{
  "text": "好的,现在市面上有很多优秀的语音识别模型,我们有whisper,有parakeet,还有canary。\n如果你只需要简单地转录文本,可能会觉得这已经是解决好的问题了。\n但最大的挑战在于,如果你想为大语言模型或正在构建的应用获取上下文,仅靠转录文本是不够的。\n你通常还需要时间出来知道某句话是何时说的。\n但往往更重要的是,你需要知道到底是谁说的。\n也就是说,如果你正在为四个人互相辩论的播客制作转录文本,你需要知道谁说了什么,谁持哪方观点等。\n如果你想做会议纪要之类的东西,道理也是一样的。\n你需要知道谁同意了哪项行动,尤其是现在我们要构建智能体来处理这些事物,能够提取引语并准确归属于对应的人。\n而不是只有一份所有人都在同一通话中逐字说出的转录文本\n就像只有一个人说话是那样\n那么一直都有办法做到这一点,对吧?\n在我之前的一些视频中\n我就展示过用Whisper配合当时的一些开源工具来做到这一点\n但这一直是一场猫鼠游戏\n看你能做到多准确\n而且你还得面对如何处理\n由不同提供商制作的多个模型等挑战\n这就引出了NVIDIA刚刚发布的东西\n这是一个专门针对该问题的新模型\n这就是NEMU TRAN 3 说话人分离\n它是一个开放权重模型\n它采用的许可证允许商业使用\n凭借这种许可证\n你当然可以做大多数你想做的事\n这里酷的一点是\n它只有大约1亿参数量\n所以只需要4GB显存的GPU就能运行\n除此之外\n当我们谈论准确率时\n它可以处理多达8位说话人\n过去大多数模型的上限大概是三到四位说话人,而它不仅能在大家清晰说话时处理八位说话人。\n即使人们互相打断说话,它也能做得相当不错。\n此外,它的设计初衷是与你正在使用的任何语音识别模型配合使用。\n所以如果你有一个特别喜欢的模型,比如PireKit,你就可以基本上使用这个模型。\n稍后我会展示一个演示。\n我们将原始播客录音转换为转录文本。\n每一行都归属于对应的说话人\n但首先让我介绍一下这个模型的背景\nNVIDIA将其作为Nemotron语音模型系列的一部分推出\n在这个频道上\n我谈过Nemotron大语言模型等\n但我认为很多人没有意识到\n这一系列还包含很多语音模型\n在识别方面\n你有PowlerKit、Canary和Nemotron ASR\n他们共同支持43种语言的流逝和批量转路\n他们还有一个叫Magpie的TTS模型\n支持多语言并具备语音定制等功能\n此外,他们还有一些进行翻译\n甚至全双公语音到语音处理的模型\n比如Personaplex\n直到最近,他们的说话人分离选项叫做Sortformer\n最多支持四位说话人\n有一段时间,他实际上相当受欢迎\n在很多方面,你可以把这个新模型看作是对它的直接替换,而且好得多。\n我们知道,人们已在生产环境中使用这些模型,因为旧版Sortformer模型仅在8月就获得了超过30万次下载。\n虽然NVIDIA没有直接说明,但在我看来,他们正在系统地改进这些模型,以构建完整的NEMOTRON语音模型套件。\n这完全说得通\n因为如果你看看对语音的需求\n无论是单用户与智能体交互\n智能体获取YouTube视频等多模态内容\n以便在研究和执行任务时\n提取信息作为上下文\n需求都在不断增加\n好的\n如果你是新手\n说话人分离到底是什么\n最简单的说法是\n语音识别告诉你说了什么\n而说话人分离告诉你谁说的\n当我们把它们结合起来,就得到了所谓的带说话人信息的转录文本。\n我认为大家凭直觉都知道这非常重要。\n如果你在转录会议,有人问了一个问题,然后有些人回答是,\n有些人回答否,你大概希望你的智能体知道谁将会跟进,谁不会采取行动。\n所有这些对于智能体确定,接下来应采取何种行动的上下文都至关重要。\n那么,他会自动知道每个说话人的名字吗?\n不\n它基本上只会说说话人\n一说话人二说话人三说话人七等等\n随着处理过程进行\n但这并不难\n你可以写一些代码将其替换掉\n一旦你告诉模型谁在说话\n你就可以将转入文本转换为包含真实性的版本\n这方面一直最大的挑战之一是\n人们会互相打断\n对吧\n即使他们并不想故意打断\n当有人在会议通话中提问时\n往往会有两三个人同时回答\n旧版说话人分离系统面临的挑战在于\n他们会因此变得混乱\n他们不擅长区分不同的声音\n尤其是在声音重叠时\n我想说\n我认为没有任何模型能在这件事上做到完美\n即使是人类\n这也可能是一项艰难的任务\n但这些模型的评分方式\n是使用所谓的说话人分离错误率\n即D.E.R\n它基本上是将三种错误相加\n模型遗漏的语音\n模型误判为非语音的内容\n以及将语音错误归属于他人的次数\n显然这里的分数越低越好\n你可以看到NVIDIA的最新模型不仅击败了竞争对手,还大幅超越了他们自己的前代模型。\n这里需要理解的另一个概念是流逝与离线的区别。\n在离线模式下,模型可以查看整个录音,包括后面的内容。\n而在流逝模式下,它必须在音频到达时做出决定。\n所以,对于已经录制好的播客等,你总是希望使用离线模式。\n对于任何实时字幕或语音智能体\n你必须使用流式模式\n好的\n那么Nemotron 3说话人分离到底有什么新特点\n它大约有1亿参数量\n不包括ASR部分\n运行它的最低GPU要求约为4GB\n所以它实际上可以在许多小型GPU上运行\n而且应该在笔记本电脑的RTX显卡上\n也能跑得相当好\n我之前提到过它支持多达8位说话人\n另一点是它是语言无关的\n并且在流逝和离线模式下都能工作\n所以我认为最好的办法是直接进入演示\n看看它实际上是如何运作的\n好的\n进入演示\n让我先展示一下设置方式\n首先感谢NVIDIA赞助这里的计算资源\n我使用DGX Spark来实际运行这里的三个语音模型\n让我带大家过一遍设置架构图\n你在这里看到的是Next.js应用,它在我的Mac上本地运行。\n然后通过Talescale调用后端到DGX。\nX Spark.\n你可以看到,我在上面有一个包含NVIDIA NEMO的Docker容器。\nVDI NEMO既提供我们正在讨论的说话人分离模型,\n也提供ParaKit ASR模型和NEMOTRON 3.5多语言 ASR模型。\n为了使用这些,我设置了一个Fast API应用。\n这是一种实现此类功能的简单方法。\n如果我要大规模部署\n我可能会选择Dynamo或Triton来提供服务\n但对于我想要的功能\n目前的设置运行得非常完美\n如果我们进入这个应用\n你可以看到我基本上可以直接使用它们的演示\n我可以选择想要使用哪个AS2模型\n我打算用ParaKit\n因为它在处理多说话人方面通常表现更好\n我可以开始转录\n你可以看到字面上只需要三四秒钟\n他就处理并转录了这里一分三十多分钟的音频\n由于我之前在摆弄这个并设置了一些名字\n他实际上记住了这里这个人的名字\n让我们听一下\n看看当多个说话人同时说话时会发生什么\n为了庆祝Nemotron三说话人分离的发布\n我们邀请了另外七位开放文本转语音声音加入派对\n我是Chris,NVDI的Nemotron声音\n大家好,谢谢邀请\n我是Jane,来自QTI的Pocket声音\n很高兴见到你们,Chris和Jane\n我是Serena,Queen三的声音\n我是来自Cocoro的Adam\n只有8200万参数量\n我还是入选了嘉宾名单\n你好,你好\n我是来自Resemble AI的Chatterbox\n我带来了富有表现力的能量\n嗨,我是来自Nari Labs\n的John Dyer声音\n我们这里已经聚集了\n很棒的人群\n大家好\n最近怎么样\n我是来自Maya Research的\nMaya