08817558428820442,
      "no_speech_prob": 1.7428562065768816e-11,
      "compression_ratio": 1.2872340425531914
    },
    {
      "start": 307.08,
      "end": 308.6,
      "text": "等一下先暂停一下",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.08817558428820442,
      "no_speech_prob": 1.7428562065768816e-11,
      "compression_ratio": 1.2872340425531914
    },
    {
      "start": 308.6,
      "end": 311.38,
      "text": "那个自回归多模态模型",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.08817558428820442,
      "no_speech_prob": 1.7428562065768816e-11,
      "compression_ratio": 1.2872340425531914
    },
    {
      "start": 311.38,
      "end": 315.12,
      "text": "这几个字听起来真的很像某种外星语言也",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.12893034861637995,
      "no_speech_prob": 1.149757806323537e-11,
      "compression_ratio": 1.2805280528052805
    },
    {
      "start": 315.12,
      "end": 318.16,
      "text": "如果我们要用最白话的方式来理解",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.12893034861637995,
      "no_speech_prob": 1.149757806323537e-11,
      "compression_ratio": 1.2805280528052805
    },
    {
      "start": 318.16,
      "end": 319.78,
      "text": "这到底是什么意思",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.12893034861637995,
      "no_speech_prob": 1.149757806323537e-11,
      "compression_ratio": 1.2805280528052805
    },
    {
      "start": 319.78,
      "end": 321.14,
      "text": "简单来说了",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.12893034861637995,
      "no_speech_prob": 1.149757806323537e-11,
      "compression_ratio": 1.2805280528052805
    },
    {
      "start": 321.14,
      "end": 322.58,
      "text": "多摩泰意味着",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.12893034861637995,
      "no_speech_prob": 1.149757806323537e-11,
      "compression_ratio": 1.2805280528052805
    },
    {
      "start": 322.58,
      "end": 325.38,
      "text": "这个模型它可以同时看懂文字",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.12893034861637995,
      "no_speech_prob": 1.149757806323537e-11,
      "compression_ratio": 1.2805280528052805
    },
    {
      "start": 325.38,
      "end": 326.92,
      "text": "也能听懂声音",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.12893034861637995,
      "no_speech_prob": 1.149757806323537e-11,
      "compression_ratio": 1.2805280528052805
    },
    {
      "start": 326.92,
      "end": 329.66,
      "text": "它不是只吃单一一种资料",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.12893034861637995,
      "no_speech_prob": 1.149757806323537e-11,
      "compression_ratio": 1.2805280528052805
    },
    {
      "start": 329.66,
      "end": 333.3,
      "text": "哦 就是眼观四面耳听八方的感觉",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.12893034861637995,
      "no_speech_prob": 1.149757806323537e-11,
      "compression_ratio": 1.2805280528052805
    },
    {
      "start": 333.3,
      "end": 333.72,
      "text": "没错",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.12893034861637995,
      "no_speech_prob": 1.149757806323537e-11,
      "compression_ratio": 1.2805280528052805
    },
    {
      "start": 333.72,
      "end": 335.1,
      "text": "那自回归的话",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.12893034861637995,
      "no_speech_prob": 1.149757806323537e-11,
      "compression_ratio": 1.2805280528052805
    },
    {
      "start": 335.1,
      "end": 337.26,
      "text": "就像是我们在玩文字接龙一样",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.12893034861637995,
      "no_speech_prob": 1.149757806323537e-11,
      "compression_ratio": 1.2805280528052805
    },
    {
      "start": 337.26,
      "end": 339.72,
      "text": "它会根据已经接收到的资讯",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.12893034861637995,
      "no_speech_prob": 1.149757806323537e-11,
      "compression_ratio": 1.2805280528052805
    },
    {
      "start": 339.72,
      "end": 341.94,
      "text": "不断去预测下一步是什么",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10895309235130608,
      "no_speech_prob": 1.2297131195271316e-11,
      "compression_ratio": 1.2403560830860534
    },
    {
      "start": 341.94,
      "end": 344.5,
      "text": "但最关键的突破在于",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10895309235130608,
      "no_speech_prob": 1.2297131195271316e-11,
      "compression_ratio": 1.2403560830860534
    },
    {
      "start": 344.5,
      "end": 347.06,
      "text": "Meta在同一个音讯串流里面",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10895309235130608,
      "no_speech_prob": 1.2297131195271316e-11,
      "compression_ratio": 1.2403560830860534
    },
    {
      "start": 347.06,
      "end": 349.12,
      "text": "直接引入了特殊的标记",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10895309235130608,
      "no_speech_prob": 1.2297131195271316e-11,
      "compression_ratio": 1.2403560830860534
    },
    {
      "start": 349.12,
      "end": 350.42,
      "text": "也就是TOKENS",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10895309235130608,
      "no_speech_prob": 1.2297131195271316e-11,
      "compression_ratio": 1.2403560830860534
    },
    {
      "start": 350.42,
      "end": 352.82,
      "text": "这完全打破了我们刚才说的",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10895309235130608,
      "no_speech_prob": 1.2297131195271316e-11,
      "compression_ratio": 1.2403560830860534
    },
    {
      "start": 352.82,
      "end": 354.34,
      "text": "那个三阶段管线",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10895309235130608,
      "no_speech_prob": 1.2297131195271316e-11,
      "compression_ratio": 1.2403560830860534
    },
    {
      "start": 354.34,
      "end": 354.92,
      "text": "标记",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10895309235130608,
      "no_speech_prob": 1.2297131195271316e-11,
      "compression_ratio": 1.2403560830860534
    },
    {
      "start": 354.92,
      "end": 357.34,
      "text": "在在实际运作的画面上",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10895309235130608,
      "no_speech_prob": 1.2297131195271316e-11,
      "compression_ratio": 1.2403560830860534
    },
    {
      "start": 357.34,
      "end": 358.84,
      "text": "到底会长什么样子",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10895309235130608,
      "no_speech_prob": 1.2297131195271316e-11,
      "compression_ratio": 1.2403560830860534
    },
    {
      "start": 358.84,
      "end": 359.86,
      "text": "你想象一下",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10895309235130608,
      "no_speech_prob": 1.2297131195271316e-11,
      "compression_ratio": 1.2403560830860534
    },
    {
      "start": 359.86,
      "end": 361.56,
      "text": "有一条连续的时间轴",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10895309235130608,
      "no_speech_prob": 1.2297131195271316e-11,
      "compression_ratio": 1.2403560830860534
    },
    {
      "start": 361.56,
      "end": 363.76,
      "text": "当模型在及时聆听的时候",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10895309235130608,
      "no_speech_prob": 1.2297131195271316e-11,
      "compression_ratio": 1.2403560830860534
    },
    {
      "start": 363.76,
      "end": 365.86,
      "text": "只要它侦测到换人说话了",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10895309235130608,
      "no_speech_prob": 1.2297131195271316e-11,
      "compression_ratio": 1.2403560830860534
    },
    {
      "start": 365.86,
      "end": 368.02,
      "text": "它不会去呼叫别的系统来帮忙",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10895309235130608,
      "no_speech_prob": 1.2297131195271316e-11,
      "compression_ratio": 1.2403560830860534
    },
    {
      "start": 368.02,
      "end": 368.9,
      "text": "那它怎么做",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10895309235130608,
      "no_speech_prob": 1.2297131195271316e-11,
      "compression_ratio": 1.2403560830860534
    },
    {
      "start": 368.9,
      "end": 371.94,
      "text": "它会直接在自己生成的文字流里面",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.12034097545875047,
      "no_speech_prob": 1.4017686317258082e-11,
      "compression_ratio": 1.30625
    },
    {
      "start": 371.94,
      "end": 373.6,
      "text": "插入一个特殊的标记",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.12034097545875047,
      "no_speech_prob": 1.4017686317258082e-11,
      "compression_ratio": 1.30625
    },
    {
      "start": 373.6,
      "end": 375.28,
      "text": "叫做Start of Turn",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.12034097545875047,
      "no_speech_prob": 1.4017686317258082e-11,
      "compression_ratio": 1.30625
    },
    {
      "start": 375.28,
      "end": 377.38,
      "text": "也就是换人说话的标记",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.12034097545875047,
      "no_speech_prob": 1.4017686317258082e-11,
      "compression_ratio": 1.30625
    },
    {
      "start": 377.38,
      "end": 378.98,
      "text": "然后紧接着标注",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.12034097545875047,
      "no_speech_prob": 1.4017686317258082e-11,
      "compression_ratio": 1.30625
    },
    {
      "start": 378.98,
      "end": 381.26,
      "text": "这句话是Speaker A说的",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.12034097545875047,
      "no_speech_prob": 1.4017686317258082e-11,
      "compression_ratio": 1.30625
    },
    {
      "start": 381.26,
      "end": 383.86,
      "text": "哇 直接在同一个时间轴上标出来",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.12034097545875047,
      "no_speech_prob": 1.4017686317258082e-11,
      "compression_ratio": 1.30625
    },
    {
      "start": 383.86,
      "end": 384.82,
      "text": "对对对",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.12034097545875047,
      "no_speech_prob": 1.4017686317258082e-11,
      "compression_ratio": 1.30625
    },
    {
      "start": 384.82,
      "end": 387.58,
      "text": "同样的 刚你停顿或者讲完的时候",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.12034097545875047,
      "no_speech_prob": 1.4017686317258082e-11,
      "compression_ratio": 1.30625
    },
    {
      "start": 387.58,
      "end": 391.66,
      "text": "它也会立刻放入一个Speech Endpoint的语音结束标记",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.12034097545875047,
      "no_speech_prob": 1.4017686317258082e-11,
      "compression_ratio": 1.30625
    },
    {
      "start": 391.66,
      "end": 393.96,
      "text": "等于说所有的文字内容",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.12034097545875047,
      "no_speech_prob": 1.4017686317258082e-11,
      "compression_ratio": 1.30625
    },
    {
      "start": 393.96,
      "end": 396.58,
      "text": "还有身份辨识 语气停顿",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.12034097545875047,
      "no_speech_prob": 1.4017686317258082e-11,
      "compression_ratio": 1.30625
    },
    {
      "start": 396.58,
      "end": 398.66,
      "text": "全都在同一个时间轴上",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.12034097545875047,
      "no_speech_prob": 1.4017686317258082e-11,
      "compression_ratio": 1.30625
    },
    {
      "start": 398.66,
      "end": 401.82,
      "text": "由这个单一模型一次性全部吐出来",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.08814321517944336,
      "no_speech_prob": 1.6402188635078474e-11,
      "compression_ratio": 1.3215434083601285
    },
    {
      "start": 401.82,
      "end": 403.02,
      "text": "没有错就是这样",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.08814321517944336,
      "no_speech_prob": 1.6402188635078474e-11,
      "compression_ratio": 1.3215434083601285
    },
    {
      "start": 403.02,
      "end": 408.42,
      "text": "哇所以这代表着我们再也不用忍受那种很尴尬的语音输入时刻了耶",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.08814321517944336,
      "no_speech_prob": 1.6402188635078474e-11,
      "compression_ratio": 1.3215434083601285
    },
    {
      "start": 408.42,
      "end": 412.44,
      "text": "你们一定有碰过平时对着手机讲话稍微停顿想个词",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.08814321517944336,
      "no_speech_prob": 1.6402188635078474e-11,
      "compression_ratio": 1.3215434083601285
    },
    {
      "start": 412.44,
      "end": 414.06,
      "text": "系统就以为你讲完了",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.08814321517944336,
      "no_speech_prob": 1.6402188635078474e-11,
      "compression_ratio": 1.3215434083601285
    },
    {
      "start": 414.06,
      "end": 416.12,
      "text": "对然后就无情的把你切断",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.08814321517944336,
      "no_speech_prob": 1.6402188635078474e-11,
      "compression_ratio": 1.3215434083601285
    },
    {
      "start": 416.12,
      "end": 420.84,
      "text": "没错现在这个模型它能真正理解对话的节奏",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.08814321517944336,
      "no_speech_prob": 1.6402188635078474e-11,
      "compression_ratio": 1.3215434083601285
    },
    {
      "start": 420.84,
      "end": 426.2,
      "text": "它知道你只是在脑袋里面那个卡词了而不是真的结束发言",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.08814321517944336,
      "no_speech_prob": 1.6402188635078474e-11,
      "compression_ratio": 1.3215434083601285
    },
    {
      "start": 426.2,
      "end": 430.32,
      "text": "没错,这正是单一串流模型带来的巨大优势",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.09676418716101337,
      "no_speech_prob": 1.0581220795391477e-11,
      "compression_ratio": 1.2659176029962547
    },
    {
      "start": 430.32,
      "end": 435.78,
      "text": "但其实,这背后隐藏着一个极度矛盾的工程难题",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.09676418716101337,
      "no_speech_prob": 1.0581220795391477e-11,
      "compression_ratio": 1.2659176029962547