E声音\n好了\n现在我们有八位合成\n开放声音\n做满了场子\n让我们来测试一下\nNemotron三说话人分离\n好的\n新的说话人分离模型\n可以在转路时\n标记我们的每个声音\n但它能在我们所有人\n一起说话时\n做到这一点吗\n让我们来看看\n即使我们互相打断\n它也能叫出\n我们俩的名字\n竹针标记\n谁在何时说话\n好的\n正如你在这里看到的\n它在转录每一个声音方面做得很好\n如果我们往下看看\n可以看到每个声音都被转录了\n我想我的UI在这方面稍微有点滞后\n但你完全可以感受到它是如何组装起来的\n实际上如果我们放大一点来自QTI的声音\n很高兴见到你们\nChris和Jen\n我是Serena\nQueen3的声音\n你可以很快看到\n我们基本上可以组装出一个能转录这些声音的东西\n如果我们想在这里输入它们的名字\n它会便利并为我们重命名所有人\n现在,如果我们想看看实际模型返回的是什么\n你可以看到这里是模型返回的原始JSON\n我们可以看到这里实际使用的是什么\n我们可以看到不同的说话人被标记等\n然后这个软件基本上将Speaker三重写为ADAM等便利这个过程\n做这样的事情很酷的一点是\n我们可以很快获得统计数据\n比如每个人在那里说了多久\n说了多少次\n有多少次发言\n如果我们想在这里放入更长的内容\n这会非常有用\n好的\n这里酷的是我们可以处理长文件\n所以我基本上放入了\nAll In 播客的一集完整节目\n他们采访了黄仁勋\n你可以看到他基本上已经处理完了\n我们得到了输出\n我命名了这里的实际说话人\n所以我们可以看到\n说话人依次出现\n现在我们有了每个人\n说了什么的完整转录文本\n这里酷的是我们可以查看统计数据\n我们可以看到他们各自的说话时长\n我们可以看到,果然,显然,黄仁勋是这里的嘉宾,所以他是说话最多的,而这一切都在148秒内完成,所以对于这个超过一小时的长文件,这大约花了两分半钟。\n这让我们基本上可以便利\n你可以看到\n现在我们基本上\n可以直接阅读转录文本\n我可以将其导出为文本文件\n或者在这里导出为srt\n这里酷的是\n如果我确实想做多语言\n或其他类似的事情\n我可以切换模型\n使用这个模型\n而不是这里的paraket模型\n你可以看到\n用这个模型\n我们支持很多不同的语言\n如果你记得我做过\n关于这个的视频\n你实际上可以微调这个模型\n使其在其他语言上表现非常好\n他们在这里称之为适配就绪\n现在我还没有真正设置这个应用去做\n但模型能做的另一件事\n是你也可以进行流逝响应\n当然\n那样你的准确率可能会下降一点\n但你在那里\n仍然应该能获得相当好的结果\n所以\n如果你有兴趣做任何语音转录\n并且希望其中包含说话人分离等\n这是一个很好的设置方法\n你可以批量完成所有转路等\n最重要的是\n能够完全在本地完成所有这一切\n所以\n如果你对这方面有任何问题\n请在评论区告诉我\n我认为看到NVIDIA\n接下来在这些语音模型上\n会做什么将会非常酷\n就像我说的\n从我看到的来看\n它们不断迭代\n以改进这些模型\n并为不同的用力\n提供不同类型的模型\n总之\n一如既往\n如果你觉得视频有用\n请点赞并订阅\n我们下个视频见\n再见",
  "language": "multilingual",
  "duration": 789.24625,
  "segments": [
    {
      "start": 0.0,
      "end": 7.1,
      "text": "好的,现在市面上有很多优秀的语音识别模型,我们有whisper,有parakeet,还有canary。",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 0,
      "language_run": 0,
      "source": "whisper"
    },
    {
      "start": 9.417,
      "end": 14.307,
      "text": "如果你只需要简单地转录文本,可能会觉得这已经是解决好的问题了。",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 1,
      "language_run": 1,
      "source": "whisper"
    },
    {
      "start": 14.807,
      "end": 23.727,
      "text": "但最大的挑战在于,如果你想为大语言模型或正在构建的应用获取上下文,仅靠转录文本是不够的。",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 1,
      "language_run": 1,
      "source": "whisper"
    },
    {
      "start": 24.207,
      "end": 28.107,
      "text": "你通常还需要时间出来知道某句话是何时说的。",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 1,
      "language_run": 1,
      "source": "whisper"
    },
    {
      "start": 28.187,
      "end": 32.944,
      "text": "但往往更重要的是,你需要知道到底是谁说的。",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 1,
      "language_run": 1,
      "source": "whisper"
    },
    {
      "start": 34.374,
      "end": 43.944,
      "text": "也就是说,如果你正在为四个人互相辩论的播客制作转录文本,你需要知道谁说了什么,谁持哪方观点等。",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 2,
      "language_run": 1,
      "source": "whisper"
    },
    {
      "start": 44.084,
      "end": 47.644,
      "text": "如果你想做会议纪要之类的东西,道理也是一样的。",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 2,
      "language_run": 1,
      "source": "whisper"
    },
    {
      "start": 48.164,
      "end": 58.644,
      "text": "你需要知道谁同意了哪项行动,尤其是现在我们要构建智能体来处理这些事物,能够提取引语并准确归属于对应的人。",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 2,
      "language_run": 1,
      "source": "whisper"
    },
    {
      "start": 58.644,
      "end": 65.224,
      "text": "而不是只有一份所有人都在同一通话中逐字说出的转录文本",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 2,
      "language_run": 1,
      "source": "whisper"
    },
    {
      "start": 65.224,
      "end": 67.364,
      "text": "就像只有一个人说话是那样",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 2,
      "language_run": 1,
      "source": "whisper"
    },
    {
      "start": 67.364,
      "end": 70.004,
      "text": "那么一直都有办法做到这一点,对吧?",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 2,
      "language_run": 1,
      "source": "whisper"
    },
    {
      "start": 70.724,
      "end": 72.084,
      "text": "在我之前的一些视频中",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 2,
      "language_run": 1,
      "source": "whisper"
    },
    {
      "start": 72.084,
      "end": 76.584,
      "text": "我就展示过用Whisper配合当时的一些开源工具来做到这一点",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 2,
      "language_run": 1,
      "source": "whisper"
    },
    {
      "start": 76.584,
      "end": 78.724,
      "text": "但这一直是一场猫鼠游戏",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 2,
      "language_run": 1,
      "source": "whisper"
    },
    {
      "start": 78.724,
      "end": 80.124,
      "text": "看你能做到多准确",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 2,
      "language_run": 1,
      "source": "whisper"
    },
    {
      "start": 80.124,
      "end": 83.184,
      "text": "而且你还得面对如何处理",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 2,
      "language_run": 1,
      "source": "whisper"
    },
    {
      "start": 83.184,
      "end": 86.118,
      "text": "由不同提供商制作的多个模型等挑战",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 2,
      "language_run": 1,
      "source": "whisper"
    },
    {
      "start": 87.504,
      "end": 90.734,
      "text": "这就引出了NVIDIA刚刚发布的东西",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 3,
      "language_run": 1,
      "source": "whisper"
    },
    {
      "start": 90.734,
      "end": 93.694,
      "text": "这是一个专门针对该问题的新模型",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 3,
      "language_run": 1,
      "source": "whisper"
    },
    {
      "start": 93.694,
      "end": 96.554,
      "text": "这就是NEMU TRAN 3 说话人分离",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 3,
      "language_run": 1,
      "source": "whisper"
    },
    {
      "start": 96.554,
      "end": 98.234,
      "text": "它是一个开放权重模型",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 3,
      "language_run": 1,
      "source": "whisper"
    },
    {
      "start": 98.234,
      "end": 100.954,
      "text": "它采用的许可证允许商业使用",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 3,
      "language_run": 1,
      "source": "whisper"
    },
    {
      "start": 100.954,
      "end": 102.214,
      "text": "凭借这种许可证",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 3,
      "language_run": 1,
      "source": "whisper"
    },
    {
      "start": 102.214,
      "end": 104.474,
      "text": "你当然可以做大多数你想做的事",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 3,
      "language_run": 1,
      "source": "whisper"
    },
    {
      "start": 104.474,
      "end": 105.914,
      "text": "这里酷的一点是",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 3,
      "language_run": 1,
      "source": "whisper"
    },
    {
      "start": 105.914,
      "end": 107.874,
      "text": "它只有大约1亿参数量",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 3,
      "language_run": 1,
      "source": "whisper"
    },
    {
      "start": 107.874,
      "end": 111.754,
      "text": "所以只需要4GB显存的GPU就能运行",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 3,
      "language_run": 1,
      "source": "whisper"
    },
    {
      "start": 111.754,
      "end": 112.654,
      "text": "除此之外",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 3,
      "language_run": 1,
      "source": "whisper"
    },
    {
      "start": 112.654,
      "end": 114.654,
      "text": "当我们谈论准确率时",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 3,
      "language_run": 1,
      "source": "whisper"
    },
    {