    },
    {
      "start": 435.78,
      "end": 437.02,
      "text": "怎么说?",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.09676418716101337,
      "no_speech_prob": 1.0581220795391477e-11,
      "compression_ratio": 1.2659176029962547
    },
    {
      "start": 437.48,
      "end": 439.32,
      "text": "既然这是一个超级模型",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.09676418716101337,
      "no_speech_prob": 1.0581220795391477e-11,
      "compression_ratio": 1.2659176029962547
    },
    {
      "start": 439.32,
      "end": 444.14,
      "text": "要及时处理所有的声音、身份和文字",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.09676418716101337,
      "no_speech_prob": 1.0581220795391477e-11,
      "compression_ratio": 1.2659176029962547
    },
    {
      "start": 444.14,
      "end": 447.98,
      "text": "那他怎么知道什么时候该把文字吐出来?",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.09676418716101337,
      "no_speech_prob": 1.0581220795391477e-11,
      "compression_ratio": 1.2659176029962547
    },
    {
      "start": 448.46,
      "end": 449.34,
      "text": "哦,对耶",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.09676418716101337,
      "no_speech_prob": 1.0581220795391477e-11,
      "compression_ratio": 1.2659176029962547
    },
    {
      "start": 449.34,
      "end": 452.1,
      "text": "如果他一听到声音就立刻输出文字",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.09676418716101337,
      "no_speech_prob": 1.0581220795391477e-11,
      "compression_ratio": 1.2659176029962547
    },
    {
      "start": 452.1,
      "end": 454.28,
      "text": "那肯定很容易听错嘛",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.09676418716101337,
      "no_speech_prob": 1.0581220795391477e-11,
      "compression_ratio": 1.2659176029962547
    },
    {
      "start": 454.28,
      "end": 456.68,
      "text": "但如果他为了追求完美",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.15502251805485906,
      "no_speech_prob": 1.5438360659314476e-11,
      "compression_ratio": 1.2992424242424243
    },
    {
      "start": 456.68,
      "end": 459.22,
      "text": "等听完整句话才输出",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.15502251805485906,
      "no_speech_prob": 1.5438360659314476e-11,
      "compression_ratio": 1.2992424242424243
    },
    {
      "start": 459.22,
      "end": 461.44,
      "text": "那又失去了我们刚才说的",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.15502251805485906,
      "no_speech_prob": 1.5438360659314476e-11,
      "compression_ratio": 1.2992424242424243
    },
    {
      "start": 461.44,
      "end": 462.72,
      "text": "及时的意义",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.15502251805485906,
      "no_speech_prob": 1.5438360659314476e-11,
      "compression_ratio": 1.2992424242424243
    },
    {
      "start": 462.72,
      "end": 463.76,
      "text": "这真的很矛盾",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.15502251805485906,
      "no_speech_prob": 1.5438360659314476e-11,
      "compression_ratio": 1.2992424242424243
    },
    {
      "start": 463.76,
      "end": 466.28,
      "text": "这就像我们人类大脑",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.15502251805485906,
      "no_speech_prob": 1.5438360659314476e-11,
      "compression_ratio": 1.2992424242424243
    },
    {
      "start": 466.28,
      "end": 467.84,
      "text": "在听别人说话一样",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.15502251805485906,
      "no_speech_prob": 1.5438360659314476e-11,
      "compression_ratio": 1.2992424242424243
    },
    {
      "start": 467.84,
      "end": 468.92,
      "text": "你想想看",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.15502251805485906,
      "no_speech_prob": 1.5438360659314476e-11,
      "compression_ratio": 1.2992424242424243
    },
    {
      "start": 468.92,
      "end": 470.72,
      "text": "如果有人对我说苹果",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.15502251805485906,
      "no_speech_prob": 1.5438360659314476e-11,
      "compression_ratio": 1.2992424242424243
    },
    {
      "start": 470.72,
      "end": 472.82,
      "text": "就很简单的日常词汇",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.15502251805485906,
      "no_speech_prob": 1.5438360659314476e-11,
      "compression_ratio": 1.2992424242424243
    },
    {
      "start": 472.82,
      "end": 475.24,
      "text": "我的大脑几乎不需要思考",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.15502251805485906,
      "no_speech_prob": 1.5438360659314476e-11,
      "compression_ratio": 1.2992424242424243
    },
    {
      "start": 475.24,
      "end": 476.52,
      "text": "马上就能写下来",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.15502251805485906,
      "no_speech_prob": 1.5438360659314476e-11,
      "compression_ratio": 1.2992424242424243
    },
    {
      "start": 476.52,
      "end": 477.38,
      "text": "对 瞄懂",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.15502251805485906,
      "no_speech_prob": 1.5438360659314476e-11,
      "compression_ratio": 1.2992424242424243
    },
    {
      "start": 477.38,
      "end": 478.7,
      "text": "但如果今天",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.15502251805485906,
      "no_speech_prob": 1.5438360659314476e-11,
      "compression_ratio": 1.2992424242424243
    },
    {
      "start": 478.7,
      "end": 480.0,
      "text": "对方正在讲一段话",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.15502251805485906,
      "no_speech_prob": 1.5438360659314476e-11,
      "compression_ratio": 1.2992424242424243
    },
    {
      "start": 480.0,
      "end": 490.24,
      "text": "但如果今天,對方正在講一個超級複雜的醫學專有名詞,或者是他講話含糊不清,甚至口音很重,我的大腦其實會稍微,呃,猶豫個零點幾秒。",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.13920612084238151,
      "no_speech_prob": 9.553661681205305e-12,
      "compression_ratio": 1.3076923076923077
    },
    {
      "start": 490.24,
      "end": 491.6,
      "text": "沒錯,會卡一下。",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.13920612084238151,
      "no_speech_prob": 9.553661681205305e-12,
      "compression_ratio": 1.3076923076923077
    },
    {
      "start": 491.6,
      "end": 499.82,
      "text": "对,就是要等它把所有的音节都发音完整了,然后我在脑海中拼错确认之后,我才敢写下来。",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.07425776035827998,
      "no_speech_prob": 1.2936205058544648e-11,
      "compression_ratio": 1.2810457516339868
    },
    {
      "start": 500.22,
      "end": 504.04,
      "text": "这个AI模型难道也会模仿人类的这种犹豫吗?",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.07425776035827998,
      "no_speech_prob": 1.2936205058544648e-11,
      "compression_ratio": 1.2810457516339868
    },
    {
      "start": 504.36,
      "end": 506.74,
      "text": "这个人类大脑的类比非常到位耶。",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.07425776035827998,
      "no_speech_prob": 1.2936205058544648e-11,
      "compression_ratio": 1.2810457516339868
    },
    {
      "start": 507.24,
      "end": 511.16,
      "text": "其实META的工程师就是循着这个逻辑去设计的。",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.07425776035827998,
      "no_speech_prob": 1.2936205058544648e-11,
      "compression_ratio": 1.2810457516339868
    },
    {
      "start": 511.68,
      "end": 516.8,
      "text": "他们在技术摆笔书里面把这个称之为自适应延迟技术。",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.07425776035827998,
      "no_speech_prob": 1.2936205058544648e-11,
      "compression_ratio": 1.2810457516339868
    },
    {
      "start": 517.16,
      "end": 520.34,
      "text": "自适应延迟,听起来蛮高级的。",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.07425776035827998,
      "no_speech_prob": 1.2936205058544648e-11,
      "compression_ratio": 1.2810457516339868
    },
    {
      "start": 520.34,
      "end": 530.52,
      "text": "为了做到这点,模型在接收音讯的时候,它会把声音切成极小极小的区块,精确来说是80毫秒一个区块。",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.07061558038416044,
      "no_speech_prob": 1.0481384857263798e-11,
      "compression_ratio": 1.3588709677419355
    },
    {
      "start": 530.76,
      "end": 533.02,
      "text": "80毫秒,这也太短了吧?",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.07061558038416044,
      "no_speech_prob": 1.0481384857263798e-11,
      "compression_ratio": 1.3588709677419355
    },
    {
      "start": 533.46,
      "end": 539.98,
      "text": "超短的,也就是说在一秒钟之内,模型会接收到12.5个音讯区块。",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.07061558038416044,
      "no_speech_prob": 1.0481384857263798e-11,
      "compression_ratio": 1.3588709677419355
    },
    {
      "start": 540.66,
      "end": 546.16,
      "text": "而在接收到每一个80毫秒区块的当下,模型都必须做出一个决定。",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.07061558038416044,
      "no_speech_prob": 1.0481384857263798e-11,
      "compression_ratio": 1.3588709677419355
    },
    {
      "start": 546.34,
      "end": 547.02,
      "text": "什么决动?",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.07061558038416044,
      "no_speech_prob": 1.0481384857263798e-11,
      "compression_ratio": 1.3588709677419355
    },
    {
      "start": 547.02,
      "end": 550.2,
      "text": "就是我是要现在立刻输出文字",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.11046802834288715,
      "no_speech_prob": 1.733228664757558e-11,
      "compression_ratio": 1.2610169491525425
    },
    {
      "start": 550.2,
      "end": 554.24,
      "text": "还是我要输出一个特殊的Next Audio标记",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.11046802834288715,
      "no_speech_prob": 1.733228664757558e-11,
      "compression_ratio": 1.2610169491525425
    },
    {
      "start": 554.24,
      "end": 555.6,
      "text": "告诉自己说",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.11046802834288715,
      "no_speech_prob": 1.733228664757558e-11,
      "compression_ratio": 1.2610169491525425
    },
    {
      "start": 555.6,
      "end": 558.34,
      "text": "我需要再多听下一个音讯巨块",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.11046802834288715,
      "no_speech_prob": 1.733228664757558e-11,
      "compression_ratio": 1.2610169491525425
    },
    {
      "start": 558.34,
      "end": 559.42,
      "text": "这很酷耶",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.11046802834288715,
      "no_speech_prob": 1.733228664757558e-11,
      "compression_ratio": 1.2610169491525425
    },
    {
      "start": 559.42,
      "end": 561.9,
      "text": "但他怎么知道自己该不该等啊",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.11046802834288715,
      "no_speech_prob": 1.733228664757558e-11,
      "compression_ratio": 1.2610169491525425
    },
    {
      "start": 561.9,
      "end": 566.24,
      "text": "工程师到底是怎么教懂一个AI学会去内心的呢",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.11046802834288715,
      "no_speech_prob": 1.733228664757558e-11,
      "compression_ratio": 1.2610169491525425
    },
    {
      "start": 566.24,
      "end": 569.24,
      "text": "这里的关键就在于强化学习",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.11046802834288715,
      "no_speech_prob": 1.733228664757558e-11,
      "compression_ratio": 1.2610169491525425
    },
    {
      "start": 569.24,
      "end": 571.18,
      "text": "工程师在训练它的时候",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.11046802834288715,
      "no_speech_prob": 1.733228664757558e-11,
      "compression_ratio": 1.2610169491525425
    },
    {
      "start": 571.18,
      "end": 574.24,
      "text": "设计了一个非常严苛的奖励机制",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.11046802834288715,
      "no_speech_prob": 1.733228664757558e-11,
      "compression_ratio": 1.2610169491525425
    },
    {
      "start": 574.24,
      "end": 575.28,
      "text": "有多严苛",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.11046802834288715,
      "no_speech_prob": 1.733228664757558e-11,
      "compression_ratio": 1.2610169491525425
    },
    {
      "start": 575.28,
      "end": 579.9,
      "text": "他们把那个字错率的惩罚跟延迟时间的惩罚",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.11697645823160807,
      "no_speech_prob": 1.4908536211111212e-11,
      "compression_ratio": 1.3214285714285714
    },
    {
      "start": 579.9,
      "end": 581.86,
      "text": "用惩罚结合在一起",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.11697645823160807,
      "no_speech_prob": 1.4908536211111212e-11,
      "compression_ratio": 1.3214285714285714
    },
    {
      "start": 581.86,
      "end": 584.48,
      "text": "你可以想象这是一场极限挑战",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.11697645823160807,
      "no_speech_prob": 1.4908536211111212e-11,
      "compression_ratio": 1.3214285714285714
    },
    {
      "start": 584.48,
      "end": 586.28,
      "text": "如果模型当个快手",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.11697645823160807,
      "no_speech_prob": 1.4908536211111212e-11,
      "compression_ratio": 1.3214285714285714
    },
    {
      "start": 586.28,
      "end": 587.8,
      "text": "一听到声音就乱拆",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.11697645823160807,
      "no_speech_prob": 1.4908536211111212e-11,
      "compression_ratio": 1.3214285714285714
    },
    {
      "start": 587.8,
      "end": 589.26,
      "text": "导致疯狂写错字",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.11697645823160807,
      "no_speech_prob": 1.4908536211111212e-11,
      "compression_ratio": 1.3214285714285714
    },
    {
      "start": 589.26,
      "end": 590.56,
      "text": "他就会被扣分",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.11697645823160807,
      "no_speech_prob": 1.4908536211111212e-11,
      "compression_ratio": 1.3214285714285714
    },
    {
      "start": 590.56,
      "end": 593.14,