      "start": 114.654,
      "end": 116.873,
      "text": "它可以处理多达8位说话人",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 3,
      "language_run": 1,
      "source": "whisper"
    },
    {
      "start": 118.296,
      "end": 126.746,
      "text": "过去大多数模型的上限大概是三到四位说话人,而它不仅能在大家清晰说话时处理八位说话人。",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 4,
      "language_run": 1,
      "source": "whisper"
    },
    {
      "start": 127.786,
      "end": 130.766,
      "text": "即使人们互相打断说话,它也能做得相当不错。",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 4,
      "language_run": 1,
      "source": "whisper"
    },
    {
      "start": 131.406,
      "end": 136.246,
      "text": "此外,它的设计初衷是与你正在使用的任何语音识别模型配合使用。",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 4,
      "language_run": 1,
      "source": "whisper"
    },
    {
      "start": 136.686,
      "end": 142.046,
      "text": "所以如果你有一个特别喜欢的模型,比如PireKit,你就可以基本上使用这个模型。",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 4,
      "language_run": 1,
      "source": "whisper"
    },
    {
      "start": 142.046,
      "end": 143.946,
      "text": "稍后我会展示一个演示。",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 4,
      "language_run": 1,
      "source": "whisper"
    },
    {
      "start": 144.046,
      "end": 147.286,
      "text": "我们将原始播客录音转换为转录文本。",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 4,
      "language_run": 1,
      "source": "whisper"
    },
    {
      "start": 147.286,
      "end": 150.346,
      "text": "每一行都归属于对应的说话人",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 4,
      "language_run": 1,
      "source": "whisper"
    },
    {
      "start": 150.346,
      "end": 153.866,
      "text": "但首先让我介绍一下这个模型的背景",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 4,
      "language_run": 1,
      "source": "whisper"
    },
    {
      "start": 153.866,
      "end": 158.406,
      "text": "NVIDIA将其作为Nemotron语音模型系列的一部分推出",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 4,
      "language_run": 1,
      "source": "whisper"
    },
    {
      "start": 158.406,
      "end": 159.726,
      "text": "在这个频道上",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 4,
      "language_run": 1,
      "source": "whisper"
    },
    {
      "start": 159.726,
      "end": 162.406,
      "text": "我谈过Nemotron大语言模型等",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 4,
      "language_run": 1,
      "source": "whisper"
    },
    {
      "start": 162.406,
      "end": 164.866,
      "text": "但我认为很多人没有意识到",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 4,
      "language_run": 1,
      "source": "whisper"
    },
    {
      "start": 164.866,
      "end": 167.846,
      "text": "这一系列还包含很多语音模型",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 4,
      "language_run": 1,
      "source": "whisper"
    },
    {
      "start": 167.846,
      "end": 169.066,
      "text": "在识别方面",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 4,
      "language_run": 1,
      "source": "whisper"
    },
    {
      "start": 169.066,
      "end": 172.261,
      "text": "你有PowlerKit、Canary和Nemotron ASR",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 4,
      "language_run": 1,
      "source": "whisper"
    },
    {
      "start": 175.053,
      "end": 179.063,
      "text": "他们共同支持43种语言的流逝和批量转路",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 5,
      "language_run": 2,
      "source": "whisper"
    },
    {
      "start": 179.063,
      "end": 182.223,