      "text": "那如果他想当个腕面主义者呢",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.11697645823160807,
      "no_speech_prob": 1.4908536211111212e-11,
      "compression_ratio": 1.3214285714285714
    },
    {
      "start": 593.14,
      "end": 593.74,
      "text": "哇",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.11697645823160807,
      "no_speech_prob": 1.4908536211111212e-11,
      "compression_ratio": 1.3214285714285714
    },
    {
      "start": 593.74,
      "end": 598.08,
      "text": "如果他为了一个字等了两三秒才输说文字",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.11697645823160807,
      "no_speech_prob": 1.4908536211111212e-11,
      "compression_ratio": 1.3214285714285714
    },
    {
      "start": 598.08,
      "end": 600.28,
      "text": "他一样会被狠狠扣分",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.11697645823160807,
      "no_speech_prob": 1.4908536211111212e-11,
      "compression_ratio": 1.3214285714285714
    },
    {
      "start": 600.28,
      "end": 601.12,
      "text": "哇",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.11697645823160807,
      "no_speech_prob": 1.4908536211111212e-11,
      "compression_ratio": 1.3214285714285714
    },
    {
      "start": 601.12,
      "end": 605.36,
      "text": "这就像是在训练一个极度严格的听打员",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.10065394272038966,
      "no_speech_prob": 2.7681410982860477e-11,
      "compression_ratio": 1.2954545454545454
    },
    {
      "start": 605.36,
      "end": 607.26,
      "text": "你就是不能错",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.10065394272038966,
      "no_speech_prob": 2.7681410982860477e-11,
      "compression_ratio": 1.2954545454545454
    },
    {
      "start": 607.26,
      "end": 608.76,
      "text": "但是你也不能慢",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.10065394272038966,
      "no_speech_prob": 2.7681410982860477e-11,
      "compression_ratio": 1.2954545454545454
    },
    {
      "start": 608.76,
      "end": 610.82,
      "text": "就是这种严格的训练",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.10065394272038966,
      "no_speech_prob": 2.7681410982860477e-11,
      "compression_ratio": 1.2954545454545454
    },
    {
      "start": 610.82,
      "end": 613.32,
      "text": "逼迫模型学会了去判断语境",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.10065394272038966,
      "no_speech_prob": 2.7681410982860477e-11,
      "compression_ratio": 1.2954545454545454
    },
    {
      "start": 613.32,
      "end": 615.52,
      "text": "碰到简单的日常对话",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.10065394272038966,
      "no_speech_prob": 2.7681410982860477e-11,
      "compression_ratio": 1.2954545454545454
    },
    {
      "start": 615.52,
      "end": 617.04,
      "text": "它就立刻秒翻",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.10065394272038966,
      "no_speech_prob": 2.7681410982860477e-11,
      "compression_ratio": 1.2954545454545454
    },
    {
      "start": 617.04,
      "end": 619.42,
      "text": "碰到尖色的专有名词",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.10065394272038966,
      "no_speech_prob": 2.7681410982860477e-11,
      "compression_ratio": 1.2954545454545454
    },
    {
      "start": 619.42,
      "end": 621.2,
      "text": "或者是模糊的发音",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.10065394272038966,
      "no_speech_prob": 2.7681410982860477e-11,
      "compression_ratio": 1.2954545454545454
    },
    {
      "start": 621.2,
      "end": 624.52,
      "text": "它就会选择多等个几十毫秒",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.10065394272038966,
      "no_speech_prob": 2.7681410982860477e-11,
      "compression_ratio": 1.2954545454545454
    },
    {
      "start": 624.52,
      "end": 627.92,
      "text": "收集更多声音特征之后再输出",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.10065394272038966,
      "no_speech_prob": 2.7681410982860477e-11,
      "compression_ratio": 1.2954545454545454
    },
    {
      "start": 627.92,
      "end": 628.96,
      "text": "这蛮合理的",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.10065394272038966,
      "no_speech_prob": 2.7681410982860477e-11,
      "compression_ratio": 1.2954545454545454
    },
    {
      "start": 628.96,
      "end": 629.52,
      "text": "对",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.10065394272038966,
      "no_speech_prob": 2.7681410982860477e-11,
      "compression_ratio": 1.2954545454545454
    },
    {
      "start": 629.52,
      "end": 631.12,
      "text": "在工程界",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.11712705272517793,
      "no_speech_prob": 2.044082784979917e-11,
      "compression_ratio": 1.1099656357388317
    },
    {
      "start": 631.12,
      "end": 633.48,
      "text": "这被称回达到了Pareto Front",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.11712705272517793,
      "no_speech_prob": 2.044082784979917e-11,
      "compression_ratio": 1.1099656357388317
    },
    {
      "start": 633.48,
      "end": 636.48,
      "text": "也就是在速度与准确度之间",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.11712705272517793,
      "no_speech_prob": 2.044082784979917e-11,
      "compression_ratio": 1.1099656357388317
    },
    {
      "start": 636.48,
      "end": 639.5,
      "text": "找到了无法再被超越的最佳边界",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.11712705272517793,
      "no_speech_prob": 2.044082784979917e-11,
      "compression_ratio": 1.1099656357388317
    },
    {
      "start": 639.5,
      "end": 641.36,
      "text": "最终的结果是",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.11712705272517793,
      "no_speech_prob": 2.044082784979917e-11,
      "compression_ratio": 1.1099656357388317
    },
    {
      "start": 641.36,
      "end": 642.84,
      "text": "Muse Voice Transcribe",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.11712705272517793,
      "no_speech_prob": 2.044082784979917e-11,
      "compression_ratio": 1.1099656357388317
    },
    {
      "start": 642.84,
      "end": 644.44,
      "text": "从你讲完话",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.11712705272517793,
      "no_speech_prob": 2.044082784979917e-11,
      "compression_ratio": 1.1099656357388317
    },
    {
      "start": 644.44,
      "end": 646.8,
      "text": "到萤幕上出现正确文字",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.11712705272517793,
      "no_speech_prob": 2.044082784979917e-11,
      "compression_ratio": 1.1099656357388317
    },
    {
      "start": 646.8,
      "end": 649.52,
      "text": "只需要惊人的0.16秒",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.11712705272517793,
      "no_speech_prob": 2.044082784979917e-11,
      "compression_ratio": 1.1099656357388317
    },
    {
      "start": 649.52,
      "end": 650.04,
      "text": "等等",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.11712705272517793,
      "no_speech_prob": 2.044082784979917e-11,
      "compression_ratio": 1.1099656357388317
    },
    {
      "start": 650.04,
      "end": 653.14,
      "text": "我要在这里稍微挑战一下",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.11712705272517793,
      "no_speech_prob": 2.044082784979917e-11,
      "compression_ratio": 1.1099656357388317
    },
    {
      "start": 653.14,
      "end": 654.96,
      "text": "0.16秒",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.11712705272517793,
      "no_speech_prob": 2.044082784979917e-11,
      "compression_ratio": 1.1099656357388317
    },
    {
      "start": 654.96,
      "end": 657.34,
      "text": "人类大脑神经传导",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.11712705272517793,
      "no_speech_prob": 2.044082784979917e-11,
      "compression_ratio": 1.1099656357388317
    },
    {
      "start": 657.34,
      "end": 658.72,
      "text": "做出反应的时间",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.11712705272517793,
      "no_speech_prob": 2.044082784979917e-11,
      "compression_ratio": 1.1099656357388317
    },
    {
      "start": 658.72,
      "end": 660.24,
      "text": "搞不好都比这个慢",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.09443648865348414,
      "no_speech_prob": 1.1075209326028013e-11,
      "compression_ratio": 1.231578947368421
    },
    {
      "start": 660.24,
      "end": 663.96,
      "text": "一个AI模型要在这0.16秒内",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.09443648865348414,
      "no_speech_prob": 1.1075209326028013e-11,
      "compression_ratio": 1.231578947368421
    },
    {
      "start": 663.96,
      "end": 665.1,
      "text": "切分音讯",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.09443648865348414,
      "no_speech_prob": 1.1075209326028013e-11,
      "compression_ratio": 1.231578947368421
    },
    {
      "start": 665.1,
      "end": 666.74,
      "text": "决定要不要等待",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.09443648865348414,
      "no_speech_prob": 1.1075209326028013e-11,
      "compression_ratio": 1.231578947368421
    },
    {
      "start": 666.74,
      "end": 669.04,
      "text": "还要辨识是谁说话",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.09443648865348414,
      "no_speech_prob": 1.1075209326028013e-11,
      "compression_ratio": 1.231578947368421
    },
    {
      "start": 669.04,
      "end": 670.38,
      "text": "并伸出文字",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.09443648865348414,
      "no_speech_prob": 1.1075209326028013e-11,
      "compression_ratio": 1.231578947368421
    },
    {
      "start": 670.38,
      "end": 673.92,
      "text": "他怎么可能在这么短的时间内运算完",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.09443648865348414,
      "no_speech_prob": 1.1075209326028013e-11,
      "compression_ratio": 1.231578947368421
    },
    {
      "start": 673.92,
      "end": 675.82,
      "text": "却不吃垮电脑的效能",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.09443648865348414,
      "no_speech_prob": 1.1075209326028013e-11,
      "compression_ratio": 1.231578947368421
    },
    {
      "start": 675.82,
      "end": 678.52,
      "text": "这真的是个非常好的质疑",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.09443648865348414,
      "no_speech_prob": 1.1075209326028013e-11,
      "compression_ratio": 1.231578947368421
    },
    {
      "start": 678.52,
      "end": 680.34,
      "text": "直觉上做这么多事情",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.09443648865348414,
      "no_speech_prob": 1.1075209326028013e-11,
      "compression_ratio": 1.231578947368421
    },
    {
      "start": 680.34,
      "end": 682.16,
      "text": "应该会严重拖慢系统",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.09443648865348414,
      "no_speech_prob": 1.1075209326028013e-11,
      "compression_ratio": 1.231578947368421
    },
    {
      "start": 682.16,
      "end": 683.06,
      "text": "对啊",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.09443648865348414,
      "no_speech_prob": 1.1075209326028013e-11,
      "compression_ratio": 1.231578947368421
    },
    {
      "start": 683.06,
      "end": 684.6,
      "text": "听起来运算量超大的",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.09443648865348414,
      "no_speech_prob": 1.1075209326028013e-11,
      "compression_ratio": 1.231578947368421
    },
    {
      "start": 684.6,
      "end": 686.66,
      "text": "但实际情况恰好相反",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.09443648865348414,
      "no_speech_prob": 1.1075209326028013e-11,
      "compression_ratio": 1.231578947368421
    },
    {
      "start": 686.66,
      "end": 689.06,
      "text": "正因为它舍去了过去那种",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.09119038581848145,
      "no_speech_prob": 1.7430697510367743e-11,
      "compression_ratio": 1.2553191489361701
    },
    {
      "start": 689.06,
      "end": 692.58,
      "text": "三套系统互相传递资料的管线架构",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.09119038581848145,
      "no_speech_prob": 1.7430697510367743e-11,
      "compression_ratio": 1.2553191489361701
    },
    {
      "start": 692.58,
      "end": 695.38,
      "text": "它省去了系统之间来回沟通",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.09119038581848145,
      "no_speech_prob": 1.7430697510367743e-11,
      "compression_ratio": 1.2553191489361701
    },
    {
      "start": 695.38,
      "end": 698.3,
      "text": "和对时间轴的庞大运算成本",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.09119038581848145,
      "no_speech_prob": 1.7430697510367743e-11,
      "compression_ratio": 1.2553191489361701
    },
    {
      "start": 698.3,
      "end": 700.8,
      "text": "就不用大家互相等来等去",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.09119038581848145,
      "no_speech_prob": 1.7430697510367743e-11,
      "compression_ratio": 1.2553191489361701
    },
    {
      "start": 700.8,
      "end": 701.4,
      "text": "没错",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.09119038581848145,
      "no_speech_prob": 1.7430697510367743e-11,
      "compression_ratio": 1.2553191489361701
    },
    {
      "start": 701.4,
      "end": 703.88,
      "text": "一个单一的自回归模型",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.09119038581848145,
      "no_speech_prob": 1.7430697510367743e-11,
      "compression_ratio": 1.2553191489361701
    },
    {
      "start": 703.88,
      "end": 706.28,
      "text": "只要硬体的记忆体频宽足够",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.09119038581848145,
      "no_speech_prob": 1.7430697510367743e-11,
      "compression_ratio": 1.2553191489361701
    },
    {
      "start": 706.28,
      "end": 708.56,
      "text": "它处理连续串流的效率",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.09119038581848145,
      "no_speech_prob": 1.7430697510367743e-11,