      "text": "他们还有一个叫Magpie的TTS模型",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 5,
      "language_run": 2,
      "source": "whisper"
    },
    {
      "start": 182.223,
      "end": 185.423,
      "text": "支持多语言并具备语音定制等功能",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 5,
      "language_run": 2,
      "source": "whisper"
    },
    {
      "start": 185.423,
      "end": 187.663,
      "text": "此外,他们还有一些进行翻译",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 5,
      "language_run": 2,
      "source": "whisper"
    },
    {
      "start": 187.663,
      "end": 190.403,
      "text": "甚至全双公语音到语音处理的模型",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 5,
      "language_run": 2,
      "source": "whisper"
    },
    {
      "start": 190.403,
      "end": 192.503,
      "text": "比如Personaplex",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 5,
      "language_run": 2,
      "source": "whisper"
    },
    {
      "start": 192.503,
      "end": 197.003,
      "text": "直到最近,他们的说话人分离选项叫做Sortformer",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 5,
      "language_run": 2,
      "source": "whisper"
    },
    {
      "start": 197.003,
      "end": 198.903,
      "text": "最多支持四位说话人",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 5,
      "language_run": 2,
      "source": "whisper"
    },
    {
      "start": 198.903,
      "end": 202.463,
      "text": "有一段时间,他实际上相当受欢迎",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 5,
      "language_run": 2,
      "source": "whisper"
    },
    {
      "start": 202.463,
      "end": 207.763,
      "text": "在很多方面,你可以把这个新模型看作是对它的直接替换,而且好得多。",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 5,
      "language_run": 2,
      "source": "whisper"
    },
    {
      "start": 208.063,
      "end": 216.023,
      "text": "我们知道,人们已在生产环境中使用这些模型,因为旧版Sortformer模型仅在8月就获得了超过30万次下载。",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 5,
      "language_run": 2,
      "source": "whisper"
    },
    {
      "start": 216.363,
      "end": 226.071,
      "text": "虽然NVIDIA没有直接说明,但在我看来,他们正在系统地改进这些模型,以构建完整的NEMOTRON语音模型套件。",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 5,
      "language_run": 2,
      "source": "whisper"
    },
    {
      "start": 227.72,
      "end": 228.81,
      "text": "这完全说得通",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 6,
      "language_run": 2,
      "source": "whisper"
    },
    {
      "start": 228.81,
      "end": 231.69,
      "text": "因为如果你看看对语音的需求",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 6,
      "language_run": 2,
      "source": "whisper"
    },
    {
      "start": 233.295,
      "end": 235.645,
      "text": "无论是单用户与智能体交互",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 7,
      "language_run": 2,
      "source": "whisper"
    },
    {
      "start": 236.783,
      "end": 241.933,
      "text": "智能体获取YouTube视频等多模态内容",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 8,
      "language_run": 2,
      "source": "whisper"
    },
    {
      "start": 243.745,
      "end": 246.165,
      "text": "以便在研究和执行任务时",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 9,
      "language_run": 2,
      "source": "whisper"
    },
    {
      "start": 246.165,
      "end": 247.865,
      "text": "提取信息作为上下文",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 9,
      "language_run": 2,
      "source": "whisper"
    },
    {
      "start": 247.865,
      "end": 249.765,
      "text": "需求都在不断增加",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 9,