      "compression_ratio": 1.2553191489361701
    },
    {
      "start": 708.56,
      "end": 711.74,
      "text": "其实远比传统的分工合作还要快得多",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.09119038581848145,
      "no_speech_prob": 1.7430697510367743e-11,
      "compression_ratio": 1.2553191489361701
    },
    {
      "start": 711.74,
      "end": 713.24,
      "text": "这也就是为什么",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.09119038581848145,
      "no_speech_prob": 1.7430697510367743e-11,
      "compression_ratio": 1.2553191489361701
    },
    {
      "start": 713.24,
      "end": 716.96,
      "text": "它能压榨出0.16秒这种极限速度",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.07607310827524384,
      "no_speech_prob": 1.0548137883981124e-11,
      "compression_ratio": 1.1932515337423313
    },
    {
      "start": 716.96,
      "end": 718.68,
      "text": "哇 原来是这样",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.07607310827524384,
      "no_speech_prob": 1.0548137883981124e-11,
      "compression_ratio": 1.1932515337423313
    },
    {
      "start": 718.68,
      "end": 722.28,
      "text": "理论和底层机制听起来确实很惊人啦",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.07607310827524384,
      "no_speech_prob": 1.0548137883981124e-11,
      "compression_ratio": 1.1932515337423313
    },
    {
      "start": 722.28,
      "end": 723.64,
      "text": "但我们都知道",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.07607310827524384,
      "no_speech_prob": 1.0548137883981124e-11,
      "compression_ratio": 1.1932515337423313
    },
    {
      "start": 723.64,
      "end": 726.72,
      "text": "实验室里面的数据是一回事",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.07607310827524384,
      "no_speech_prob": 1.0548137883981124e-11,
      "compression_ratio": 1.1932515337423313
    },
    {
      "start": 726.72,
      "end": 729.62,
      "text": "把模型丢进残酷的真实世界",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.07607310827524384,
      "no_speech_prob": 1.0548137883981124e-11,
      "compression_ratio": 1.1932515337423313
    },
    {
      "start": 729.62,
      "end": 732.12,
      "text": "面对各种混乱的对话场景",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.07607310827524384,
      "no_speech_prob": 1.0548137883981124e-11,
      "compression_ratio": 1.1932515337423313
    },
    {
      "start": 732.12,
      "end": 734.0,
      "text": "那又是另外一回事了嘛",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.07607310827524384,
      "no_speech_prob": 1.0548137883981124e-11,
      "compression_ratio": 1.1932515337423313
    },
    {
      "start": 734.0,
      "end": 736.96,
      "text": "它在真实环境中的表现到底怎么样",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.07607310827524384,
      "no_speech_prob": 1.0548137883981124e-11,
      "compression_ratio": 1.1932515337423313
    },
    {
      "start": 736.96,
      "end": 740.04,
      "text": "那我们就来看看知名的AI评测机构",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.07607310827524384,
      "no_speech_prob": 1.0548137883981124e-11,
      "compression_ratio": 1.1932515337423313
    },
    {
      "start": 740.04,
      "end": 743.2,
      "text": "Artificial Analysis给出的硬数据吧",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.07607310827524384,
      "no_speech_prob": 1.0548137883981124e-11,
      "compression_ratio": 1.1932515337423313
    },
    {
      "start": 743.2,
      "end": 746.9,
      "text": "在纯英文的即时串流测试里面",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.17158293898088217,
      "no_speech_prob": 2.1584730530710416e-11,
      "compression_ratio": 1.069090909090909
    },
    {
      "start": 746.9,
      "end": 751.24,
      "text": "Muse Voice Transcribe的字错率仅仅只有3.1%",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.17158293898088217,
      "no_speech_prob": 2.1584730530710416e-11,
      "compression_ratio": 1.069090909090909
    },
    {
      "start": 751.24,
      "end": 752.48,
      "text": "3.1%",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.17158293898088217,
      "no_speech_prob": 2.1584730530710416e-11,
      "compression_ratio": 1.069090909090909
    },
    {
      "start": 752.48,
      "end": 754.46,
      "text": "这算是很好的成绩吗",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.17158293898088217,
      "no_speech_prob": 2.1584730530710416e-11,
      "compression_ratio": 1.069090909090909
    },
    {
      "start": 754.46,
      "end": 755.56,
      "text": "非常好",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.17158293898088217,
      "no_speech_prob": 2.1584730530710416e-11,
      "compression_ratio": 1.069090909090909
    },
    {
      "start": 755.56,
      "end": 756.7,
      "text": "这是什么概面",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.17158293898088217,
      "no_speech_prob": 2.1584730530710416e-11,
      "compression_ratio": 1.069090909090909
    },
    {
      "start": 756.7,
      "end": 760.14,
      "text": "它直接击败了目前语音领域的王者",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.17158293898088217,
      "no_speech_prob": 2.1584730530710416e-11,
      "compression_ratio": 1.069090909090909
    },
    {
      "start": 760.14,
      "end": 762.18,
      "text": "也就是那个11实验室",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.17158293898088217,
      "no_speech_prob": 2.1584730530710416e-11,
      "compression_ratio": 1.069090909090909
    },
    {
      "start": 762.18,
      "end": 764.4,
      "text": "他们的3.6%",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.17158293898088217,
      "no_speech_prob": 2.1584730530710416e-11,
      "compression_ratio": 1.069090909090909
    },
    {
      "start": 764.4,
      "end": 765.94,
      "text": "哦 11 labs",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.17158293898088217,
      "no_speech_prob": 2.1584730530710416e-11,
      "compression_ratio": 1.069090909090909
    },
    {
      "start": 765.94,
      "end": 766.48,
      "text": "对",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.17158293898088217,
      "no_speech_prob": 2.1584730530710416e-11,
      "compression_ratio": 1.069090909090909
    },
    {
      "start": 766.48,
      "end": 771.36,
      "text": "它也赢过了Oper AI的GPT Live Transcribe的3.9",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.17158293898088217,
      "no_speech_prob": 2.1584730530710416e-11,
      "compression_ratio": 1.069090909090909
    },
    {
      "start": 771.36,
      "end": 774.26,
      "text": "甚至把Google在8月才刚发布的",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.10946220349354349,
      "no_speech_prob": 3.050974403673301e-11,
      "compression_ratio": 1.1428571428571428
    },
    {
      "start": 774.26,
      "end": 776.9,
      "text": "German Live 3.5 Transcribe Live的4.0",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.10946220349354349,
      "no_speech_prob": 3.050974403673301e-11,
      "compression_ratio": 1.1428571428571428
    },
    {
      "start": 776.9,
      "end": 778.42,
      "text": "都抛在脑后",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.10946220349354349,
      "no_speech_prob": 3.050974403673301e-11,
      "compression_ratio": 1.1428571428571428
    },
    {
      "start": 778.42,
      "end": 780.82,
      "text": "哇 全部斩压耶",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.10946220349354349,
      "no_speech_prob": 3.050974403673301e-11,
      "compression_ratio": 1.1428571428571428
    },
    {
      "start": 780.82,
      "end": 781.34,
      "text": "没错",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.10946220349354349,
      "no_speech_prob": 3.050974403673301e-11,
      "compression_ratio": 1.1428571428571428
    },
    {
      "start": 781.34,
      "end": 784.3,
      "text": "而且在讲者辨识错误率上",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.10946220349354349,
      "no_speech_prob": 3.050974403673301e-11,
      "compression_ratio": 1.1428571428571428
    },
    {
      "start": 784.3,
      "end": 786.5,
      "text": "它的错误率只有17.5%",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.10946220349354349,
      "no_speech_prob": 3.050974403673301e-11,
      "compression_ratio": 1.1428571428571428
    },
    {
      "start": 786.5,
      "end": 789.32,
      "text": "这个数字甚至赢过了市面上许多",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.10946220349354349,
      "no_speech_prob": 3.050974403673301e-11,
      "compression_ratio": 1.1428571428571428
    },
    {
      "start": 789.32,
      "end": 793.08,
      "text": "允许你事后慢慢处理的那种非即时模型",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.10946220349354349,
      "no_speech_prob": 3.050974403673301e-11,
      "compression_ratio": 1.1428571428571428
    },
    {
      "start": 793.08,
      "end": 795.54,
      "text": "不过我觉得我们必须拉回现实一下",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.10946220349354349,
      "no_speech_prob": 3.050974403673301e-11,
      "compression_ratio": 1.1428571428571428
    },
    {
      "start": 795.54,
      "end": 797.64,
      "text": "我们的听众最关心的",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.10946220349354349,
      "no_speech_prob": 3.050974403673301e-11,
      "compression_ratio": 1.1428571428571428
    },
    {
      "start": 797.64,
      "end": 800.7,
      "text": "绝对不是在一个纯英文",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.10946220349354349,
      "no_speech_prob": 3.050974403673301e-11,
      "compression_ratio": 1.1428571428571428
    },
    {
      "start": 800.7,
      "end": 803.16,
      "text": "大家都自证强缘的环境里面",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.12003931214537801,
      "no_speech_prob": 1.9943153031776184e-11,
      "compression_ratio": 1.189710610932476
    },
    {
      "start": 803.16,
      "end": 803.9,
      "text": "它有多准",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.12003931214537801,
      "no_speech_prob": 1.9943153031776184e-11,
      "compression_ratio": 1.189710610932476
    },
    {
      "start": 803.9,
      "end": 804.98,
      "text": "嗯 确实",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.12003931214537801,
      "no_speech_prob": 1.9943153031776184e-11,
      "compression_ratio": 1.189710610932476
    },
    {
      "start": 804.98,
      "end": 805.48,
      "text": "对嘛",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.12003931214537801,
      "no_speech_prob": 1.9943153031776184e-11,
      "compression_ratio": 1.189710610932476
    },
    {
      "start": 805.48,
      "end": 806.52,
      "text": "我们在排湾",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.12003931214537801,
      "no_speech_prob": 1.9943153031776184e-11,
      "compression_ratio": 1.189710610932476
    },
    {
      "start": 806.52,
      "end": 809.42,
      "text": "平时讲话可是中阴夹杂的耶",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.12003931214537801,
      "no_speech_prob": 1.9943153031776184e-11,
      "compression_ratio": 1.189710610932476
    },
    {
      "start": 809.42,
      "end": 810.54,
      "text": "像是那个",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.12003931214537801,
      "no_speech_prob": 1.9943153031776184e-11,
      "compression_ratio": 1.189710610932476
    },
    {
      "start": 810.54,
      "end": 811.8,
      "text": "我明天有个meeting",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.12003931214537801,
      "no_speech_prob": 1.9943153031776184e-11,
      "compression_ratio": 1.189710610932476
    },
    {
      "start": 811.8,
      "end": 813.46,
      "text": "记得remind我一下",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.12003931214537801,
      "no_speech_prob": 1.9943153031776184e-11,
      "compression_ratio": 1.189710610932476
    },
    {
      "start": 813.46,
      "end": 816.34,
      "text": "这种句子才是我们最正式的日常啊",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.12003931214537801,
      "no_speech_prob": 1.9943153031776184e-11,
      "compression_ratio": 1.189710610932476
    },
    {
      "start": 816.34,
      "end": 818.94,
      "text": "Meta其实在这方面下了很大的功夫",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.12003931214537801,
      "no_speech_prob": 1.9943153031776184e-11,
      "compression_ratio": 1.189710610932476
    },
    {
      "start": 818.94,
      "end": 820.84,
      "text": "这款模型发布的时候",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.12003931214537801,
      "no_speech_prob": 1.9943153031776184e-11,
      "compression_ratio": 1.189710610932476
    },
    {
      "start": 820.84,
      "end": 823.16,
      "text": "就已经支援了70多种语言",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.12003931214537801,
      "no_speech_prob": 1.9943153031776184e-11,
      "compression_ratio": 1.189710610932476
    },
    {
      "start": 823.16,
      "end": 826.2,
      "text": "其中有25种是经过完整验证的",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.12003931214537801,
      "no_speech_prob": 1.9943153031776184e-11,
      "compression_ratio": 1.189710610932476
    },
    {
      "start": 826.2,
      "end": 827.38,
      "text": "最厉害的是",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.12003931214537801,
      "no_speech_prob": 1.9943153031776184e-11,
      "compression_ratio": 1.189710610932476
    },
    {
      "start": 827.38,
      "end": 831.28,
      "text": "它原生就支援句子内部的鱼马转换",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.1082120769752,
      "no_speech_prob": 2.1959715296171467e-11,
      "compression_ratio": 1.1369426751592357
    },
    {
      "start": 831.28,
      "end": 832.14,
      "text": "鱼马转换",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.1082120769752,
      "no_speech_prob": 2.1959715296171467e-11,
      "compression_ratio": 1.1369426751592357
    },
    {
      "start": 832.14,
      "end": 835.22,
      "text": "意思是它不需要我去切换那个语言设定",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.1082120769752,
      "no_speech_prob": 2.1959715296171467e-11,
      "compression_ratio": 1.1369426751592357
    },
    {
      "start": 835.22,
      "end": 836.78,
      "text": "就能自动听懂吗",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.1082120769752,
      "no_speech_prob": 2.1959715296171467e-11,
      "compression_ratio": 1.1369426751592357
    },
    {
      "start": 836.78,
      "end": 837.1,
      "text": "是的",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.1082120769752,
      "no_speech_prob": 2.1959715296171467e-11,
      "compression_ratio": 1.1369426751592357
    },
    {
      "start": 837.1,
      "end": 839.74,
      "text": "在官方释出的测试来源里面",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.1082120769752,
      "no_speech_prob": 2.1959715296171467e-11,
      "compression_ratio": 1.1369426751592357
    },
    {
      "start": 839.74,
      "end": 842.82,
      "text": "有一段充满硬核科技术语的中文录音",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.1082120769752,
      "no_speech_prob": 2.1959715296171467e-11,
      "compression_ratio": 1.1369426751592357
    },
    {
      "start": 842.82,
      "end": 844.08,
      "text": "测试者说",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.1082120769752,
      "no_speech_prob": 2.1959715296171467e-11,
      "compression_ratio": 1.1369426751592357
    },
    {
      "start": 844.08,
      "end": 846.98,
      "text": "用Olama跑了向Meta的Muse Glamour",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.1082120769752,
      "no_speech_prob": 2.1959715296171467e-11,
      "compression_ratio": 1.1369426751592357
    },
    {
      "start": 846.98,
      "end": 851.54,
      "text": "直接deploy在我的NVIDIA RTX 3090上",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.1082120769752,
      "no_speech_prob": 2.1959715296171467e-11,
      "compression_ratio": 1.1369426751592357
    },
    {
      "start": 851.54,
      "end": 854.96,
      "text": "我这句话里面充满了英文专有名词数字",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.1082120769752,
      "no_speech_prob": 2.1959715296171467e-11,
      "compression_ratio": 1.1369426751592357
    },
    {
      "start": 854.96,
      "end": 858.32,
      "text": "而且中英文切换完全是随机的耶",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.10911039000782886,
      "no_speech_prob": 2.3919994457388682e-11,
      "compression_ratio": 1.3129032258064517
    },
    {
      "start": 858.32,
      "end": 858.72,
      "text": "没错",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.10911039000782886,
      "no_speech_prob": 2.3919994457388682e-11,
      "compression_ratio": 1.3129032258064517
    },
    {
      "start": 858.72,
      "end": 861.86,
      "text": "母亲不需要任何提示就能无缝辨识",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.10911039000782886,
      "no_speech_prob": 2.3919994457388682e-11,
      "compression_ratio": 1.3129032258064517
    },
    {
      "start": 861.86,
      "end": 864.02,
      "text": "还能根据文法的上下文",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.10911039000782886,
      "no_speech_prob": 2.3919994457388682e-11,
      "compression_ratio": 1.3129032258064517
    },
    {
      "start": 864.02,
      "end": 865.78,
      "text": "去自动校正评字",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.10911039000782886,
      "no_speech_prob": 2.3919994457388682e-11,
      "compression_ratio": 1.3129032258064517
    },
    {
      "start": 865.78,
      "end": 867.58,
      "text": "说到真实世界的混乱",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.10911039000782886,
      "no_speech_prob": 2.3919994457388682e-11,
      "compression_ratio": 1.3129032258064517
    },
    {
      "start": 867.58,
      "end": 869.08,
      "text": "我看到资料里面",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.10911039000782886,
      "no_speech_prob": 2.3919994457388682e-11,
      "compression_ratio": 1.3129032258064517
    },
    {
      "start": 869.08,
      "end": 871.8,
      "text": "还有一个让我下巴掉下来的测试",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.10911039000782886,
      "no_speech_prob": 2.3919994457388682e-11,
      "compression_ratio": 1.3129032258064517
    },
    {
      "start": 871.8,
      "end": 873.2,
      "text": "你说那个爬山的那个吗",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.10911039000782886,
      "no_speech_prob": 2.3919994457388682e-11,
      "compression_ratio": 1.3129032258064517
    },
    {
      "start": 873.2,
      "end": 873.84,
      "text": "对对对",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.10911039000782886,
      "no_speech_prob": 2.3919994457388682e-11,
      "compression_ratio": 1.3129032258064517
    },
    {
      "start": 873.84,
      "end": 876.26,
      "text": "那个是一个长达一小时的音档",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.10911039000782886,
      "no_speech_prob": 2.3919994457388682e-11,
      "compression_ratio": 1.3129032258064517
    },
    {
      "start": 876.26,
      "end": 879.14,
      "text": "里面竟然有11个人",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.10911039000782886,
      "no_speech_prob": 2.3919994457388682e-11,
      "compression_ratio": 1.3129032258064517
    },
    {
      "start": 879.14,
      "end": 881.04,
      "text": "在洛杉矶讨论爬山",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.10911039000782886,
      "no_speech_prob": 2.3919994457388682e-11,
      "compression_ratio": 1.3129032258064517
    },
    {
      "start": 881.04,
      "end": 883.5,
      "text": "甚至还聊到遇到土狼的经历",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.10911039000782886,
      "no_speech_prob": 2.3919994457388682e-11,
      "compression_ratio": 1.3129032258064517
    },
    {
      "start": 883.5,
      "end": 884.66,
      "text": "你想想看",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.10911039000782886,
      "no_speech_prob": 2.3919994457388682e-11,
      "compression_ratio": 1.3129032258064517
    },
    {
      "start": 884.66,
      "end": 887.08,
      "text": "11个人在闲聊耶",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.10509483337402344,
      "no_speech_prob": 1.4774532292038955e-11,
      "compression_ratio": 1.2408026755852843
    },
    {
      "start": 887.08,
      "end": 888.66,
      "text": "一定有人在大笑",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.10509483337402344,
      "no_speech_prob": 1.4774532292038955e-11,
      "compression_ratio": 1.2408026755852843
    },
    {
      "start": 888.66,
      "end": 890.04,
      "text": "有人互相抢话",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.10509483337402344,
      "no_speech_prob": 1.4774532292038955e-11,
      "compression_ratio": 1.2408026755852843
    },
    {
      "start": 890.04,
      "end": 892.34,
      "text": "背景绝对还有各种环境噪音",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.10509483337402344,
      "no_speech_prob": 1.4774532292038955e-11,
      "compression_ratio": 1.2408026755852843
    },
    {
      "start": 892.34,
      "end": 893.42,
      "text": "这对AI来说",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.10509483337402344,
      "no_speech_prob": 1.4774532292038955e-11,
      "compression_ratio": 1.2408026755852843
    },
    {
      "start": 893.42,
      "end": 895.04,
      "text": "根本是地狱级的考验",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.10509483337402344,
      "no_speech_prob": 1.4774532292038955e-11,
      "compression_ratio": 1.2408026755852843
    },
    {
      "start": 895.04,
      "end": 895.58,
      "text": "真的",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.10509483337402344,
      "no_speech_prob": 1.4774532292038955e-11,
      "compression_ratio": 1.2408026755852843
    },
    {
      "start": 895.58,
      "end": 898.22,
      "text": "结果系统竟然能精准的",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.10509483337402344,
      "no_speech_prob": 1.4774532292038955e-11,
      "compression_ratio": 1.2408026755852843
    },
    {
      "start": 898.22,
      "end": 900.04,
      "text": "把那11个人的发言",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.10509483337402344,
      "no_speech_prob": 1.4774532292038955e-11,
      "compression_ratio": 1.2408026755852843
    },
    {
      "start": 900.04,
      "end": 902.12,
      "text": "一次不漏的拆解开来",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.10509483337402344,
      "no_speech_prob": 1.4774532292038955e-11,
      "compression_ratio": 1.2408026755852843
    },
    {
      "start": 902.12,
      "end": 904.28,
      "text": "还标上speakerA到K",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.10509483337402344,
      "no_speech_prob": 1.4774532292038955e-11,
      "compression_ratio": 1.2408026755852843
    },
    {
      "start": 904.28,
      "end": 906.9,
      "text": "这如果叫人类的听答员来做",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.10509483337402344,
      "no_speech_prob": 1.4774532292038955e-11,
      "compression_ratio": 1.2408026755852843
    },
    {
      "start": 906.9,
      "end": 908.14,
      "text": "听个10分钟",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.10509483337402344,
      "no_speech_prob": 1.4774532292038955e-11,
      "compression_ratio": 1.2408026755852843
    },
    {
      "start": 908.14,
      "end": 909.74,
      "text": "绝对崩溃磁石吧",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.10509483337402344,
      "no_speech_prob": 1.4774532292038955e-11,
      "compression_ratio": 1.2408026755852843
    },
    {
      "start": 909.74,
      "end": 911.36,
      "text": "但这在技术上",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.10509483337402344,
      "no_speech_prob": 1.4774532292038955e-11,
      "compression_ratio": 1.2408026755852843
    },
    {
      "start": 911.36,
      "end": 912.74,
      "text": "到底是怎么办到的",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.10509483337402344,
      "no_speech_prob": 1.4774532292038955e-11,
      "compression_ratio": 1.2408026755852843
    },
    {
      "start": 912.74,
      "end": 913.82,
      "text": "如果两个人",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.10509483337402344,
      "no_speech_prob": 1.4774532292038955e-11,
      "compression_ratio": 1.2408026755852843
    },
    {
      "start": 913.82,
      "end": 916.06,
      "text": "在同一微秒重叠讲话呢",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.14219184038115712,
      "no_speech_prob": 2.5866322972412092e-11,
      "compression_ratio": 1.1934426229508197
    },
    {
      "start": 916.06,
      "end": 917.38,
      "text": "这就是我们刚刚说的",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.14219184038115712,
      "no_speech_prob": 2.5866322972412092e-11,
      "compression_ratio": 1.1934426229508197
    },
    {
      "start": 917.38,
      "end": 920.18,
      "text": "多模态标记发挥威力的地方了",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.14219184038115712,
      "no_speech_prob": 2.5866322972412092e-11,
      "compression_ratio": 1.1934426229508197
    },
    {
      "start": 920.18,
      "end": 922.52,
      "text": "当两个人同时讲话的时候",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.14219184038115712,
      "no_speech_prob": 2.5866322972412092e-11,
      "compression_ratio": 1.1934426229508197
    },
    {
      "start": 922.52,
      "end": 925.56,
      "text": "传统系统的音轨会整个混在一起",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.14219184038115712,
      "no_speech_prob": 2.5866322972412092e-11,
      "compression_ratio": 1.1934426229508197
    },
    {
      "start": 925.56,
      "end": 927.56,
      "text": "变成无法辨识的噪音",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.14219184038115712,
      "no_speech_prob": 2.5866322972412092e-11,
      "compression_ratio": 1.1934426229508197
    },
    {
      "start": 927.56,
      "end": 929.68,
      "text": "嗯就胡称一团",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.14219184038115712,
      "no_speech_prob": 2.5866322972412092e-11,
      "compression_ratio": 1.1934426229508197
    },
    {
      "start": 929.68,
      "end": 930.14,
      "text": "对",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.14219184038115712,
      "no_speech_prob": 2.5866322972412092e-11,
      "compression_ratio": 1.1934426229508197
    },
    {
      "start": 930.14,
      "end": 931.94,
      "text": "但Muse Voice Transcribe",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.14219184038115712,
      "no_speech_prob": 2.5866322972412092e-11,
      "compression_ratio": 1.1934426229508197
    },
    {
      "start": 931.94,
      "end": 935.44,
      "text": "因为是从极短的80毫秒音讯区块",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.14219184038115712,
      "no_speech_prob": 2.5866322972412092e-11,
      "compression_ratio": 1.1934426229508197
    },
    {
      "start": 935.44,
      "end": 936.46,
      "text": "去提取特征",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.14219184038115712,
      "no_speech_prob": 2.5866322972412092e-11,
      "compression_ratio": 1.1934426229508197
    },
    {
      "start": 936.46,
      "end": 939.44,
      "text": "它可以同时触发多个讲者的token",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.14219184038115712,
      "no_speech_prob": 2.5866322972412092e-11,
      "compression_ratio": 1.1934426229508197
    },
    {
      "start": 939.44,
      "end": 941.06,
      "text": "可以同时触发",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.14219184038115712,
      "no_speech_prob": 2.5866322972412092e-11,
      "compression_ratio": 1.1934426229508197
    },