      "language_run": 2,
      "source": "whisper"
    },
    {
      "start": 249.765,
      "end": 250.125,
      "text": "好的",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 9,
      "language_run": 2,
      "source": "whisper"
    },
    {
      "start": 250.125,
      "end": 251.565,
      "text": "如果你是新手",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 9,
      "language_run": 2,
      "source": "whisper"
    },
    {
      "start": 251.565,
      "end": 253.485,
      "text": "说话人分离到底是什么",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 9,
      "language_run": 2,
      "source": "whisper"
    },
    {
      "start": 253.485,
      "end": 254.965,
      "text": "最简单的说法是",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 9,
      "language_run": 2,
      "source": "whisper"
    },
    {
      "start": 254.965,
      "end": 257.285,
      "text": "语音识别告诉你说了什么",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 9,
      "language_run": 2,
      "source": "whisper"
    },
    {
      "start": 257.285,
      "end": 259.909,
      "text": "而说话人分离告诉你谁说的",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 9,
      "language_run": 2,
      "source": "whisper"
    },
    {
      "start": 261.623,
      "end": 266.713,
      "text": "当我们把它们结合起来,就得到了所谓的带说话人信息的转录文本。",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 10,
      "language_run": 2,
      "source": "whisper"
    },
    {
      "start": 267.193,
      "end": 270.093,
      "text": "我认为大家凭直觉都知道这非常重要。",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 10,
      "language_run": 2,
      "source": "whisper"
    },
    {
      "start": 270.573,
      "end": 274.613,
      "text": "如果你在转录会议,有人问了一个问题,然后有些人回答是,",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 10,
      "language_run": 2,
      "source": "whisper"
    },
    {
      "start": 275.153,
      "end": 281.993,
      "text": "有些人回答否,你大概希望你的智能体知道谁将会跟进,谁不会采取行动。",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 10,
      "language_run": 2,
      "source": "whisper"
    },
    {
      "start": 282.373,
      "end": 288.113,
      "text": "所有这些对于智能体确定,接下来应采取何种行动的上下文都至关重要。",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 10,
      "language_run": 2,
      "source": "whisper"
    },
    {
      "start": 288.113,
      "end": 290.493,
      "text": "那么,他会自动知道每个说话人的名字吗?",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 10,
      "language_run": 2,
      "source": "whisper"
    },
    {
      "start": 290.493,
      "end": 291.333,
      "text": "不",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 10,
      "language_run": 2,
      "source": "whisper"
    },
    {
      "start": 291.333,
      "end": 293.233,
      "text": "它基本上只会说说话人",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 10,
      "language_run": 2,
      "source": "whisper"
    },
    {
      "start": 293.233,
      "end": 297.833,
      "text": "一说话人二说话人三说话人七等等",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 10,
      "language_run": 2,
      "source": "whisper"
    },
    {
      "start": 297.833,
      "end": 299.653,
      "text": "随着处理过程进行",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 10,
      "language_run": 2,
      "source": "whisper"
    },
    {
      "start": 299.653,
      "end": 301.193,
      "text": "但这并不难",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 10,
      "language_run": 2,
      "source": "whisper"
    },
    {
      "start": 301.193,
      "end": 303.333,