    {
      "start": 941.06,
      "end": 941.6,
      "text": "没错",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.14219184038115712,
      "no_speech_prob": 2.5866322972412092e-11,
      "compression_ratio": 1.1934426229508197
    },
    {
      "start": 941.6,
      "end": 944.46,
      "text": "它会在文字流里面并行输出",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.09993199542560409,
      "no_speech_prob": 1.8325178180456092e-11,
      "compression_ratio": 1.1266375545851528
    },
    {
      "start": 944.46,
      "end": 946.3,
      "text": "Speaker A说了这一段",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.09993199542560409,
      "no_speech_prob": 1.8325178180456092e-11,
      "compression_ratio": 1.1266375545851528
    },
    {
      "start": 946.3,
      "end": 948.66,
      "text": "同时Speaker B正在说那一段",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.09993199542560409,
      "no_speech_prob": 1.8325178180456092e-11,
      "compression_ratio": 1.1266375545851528
    },
    {
      "start": 948.66,
      "end": 950.5,
      "text": "它不是单线程的处理",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.09993199542560409,
      "no_speech_prob": 1.8325178180456092e-11,
      "compression_ratio": 1.1266375545851528
    },
    {
      "start": 950.5,
      "end": 952.46,
      "text": "而是能像解开毛线球一样",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.09993199542560409,
      "no_speech_prob": 1.8325178180456092e-11,
      "compression_ratio": 1.1266375545851528
    },
    {
      "start": 952.46,
      "end": 955.02,
      "text": "把交织在一起的音讯分离出来",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.09993199542560409,
      "no_speech_prob": 1.8325178180456092e-11,
      "compression_ratio": 1.1266375545851528
    },
    {
      "start": 955.02,
      "end": 956.92,
      "text": "这技术真的蛮扯的",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.09993199542560409,
      "no_speech_prob": 1.8325178180456092e-11,
      "compression_ratio": 1.1266375545851528
    },
    {
      "start": 956.92,
      "end": 959.76,
      "text": "不过身为替台湾听众把关的酒色",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.09993199542560409,
      "no_speech_prob": 1.8325178180456092e-11,
      "compression_ratio": 1.1266375545851528
    },
    {
      "start": 964.5,
      "end": 966.0,
      "text": "哦,什麼事實?",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.13766370182148893,
      "no_speech_prob": 8.94035713516761e-12,
      "compression_ratio": 1.2557471264367817
    },
    {
      "start": 966.0,
      "end": 974.7,
      "text": "就是,根據台灣知名科技媒體電腦王阿達的實際測試,這項技術目前有一個在地化的致命缺陷。",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.13766370182148893,
      "no_speech_prob": 8.94035713516761e-12,
      "compression_ratio": 1.2557471264367817
    },
    {
      "start": 974.7,
      "end": 976.1,
      "text": "嗯,在地化",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.10730937549046107,
      "no_speech_prob": 1.9723210911704037e-11,
      "compression_ratio": 1.3664383561643836
    },
    {
      "start": 976.1,
      "end": 980.66,
      "text": "对,尽管你对着麦克风说的是很标准的台湾繁体中文",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.10730937549046107,
      "no_speech_prob": 1.9723210911704037e-11,
      "compression_ratio": 1.3664383561643836
    },
    {
      "start": 980.66,
      "end": 986.04,
      "text": "它目前在荧幕上输出的文字依然全都是,呃,简体中文",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.10730937549046107,
      "no_speech_prob": 1.9723210911704037e-11,
      "compression_ratio": 1.3664383561643836
    },
    {
      "start": 986.04,
      "end": 988.14,
      "text": "这确实蛮麻烦的",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.10730937549046107,
      "no_speech_prob": 1.9723210911704037e-11,
      "compression_ratio": 1.3664383561643836
    },
    {
      "start": 988.14,
      "end": 992.56,
      "text": "对啊,不管设定怎么调,目前就是无法原声输出繁体字",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.10730937549046107,
      "no_speech_prob": 1.9723210911704037e-11,
      "compression_ratio": 1.3664383561643836
    },
    {
      "start": 992.56,
      "end": 996.6,
      "text": "这对台湾的使用者来说绝对还是一个不小的痛点啊",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.10730937549046107,
      "no_speech_prob": 1.9723210911704037e-11,
      "compression_ratio": 1.3664383561643836
    },
    {
      "start": 996.6,
      "end": 999.44,
      "text": "这的确是一个很明显的在地化缺陷",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.10730937549046107,
      "no_speech_prob": 1.9723210911704037e-11,
      "compression_ratio": 1.3664383561643836
    },
    {
      "start": 999.44,
      "end": 1003.14,
      "text": "其实这显示出虽然模型的听力没有问题",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.10730937549046107,
      "no_speech_prob": 1.9723210911704037e-11,
      "compression_ratio": 1.3664383561643836
    },
    {
      "start": 1003.14,
      "end": 1006.74,
      "text": "它的神经网络完全能理解中文的语义",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.08355503760013483,
      "no_speech_prob": 1.3391550889030324e-11,
      "compression_ratio": 1.2997032640949555
    },
    {
      "start": 1006.74,
      "end": 1008.26,
      "text": "听得懂但写出来不对",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.08355503760013483,
      "no_speech_prob": 1.3391550889030324e-11,
      "compression_ratio": 1.2997032640949555
    },
    {
      "start": 1008.26,
      "end": 1008.72,
      "text": "没错",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.08355503760013483,
      "no_speech_prob": 1.3391550889030324e-11,
      "compression_ratio": 1.2997032640949555
    },
    {
      "start": 1008.72,
      "end": 1011.78,
      "text": "但在最后的文字渲染与输出层面",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.08355503760013483,
      "no_speech_prob": 1.3391550889030324e-11,
      "compression_ratio": 1.2997032640949555
    },
    {
      "start": 1011.78,
      "end": 1013.36,
      "text": "META的训练资料库",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.08355503760013483,
      "no_speech_prob": 1.3391550889030324e-11,
      "compression_ratio": 1.2997032640949555
    },
    {
      "start": 1013.36,
      "end": 1016.16,
      "text": "可能还是过度向简体中文倾斜了",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.08355503760013483,
      "no_speech_prob": 1.3391550889030324e-11,
      "compression_ratio": 1.2997032640949555
    },
    {
      "start": 1016.16,
      "end": 1018.88,
      "text": "这也反映了全球化大语言模型",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.08355503760013483,
      "no_speech_prob": 1.3391550889030324e-11,
      "compression_ratio": 1.2997032640949555
    },
    {
      "start": 1018.88,
      "end": 1019.86,
      "text": "在落地的时候",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.08355503760013483,
      "no_speech_prob": 1.3391550889030324e-11,
      "compression_ratio": 1.2997032640949555
    },
    {
      "start": 1019.86,
      "end": 1023.62,
      "text": "必然会面临到的文化与区域细节挑战了",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.08355503760013483,
      "no_speech_prob": 1.3391550889030324e-11,
      "compression_ratio": 1.2997032640949555
    },
    {
      "start": 1023.62,
      "end": 1025.4,
      "text": "这部分就需要META后续",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.08355503760013483,
      "no_speech_prob": 1.3391550889030324e-11,
      "compression_ratio": 1.2997032640949555
    },
    {
      "start": 1025.4,
      "end": 1026.82,
      "text": "透过微调来解决了",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.08355503760013483,
      "no_speech_prob": 1.3391550889030324e-11,
      "compression_ratio": 1.2997032640949555
    },
    {
      "start": 1026.82,
      "end": 1028.64,
      "text": "既然提到META的策略",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.08355503760013483,
      "no_speech_prob": 1.3391550889030324e-11,
      "compression_ratio": 1.2997032640949555
    },
    {
      "start": 1028.64,
      "end": 1031.46,
      "text": "我们来聊聊一个很反常的商业布局好了",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.08355503760013483,
      "no_speech_prob": 1.3391550889030324e-11,
      "compression_ratio": 1.2997032640949555
    },
    {
      "start": 1031.46,
      "end": 1032.54,
      "text": "我们都知道",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.08355503760013483,
      "no_speech_prob": 1.3391550889030324e-11,
      "compression_ratio": 1.2997032640949555
    },
    {
      "start": 1032.54,
      "end": 1036.32,
      "text": "Meta过去这一年来一直高举着开源的大旗",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.07547149996785722,
      "no_speech_prob": 1.3498267607825465e-11,
      "compression_ratio": 1.229607250755287
    },
    {
      "start": 1036.32,
      "end": 1039.92,
      "text": "把他们最大的AI模型免费开放给全世界",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.07547149996785722,
      "no_speech_prob": 1.3498267607825465e-11,
      "compression_ratio": 1.229607250755287
    },
    {
      "start": 1039.92,
      "end": 1041.6,
      "text": "对 开源界的老大哥",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.07547149996785722,
      "no_speech_prob": 1.3498267607825465e-11,
      "compression_ratio": 1.229607250755287
    },
    {
      "start": 1041.6,
      "end": 1042.3,
      "text": "没错",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.07547149996785722,
      "no_speech_prob": 1.3498267607825465e-11,
      "compression_ratio": 1.229607250755287
    },
    {
      "start": 1042.3,
      "end": 1046.62,
      "text": "但这一次Mail Voice Transcribe却选择了封闭权重",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.07547149996785722,
      "no_speech_prob": 1.3498267607825465e-11,
      "compression_ratio": 1.229607250755287
    },
    {
      "start": 1046.62,
      "end": 1048.8,
      "text": "也就是不开源了",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.07547149996785722,
      "no_speech_prob": 1.3498267607825465e-11,
      "compression_ratio": 1.229607250755287
    },
    {
      "start": 1048.8,
      "end": 1049.38,
      "text": "为什么",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.07547149996785722,
      "no_speech_prob": 1.3498267607825465e-11,
      "compression_ratio": 1.229607250755287
    },
    {
      "start": 1049.38,
      "end": 1052.22,
      "text": "这次Meta确实选择了一条不一样的路",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.07547149996785722,
      "no_speech_prob": 1.3498267607825465e-11,
      "compression_ratio": 1.229607250755287
    },
    {
      "start": 1052.22,
      "end": 1053.6,
      "text": "如果你是开发者",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.07547149996785722,
      "no_speech_prob": 1.3498267607825465e-11,
      "compression_ratio": 1.229607250755287
    },
    {
      "start": 1053.6,
      "end": 1055.78,
      "text": "你无法直接下载模型版身",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.07547149996785722,
      "no_speech_prob": 1.3498267607825465e-11,
      "compression_ratio": 1.229607250755287
    },
    {
      "start": 1055.78,
      "end": 1058.76,
      "text": "你必须透过Meta Model API来使用",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.07547149996785722,
      "no_speech_prob": 1.3498267607825465e-11,
      "compression_ratio": 1.229607250755287
    },
    {
      "start": 1058.76,
      "end": 1059.98,
      "text": "哦 要收费了",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.07547149996785722,
      "no_speech_prob": 1.3498267607825465e-11,
      "compression_ratio": 1.229607250755287
    },
    {
      "start": 1059.98,
      "end": 1062.08,
      "text": "不过收费其实相当低廉啦",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.07547149996785722,
      "no_speech_prob": 1.3498267607825465e-11,
      "compression_ratio": 1.229607250755287
    },
    {
      "start": 1062.08,
      "end": 1065.86,
      "text": "大约是每1000分钟的音讯收费3美元",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.09577218482368871,
      "no_speech_prob": 1.0572336409109262e-11,
      "compression_ratio": 1.165644171779141
    },
    {
      "start": 1065.86,
      "end": 1069.72,
      "text": "换算下来一个小时大概才0.18美元",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.09577218482368871,
      "no_speech_prob": 1.0572336409109262e-11,
      "compression_ratio": 1.165644171779141
    },
    {
      "start": 1069.72,
      "end": 1071.52,
      "text": "那真的蛮便宜的",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.09577218482368871,
      "no_speech_prob": 1.0572336409109262e-11,
      "compression_ratio": 1.165644171779141
    },
    {
      "start": 1071.52,
      "end": 1072.06,
      "text": "对",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.09577218482368871,
      "no_speech_prob": 1.0572336409109262e-11,
      "compression_ratio": 1.165644171779141
    },
    {
      "start": 1072.06,
      "end": 1076.68,
      "text": "另外一般使用者其实可以透过他们新推出的程式撰写工具",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.09577218482368871,
      "no_speech_prob": 1.0572336409109262e-11,
      "compression_ratio": 1.165644171779141
    },
    {
      "start": 1076.68,
      "end": 1078.02,
      "text": "叫做Muse Code",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.09577218482368871,
      "no_speech_prob": 1.0572336409109262e-11,
      "compression_ratio": 1.165644171779141
    },
    {
      "start": 1078.02,
      "end": 1083.5,
      "text": "或者是直接在Mac电脑上下载Meta AI的应用程式来使用这项技术",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.09577218482368871,
      "no_speech_prob": 1.0572336409109262e-11,
      "compression_ratio": 1.165644171779141
    },
    {
      "start": 1083.5,
      "end": 1085.7,
      "text": "所以这到底意味着什么",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.09577218482368871,
      "no_speech_prob": 1.0572336409109262e-11,
      "compression_ratio": 1.165644171779141
    },
    {
      "start": 1085.7,
      "end": 1091.62,
      "text": "Meta把这个听力超强的模型紧紧锁在自家的生态系里面",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.09577218482368871,
      "no_speech_prob": 1.0572336409109262e-11,
      "compression_ratio": 1.165644171779141
    },
    {
      "start": 1091.62,
      "end": 1094.6,
      "text": "甚至还跟Mac的键盘深度绑定",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.08092220100813997,
      "no_speech_prob": 1.2052917691685039e-11,
      "compression_ratio": 1.243076923076923
    },
    {
      "start": 1094.6,
      "end": 1098.58,
      "text": "Mark Zuckerberg到底在盘算什么更大的棋局",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.08092220100813997,
      "no_speech_prob": 1.2052917691685039e-11,
      "compression_ratio": 1.243076923076923
    },
    {
      "start": 1098.58,
      "end": 1102.52,
      "text": "这不仅仅是为了赚那个0.18美元吧",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.08092220100813997,
      "no_speech_prob": 1.2052917691685039e-11,
      "compression_ratio": 1.243076923076923
    },
    {
      "start": 1102.52,
      "end": 1105.18,
      "text": "绝对不是为了这点API的零头啦",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.08092220100813997,
      "no_speech_prob": 1.2052917691685039e-11,
      "compression_ratio": 1.243076923076923
    },
    {
      "start": 1105.18,
      "end": 1108.46,
      "text": "如果我们深入观察 Zuckerberg近期的内部发言",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.08092220100813997,
      "no_speech_prob": 1.2052917691685039e-11,
      "compression_ratio": 1.243076923076923
    },
    {
      "start": 1108.46,
      "end": 1112.3,
      "text": "他不断强调未来的AI必须是个人化的超级智慧",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.08092220100813997,
      "no_speech_prob": 1.2052917691685039e-11,
      "compression_ratio": 1.243076923076923
    },
    {
      "start": 1112.3,
      "end": 1113.98,
      "text": "个人化的超级智慧",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.08092220100813997,
      "no_speech_prob": 1.2052917691685039e-11,
      "compression_ratio": 1.243076923076923
    },
    {
      "start": 1113.98,
      "end": 1116.12,
      "text": "这听起来很抽象耶",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.08092220100813997,
      "no_speech_prob": 1.2052917691685039e-11,
      "compression_ratio": 1.243076923076923
    },
    {
      "start": 1116.12,
      "end": 1120.1,
      "text": "这跟一个语音转文字的模型有什么直接关系",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.08092220100813997,
      "no_speech_prob": 1.2052917691685039e-11,
      "compression_ratio": 1.243076923076923
    },
    {
      "start": 1120.1,
      "end": 1121.34,
      "text": "关系太大了",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.08092220100813997,
      "no_speech_prob": 1.2052917691685039e-11,
      "compression_ratio": 1.243076923076923
    },
    {
      "start": 1121.34,
      "end": 1123.92,
      "text": "如果AI要成为你最贴身",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.09377633119240786,
      "no_speech_prob": 1.2659013594318314e-11,
      "compression_ratio": 1.2822822822822824
    },
    {
      "start": 1123.92,
      "end": 1125.42,
      "text": "最懂你的个人助理",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.09377633119240786,
      "no_speech_prob": 1.2659013594318314e-11,
      "compression_ratio": 1.2822822822822824
    },
    {
      "start": 1125.42,
      "end": 1126.72,
      "text": "它就不能只是一个",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.09377633119240786,
      "no_speech_prob": 1.2659013594318314e-11,
      "compression_ratio": 1.2822822822822824
    },
    {
      "start": 1126.72,
      "end": 1127.9,
      "text": "你对着手机大喊",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.09377633119240786,
      "no_speech_prob": 1.2659013594318314e-11,
      "compression_ratio": 1.2822822822822824
    },
    {
      "start": 1127.9,
      "end": 1129.22,
      "text": "嘿帮我设定闹钟",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.09377633119240786,
      "no_speech_prob": 1.2659013594318314e-11,
      "compression_ratio": 1.2822822822822824
    },
    {
      "start": 1129.22,
      "end": 1130.9,
      "text": "那种死板的声控工具",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.09377633119240786,
      "no_speech_prob": 1.2659013594318314e-11,
      "compression_ratio": 1.2822822822822824
    },
    {
      "start": 1130.9,
      "end": 1133.58,
      "text": "Meta的最终目标是硬体",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.09377633119240786,
      "no_speech_prob": 1.2659013594318314e-11,
      "compression_ratio": 1.2822822822822824
    },
    {
      "start": 1133.58,
      "end": 1134.84,
      "text": "特别是他们正在",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.09377633119240786,
      "no_speech_prob": 1.2659013594318314e-11,
      "compression_ratio": 1.2822822822822824
    },
    {
      "start": 1134.84,
      "end": 1136.74,
      "text": "大力发展的AI视觉眼镜",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.09377633119240786,
      "no_speech_prob": 1.2659013594318314e-11,
      "compression_ratio": 1.2822822822822824
    },
    {
      "start": 1136.74,
      "end": 1138.2,
      "text": "哦 AI眼镜",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.09377633119240786,
      "no_speech_prob": 1.2659013594318314e-11,
      "compression_ratio": 1.2822822822822824
    },
    {
      "start": 1138.2,
      "end": 1138.96,
      "text": "你试想一下",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.09377633119240786,
      "no_speech_prob": 1.2659013594318314e-11,
      "compression_ratio": 1.2822822822822824
    },
    {
      "start": 1138.96,
      "end": 1140.66,
      "text": "你戴着AI眼镜的一天好了",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.09377633119240786,
      "no_speech_prob": 1.2659013594318314e-11,
      "compression_ratio": 1.2822822822822824
    },
    {
      "start": 1140.66,
      "end": 1142.84,
      "text": "你走在台北吵杂的街头",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.09377633119240786,
      "no_speech_prob": 1.2659013594318314e-11,
      "compression_ratio": 1.2822822822822824
    },
    {
      "start": 1142.84,
      "end": 1144.26,
      "text": "旁边有车流声",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.09377633119240786,
      "no_speech_prob": 1.2659013594318314e-11,
      "compression_ratio": 1.2822822822822824
    },
    {
      "start": 1144.26,
      "end": 1146.26,
      "text": "你正在点一杯珍珠奶茶",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.09377633119240786,
      "no_speech_prob": 1.2659013594318314e-11,
      "compression_ratio": 1.2822822822822824
    },
    {
      "start": 1146.26,
      "end": 1148.16,
      "text": "店员问你要几分甜",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.09377633119240786,
      "no_speech_prob": 1.2659013594318314e-11,
      "compression_ratio": 1.2822822822822824
    },
    {
      "start": 1148.16,
      "end": 1149.7,
      "text": "这时你旁边的朋友",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.09377633119240786,
      "no_speech_prob": 1.2659013594318314e-11,
      "compression_ratio": 1.2822822822822824
    },
    {
      "start": 1149.7,
      "end": 1150.78,
      "text": "突然插花问你",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.09377633119240786,
      "no_speech_prob": 1.2659013594318314e-11,
      "compression_ratio": 1.2822822822822824
    },
    {
      "start": 1150.78,
      "end": 1151.94,
      "text": "等下要去哪里",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.09655422168773609,
      "no_speech_prob": 1.9158622605042197e-11,
      "compression_ratio": 1.206798866855524
    },
    {
      "start": 1151.94,
      "end": 1154.2,
      "text": "哇 这个场景非常真实",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.09655422168773609,
      "no_speech_prob": 1.9158622605042197e-11,
      "compression_ratio": 1.206798866855524
    },
    {
      "start": 1154.2,
      "end": 1155.62,
      "text": "也非常混乱",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.09655422168773609,
      "no_speech_prob": 1.9158622605042197e-11,
      "compression_ratio": 1.206798866855524
    },
    {
      "start": 1155.62,
      "end": 1157.58,
      "text": "对 如果你的AI眼镜",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.09655422168773609,
      "no_speech_prob": 1.9158622605042197e-11,
      "compression_ratio": 1.206798866855524
    },
    {
      "start": 1157.58,
      "end": 1159.0,
      "text": "没有完美的机器听觉",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.09655422168773609,
      "no_speech_prob": 1.9158622605042197e-11,
      "compression_ratio": 1.206798866855524
    },
    {
      "start": 1159.0,
      "end": 1160.1,
      "text": "它就只是一个",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.09655422168773609,
      "no_speech_prob": 1.9158622605042197e-11,
      "compression_ratio": 1.206798866855524
    },
    {
      "start": 1160.1,
      "end": 1161.92,
      "text": "戴在脸上的笨重摄影机",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.09655422168773609,
      "no_speech_prob": 1.9158622605042197e-11,
      "compression_ratio": 1.206798866855524
    },
    {
      "start": 1161.92,
      "end": 1163.96,
      "text": "但有了Muse Voice Transcribe",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.09655422168773609,
      "no_speech_prob": 1.9158622605042197e-11,
      "compression_ratio": 1.206798866855524
    },
    {
      "start": 1163.96,
      "end": 1166.02,
      "text": "眼睛能在0.16秒内",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.09655422168773609,
      "no_speech_prob": 1.9158622605042197e-11,
      "compression_ratio": 1.206798866855524
    },
    {
      "start": 1166.02,
      "end": 1167.82,
      "text": "分辨出哪句是电源说的",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.09655422168773609,
      "no_speech_prob": 1.9158622605042197e-11,
      "compression_ratio": 1.206798866855524
    },
    {
      "start": 1167.82,
      "end": 1169.42,
      "text": "哪句是朋友说的",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.09655422168773609,
      "no_speech_prob": 1.9158622605042197e-11,
      "compression_ratio": 1.206798866855524
    },
    {
      "start": 1169.42,
      "end": 1171.16,
      "text": "还能过滤掉车流声",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.09655422168773609,
      "no_speech_prob": 1.9158622605042197e-11,
      "compression_ratio": 1.206798866855524
    },
    {
      "start": 1171.16,
      "end": 1173.12,
      "text": "没错 甚至刚你转头",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.09655422168773609,
      "no_speech_prob": 1.9158622605042197e-11,
      "compression_ratio": 1.206798866855524
    },
    {
      "start": 1173.12,
      "end": 1174.92,
      "text": "忘记朋友刚才问什么的时候",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.09655422168773609,
      "no_speech_prob": 1.9158622605042197e-11,
      "compression_ratio": 1.206798866855524
    },
    {
      "start": 1174.92,
      "end": 1176.74,
      "text": "AI已经把答案准备好",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.09655422168773609,
      "no_speech_prob": 1.9158622605042197e-11,
      "compression_ratio": 1.206798866855524
    },
    {
      "start": 1176.74,
      "end": 1178.32,
      "text": "投射在你的镜片上了",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.09655422168773609,
      "no_speech_prob": 1.9158622605042197e-11,
      "compression_ratio": 1.206798866855524
    },
    {
      "start": 1178.32,
      "end": 1180.04,
      "text": "哇 这画面感很强耶",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.09655422168773609,
      "no_speech_prob": 1.9158622605042197e-11,
      "compression_ratio": 1.206798866855524
    },
    {
      "start": 1180.04,
      "end": 1184.22,
      "text": "这就是为什么Meta现在让你只要在Mac电脑上按住FN键",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.07923281192779541,
      "no_speech_prob": 1.2909641237957015e-11,
      "compression_ratio": 1.2310030395136777
    },
    {
      "start": 1184.22,
      "end": 1187.8,
      "text": "就能在任何视窗进行痊愈的及时听写",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.07923281192779541,
      "no_speech_prob": 1.2909641237957015e-11,
      "compression_ratio": 1.2310030395136777
    },
    {
      "start": 1187.8,
      "end": 1190.2,
      "text": "这不是一个单纯的打字工具",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.07923281192779541,
      "no_speech_prob": 1.2909641237957015e-11,
      "compression_ratio": 1.2310030395136777
    },
    {
      "start": 1190.2,
      "end": 1193.78,
      "text": "这是Meta在训练你习惯与AI无缝对话",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.07923281192779541,
      "no_speech_prob": 1.2909641237957015e-11,
      "compression_ratio": 1.2310030395136777
    },
    {
      "start": 1193.78,
      "end": 1198.32,
      "text": "为未来的穿戴式硬体生态系铺设最底层的基础设施",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.07923281192779541,
      "no_speech_prob": 1.2909641237957015e-11,
      "compression_ratio": 1.2310030395136777
    },
    {
      "start": 1198.32,
      "end": 1202.28,
      "text": "今天的探讨真的把很多点都串起来了",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.07923281192779541,
      "no_speech_prob": 1.2909641237957015e-11,
      "compression_ratio": 1.2310030395136777
    },
    {
      "start": 1202.28,
      "end": 1205.8,
      "text": "我们现在正在见证一个历史性的转捩点",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.07923281192779541,