      "text": "你可以写一些代码将其替换掉",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 10,
      "language_run": 2,
      "source": "whisper"
    },
    {
      "start": 303.333,
      "end": 305.473,
      "text": "一旦你告诉模型谁在说话",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 10,
      "language_run": 2,
      "source": "whisper"
    },
    {
      "start": 305.473,
      "end": 308.533,
      "text": "你就可以将转入文本转换为包含真实性的版本",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 10,
      "language_run": 2,
      "source": "whisper"
    },
    {
      "start": 308.533,
      "end": 310.413,
      "text": "这方面一直最大的挑战之一是",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 10,
      "language_run": 2,
      "source": "whisper"
    },
    {
      "start": 310.413,
      "end": 311.973,
      "text": "人们会互相打断",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 10,
      "language_run": 2,
      "source": "whisper"
    },
    {
      "start": 311.973,
      "end": 312.593,
      "text": "对吧",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 10,
      "language_run": 2,
      "source": "whisper"
    },
    {
      "start": 312.593,
      "end": 314.633,
      "text": "即使他们并不想故意打断",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 10,
      "language_run": 2,
      "source": "whisper"
    },
    {
      "start": 314.633,
      "end": 317.273,
      "text": "当有人在会议通话中提问时",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 10,
      "language_run": 2,
      "source": "whisper"
    },
    {
      "start": 317.273,
      "end": 319.798,
      "text": "往往会有两三个人同时回答",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 10,
      "language_run": 2,
      "source": "whisper"
    },
    {
      "start": 321.786,
      "end": 324.836,
      "text": "旧版说话人分离系统面临的挑战在于",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 11,
      "language_run": 2,
      "source": "whisper"
    },
    {
      "start": 324.836,
      "end": 326.756,
      "text": "他们会因此变得混乱",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 11,
      "language_run": 2,
      "source": "whisper"
    },
    {
      "start": 326.756,
      "end": 329.576,
      "text": "他们不擅长区分不同的声音",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 11,
      "language_run": 2,
      "source": "whisper"
    },
    {
      "start": 329.576,
      "end": 331.793,
      "text": "尤其是在声音重叠时",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 11,
      "language_run": 2,
      "source": "whisper"
    },
    {
      "start": 333.547,
      "end": 334.097,
      "text": "我想说",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 12,
      "language_run": 2,
      "source": "whisper"
    },
    {
      "start": 334.097,
      "end": 337.817,
      "text": "我认为没有任何模型能在这件事上做到完美",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 12,
      "language_run": 2,
      "source": "whisper"
    },
    {
      "start": 337.817,
      "end": 338.797,
      "text": "即使是人类",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 12,
      "language_run": 2,
      "source": "whisper"
    },
    {
      "start": 338.797,
      "end": 341.457,
      "text": "这也可能是一项艰难的任务",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 12,
      "language_run": 2,
      "source": "whisper"
    },
    {
      "start": 341.457,
      "end": 343.137,
      "text": "但这些模型的评分方式",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 12,
      "language_run": 2,
      "source": "whisper"
    },
    {
      "start": 343.137,
      "end": 346.037,