      "no_speech_prob": 1.2909641237957015e-11,
      "compression_ratio": 1.2310030395136777
    },
    {
      "start": 1205.8,
      "end": 1209.88,
      "text": "机器在复杂环境下的听觉处理能力",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.07923281192779541,
      "no_speech_prob": 1.2909641237957015e-11,
      "compression_ratio": 1.2310030395136777
    },
    {
      "start": 1209.88,
      "end": 1213.32,
      "text": "正在以前所未有的速度超越人类",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.06832064880183869,
      "no_speech_prob": 1.212039236336837e-11,
      "compression_ratio": 1.231578947368421
    },
    {
      "start": 1213.32,
      "end": 1214.62,
      "text": "对,真的很快",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.06832064880183869,
      "no_speech_prob": 1.212039236336837e-11,
      "compression_ratio": 1.231578947368421
    },
    {
      "start": 1214.62,
      "end": 1218.78,
      "text": "他能毫不犹豫地在0.16秒内处理资讯",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.06832064880183869,
      "no_speech_prob": 1.212039236336837e-11,
      "compression_ratio": 1.231578947368421
    },
    {
      "start": 1218.78,
      "end": 1222.12,
      "text": "他能同时分辨十几个人的声音",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.06832064880183869,
      "no_speech_prob": 1.212039236336837e-11,
      "compression_ratio": 1.231578947368421
    },
    {
      "start": 1222.12,
      "end": 1224.48,
      "text": "他甚至学会了像人类一样",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.06832064880183869,
      "no_speech_prob": 1.212039236336837e-11,
      "compression_ratio": 1.231578947368421
    },
    {
      "start": 1224.48,
      "end": 1227.42,
      "text": "在困难的词汇上稍微犹豫一下",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.06832064880183869,
      "no_speech_prob": 1.212039236336837e-11,
      "compression_ratio": 1.231578947368421
    },
    {
      "start": 1227.42,
      "end": 1231.44,
      "text": "没错,从过去笨重的三步骤拼装车系统",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.06832064880183869,
      "no_speech_prob": 1.212039236336837e-11,
      "compression_ratio": 1.231578947368421
    },
    {
      "start": 1231.44,
      "end": 1234.56,
      "text": "进化到现在单一串流的超级模型",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.06832064880183869,
      "no_speech_prob": 1.212039236336837e-11,
      "compression_ratio": 1.231578947368421
    },
    {
      "start": 1234.56,
      "end": 1237.84,
      "text": "这不仅仅是技术指标上的胜利而已",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.06832064880183869,
      "no_speech_prob": 1.212039236336837e-11,
      "compression_ratio": 1.231578947368421
    },
    {
      "start": 1237.84,
      "end": 1241.72,
      "text": "更是人机互动方式的一次彻底革命",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.08197396356354719,
      "no_speech_prob": 1.1931359546191178e-11,
      "compression_ratio": 1.3141891891891893
    },
    {
      "start": 1241.72,
      "end": 1242.9,
      "text": "真的是革命",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.08197396356354719,
      "no_speech_prob": 1.1931359546191178e-11,
      "compression_ratio": 1.3141891891891893
    },
    {
      "start": 1242.9,
      "end": 1246.62,
      "text": "虽然目前还像是繁体中文资源不足这样的缺点",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.08197396356354719,
      "no_speech_prob": 1.1931359546191178e-11,
      "compression_ratio": 1.3141891891891893
    },
    {
      "start": 1246.62,
      "end": 1249.72,
      "text": "但这项技术为未来AI眼镜",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.08197396356354719,
      "no_speech_prob": 1.1931359546191178e-11,
      "compression_ratio": 1.3141891891891893
    },
    {
      "start": 1249.72,
      "end": 1252.6,
      "text": "和无缝语音互动所打下的基础",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.08197396356354719,
      "no_speech_prob": 1.1931359546191178e-11,
      "compression_ratio": 1.3141891891891893
    },
    {
      "start": 1252.6,
      "end": 1255.9,
      "text": "其实已经预示了科技的下一步走向了",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.08197396356354719,
      "no_speech_prob": 1.1931359546191178e-11,
      "compression_ratio": 1.3141891891891893
    },
    {
      "start": 1255.9,
      "end": 1256.22,
      "text": "真的",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.08197396356354719,
      "no_speech_prob": 1.1931359546191178e-11,
      "compression_ratio": 1.3141891891891893
    },
    {
      "start": 1256.22,
      "end": 1259.28,
      "text": "在我们结束今天的深度探讨之前",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.08197396356354719,
      "no_speech_prob": 1.1931359546191178e-11,
      "compression_ratio": 1.3141891891891893
    },
    {
      "start": 1259.28,
      "end": 1262.56,
      "text": "我想留给你们一个发人深审的问题",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.08197396356354719,
      "no_speech_prob": 1.1931359546191178e-11,
      "compression_ratio": 1.3141891891891893
    },
    {
      "start": 1262.56,
      "end": 1264.08,
      "text": "想象一下",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.08197396356354719,
      "no_speech_prob": 1.1931359546191178e-11,
      "compression_ratio": 1.3141891891891893
    },
    {
      "start": 1264.08,
      "end": 1266.32,
      "text": "如果未来的每一台装置",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.08197396356354719,
      "no_speech_prob": 1.1931359546191178e-11,
      "compression_ratio": 1.3141891891891893
    },
    {
      "start": 1266.32,
      "end": 1267.7,
      "text": "每一副眼镜",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.08197396356354719,
      "no_speech_prob": 1.1931359546191178e-11,
      "compression_ratio": 1.3141891891891893
    },
    {
      "start": 1267.7,
      "end": 1269.8,
      "text": "都能在0.16秒内",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.0950772212101863,
      "no_speech_prob": 1.1039175649985022e-11,
      "compression_ratio": 1.2614840989399294
    },
    {
      "start": 1269.8,
      "end": 1271.3,
      "text": "完美记录下",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.0950772212101863,
      "no_speech_prob": 1.1039175649985022e-11,
      "compression_ratio": 1.2614840989399294
    },
    {
      "start": 1271.3,
      "end": 1273.16,
      "text": "20个人在会议室里",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.0950772212101863,
      "no_speech_prob": 1.1039175649985022e-11,
      "compression_ratio": 1.2614840989399294
    },
    {
      "start": 1273.16,
      "end": 1274.16,
      "text": "互相强化",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.0950772212101863,
      "no_speech_prob": 1.1039175649985022e-11,
      "compression_ratio": 1.2614840989399294
    },
    {
      "start": 1274.16,
      "end": 1276.76,
      "text": "甚至私下碎碎念的内容",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.0950772212101863,
      "no_speech_prob": 1.1039175649985022e-11,
      "compression_ratio": 1.2614840989399294
    },
    {
      "start": 1276.76,
      "end": 1278.46,
      "text": "并精准标记",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.0950772212101863,
      "no_speech_prob": 1.1039175649985022e-11,
      "compression_ratio": 1.2614840989399294
    },
    {
      "start": 1278.46,
      "end": 1279.74,
      "text": "到底是谁说的",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.0950772212101863,
      "no_speech_prob": 1.1039175649985022e-11,
      "compression_ratio": 1.2614840989399294
    },
    {
      "start": 1279.74,
      "end": 1281.28,
      "text": "这对我们的社会",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.0950772212101863,
      "no_speech_prob": 1.1039175649985022e-11,
      "compression_ratio": 1.2614840989399294
    },
    {
      "start": 1281.28,
      "end": 1282.52,
      "text": "到底意味着什么",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.0950772212101863,
      "no_speech_prob": 1.1039175649985022e-11,
      "compression_ratio": 1.2614840989399294
    },
    {
      "start": 1282.52,
      "end": 1283.06,
      "text": "嗯",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.0950772212101863,
      "no_speech_prob": 1.1039175649985022e-11,
      "compression_ratio": 1.2614840989399294
    },
    {
      "start": 1283.06,
      "end": 1283.96,
      "text": "这蛮可怕的",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.0950772212101863,
      "no_speech_prob": 1.1039175649985022e-11,
      "compression_ratio": 1.2614840989399294
    },
    {
      "start": 1283.96,
      "end": 1284.62,
      "text": "对吧",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.0950772212101863,
      "no_speech_prob": 1.1039175649985022e-11,
      "compression_ratio": 1.2614840989399294
    },
    {
      "start": 1284.62,
      "end": 1286.26,
      "text": "当过去那种",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.0950772212101863,
      "no_speech_prob": 1.1039175649985022e-11,
      "compression_ratio": 1.2614840989399294
    },
    {
      "start": 1286.26,
      "end": 1288.86,
      "text": "听过就算了的随性闲聊",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.0950772212101863,
      "no_speech_prob": 1.1039175649985022e-11,
      "compression_ratio": 1.2614840989399294
    },
    {
      "start": 1288.86,
      "end": 1291.38,
      "text": "全部变成可以被精准搜寻",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.0950772212101863,
      "no_speech_prob": 1.1039175649985022e-11,
      "compression_ratio": 1.2614840989399294
    },
    {
      "start": 1291.38,
      "end": 1292.78,
      "text": "被永久记录的",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.0950772212101863,
      "no_speech_prob": 1.1039175649985022e-11,
      "compression_ratio": 1.2614840989399294
    },
    {
      "start": 1292.78,
      "end": 1295.1,
      "text": "完美数位逐字稿的时候",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.0950772212101863,
      "no_speech_prob": 1.1039175649985022e-11,
      "compression_ratio": 1.2614840989399294
    },
    {
      "start": 1295.1,
      "end": 1297.5,
      "text": "这会让我们的工作与生活",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.0950772212101863,
      "no_speech_prob": 1.1039175649985022e-11,
      "compression_ratio": 1.2614840989399294
    },
    {
      "start": 1297.5,
      "end": 1298.84,
      "text": "变得极度高效",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.10731487712640872,
      "no_speech_prob": 1.5939100733719513e-11,
      "compression_ratio": 1.1105263157894736
    },
    {
      "start": 1298.84,
      "end": 1301.34,
      "text": "还是会让我们在每次开口之前",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.10731487712640872,
      "no_speech_prob": 1.5939100733719513e-11,
      "compression_ratio": 1.1105263157894736
    },
    {
      "start": 1301.34,
      "end": 1303.82,
      "text": "都变得极度焦虑多疑呢",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.10731487712640872,
      "no_speech_prob": 1.5939100733719513e-11,
      "compression_ratio": 1.1105263157894736
    },
    {
      "start": 1303.82,
      "end": 1306.24,
      "text": "这是一个值得你在关掉节目后",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.10731487712640872,
      "no_speech_prob": 1.5939100733719513e-11,
      "compression_ratio": 1.1105263157894736
    },
    {
      "start": 1306.24,
      "end": 1307.86,
      "text": "好好思考的问题",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.10731487712640872,
      "no_speech_prob": 1.5939100733719513e-11,
      "compression_ratio": 1.1105263157894736
    },
    {
      "start": 1307.86,
      "end": 1309.16,
      "text": "谢谢收听",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.10731487712640872,
      "no_speech_prob": 1.5939100733719513e-11,
      "compression_ratio": 1.1105263157894736
    },
    {
      "start": 1309.16,
      "end": 1310.36,
      "text": "欢迎订阅分享",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.10731487712640872,
      "no_speech_prob": 1.5939100733719513e-11,
      "compression_ratio": 1.1105263157894736
    },
    {
      "start": 1310.36,
      "end": 1311.6,
      "text": "听书Podcast频道",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.10731487712640872,
      "no_speech_prob": 1.5939100733719513e-11,
      "compression_ratio": 1.1105263157894736
    },
    {
      "start": 1311.6,
      "end": 1312.62,
      "text": "我们下次见",
      "chunk": 2,
      "language": "zh",
      "avg_logprob": -0.10731487712640872,
      "no_speech_prob": 1.5939100733719513e-11,
      "compression_ratio": 1.1105263157894736
    }
  ],
  "segmented_transcription": {
    "status": "SEGMENTED_TRANSCRIPT_OK",
    "created_at": "2026-09-07T16:23:51",
    "manifest": "/Users/sagawa/AI_WORK/video_notes/url/20260907_162302__Muse_Voice_Transcribe-Meta_即時語音感知模型/segmented_work/chunk_manifest.json",
    "audio": "/Users/sagawa/AI_WORK/video_notes/url/20260907_162302__Muse_Voice_Transcribe-Meta_即時語音感知模型/audio_16k_mono.wav",
    "audio_sha256": "abb05d0eedc77a1fe88221f4c5a31beb78b667c3670a64e34e16d9abe8fa97a0",
    "audio_duration": 1312.750313,
    "model": "/Users/sagawa/AI_WORK/cache/huggingface/hub/models--mlx-community--whisper-large-v3-turbo/snapshots/a4aaeec0636e6fef84abdcbe3544cb2bf7e9f6fb",
    "language_hint": "zh",
    "condition_on_previous_text": false,
    "chunk_count": 3,
    "resumed_chunks": 0,
    "merged_segment_count": 565,
    "cleaned_segment_count": 565,
    "removed_hallucination_segment_count": 0,
    "final_segment_end": 1312.62,
    "global_repeat_run": 2,
    "global_repeat_text": "对吧",
    "tail_repeat_run": 1,
    "tail_repeat_text": "想象一下",
    "hallucination_region_count": 0,
    "hallucination_regions": [],
    "qc_ok": true,
    "qc_errors": []
  }
}