      "text": "是使用所谓的说话人分离错误率",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 12,
      "language_run": 2,
      "source": "whisper"
    },
    {
      "start": 346.037,
      "end": 347.417,
      "text": "即D.E.R",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 12,
      "language_run": 2,
      "source": "whisper"
    },
    {
      "start": 347.417,
      "end": 349.997,
      "text": "它基本上是将三种错误相加",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 12,
      "language_run": 2,
      "source": "whisper"
    },
    {
      "start": 349.997,
      "end": 351.997,
      "text": "模型遗漏的语音",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 12,
      "language_run": 2,
      "source": "whisper"
    },
    {
      "start": 351.997,
      "end": 354.457,
      "text": "模型误判为非语音的内容",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 12,
      "language_run": 2,
      "source": "whisper"
    },
    {
      "start": 354.457,
      "end": 358.177,
      "text": "以及将语音错误归属于他人的次数",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 12,
      "language_run": 2,
      "source": "whisper"
    },
    {
      "start": 358.177,
      "end": 360.757,
      "text": "显然这里的分数越低越好",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 12,
      "language_run": 2,
      "source": "whisper"
    },
    {
      "start": 360.757,
      "end": 368.977,
      "text": "你可以看到NVIDIA的最新模型不仅击败了竞争对手,还大幅超越了他们自己的前代模型。",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 12,
      "language_run": 2,
      "source": "whisper"
    },
    {
      "start": 369.757,
      "end": 373.837,
      "text": "这里需要理解的另一个概念是流逝与离线的区别。",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 12,
      "language_run": 2,
      "source": "whisper"
    },
    {
      "start": 374.317,
      "end": 379.577,
      "text": "在离线模式下,模型可以查看整个录音,包括后面的内容。",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 12,
      "language_run": 2,
      "source": "whisper"
    },
    {
      "start": 379.837,
      "end": 383.837,
      "text": "而在流逝模式下,它必须在音频到达时做出决定。",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 12,
      "language_run": 2,
      "source": "whisper"
    },
    {
      "start": 384.057,
      "end": 388.997,
      "text": "所以,对于已经录制好的播客等,你总是希望使用离线模式。",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 12,
      "language_run": 2,
      "source": "whisper"
    },
    {
      "start": 388.997,
      "end": 392.357,
      "text": "对于任何实时字幕或语音智能体",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 12,
      "language_run": 2,
      "source": "whisper"
    },
    {
      "start": 392.357,
      "end": 394.257,
      "text": "你必须使用流式模式",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 12,
      "language_run": 2,
      "source": "whisper"
    },
    {
      "start": 394.257,
      "end": 394.817,
      "text": "好的",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 12,
      "language_run": 2,
      "source": "whisper"
    },
    {
      "start": 394.817,
      "end": 398.277,
      "text": "那么Nemotron 3说话人分离到底有什么新特点",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 12,
      "language_run": 2,
      "source": "whisper"
    },
    {
      "start": 398.277,
      "end": 400.217,
      "text": "它大约有1亿参数量",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 12,
      "language_run": 2,
      "source": "whisper"
    },
    {
      "start": 400.217,
      "end": 401.897,
      "text": "不包括ASR部分",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 12,
      "language_run": 2,
      "source": "whisper"
    },
    {
      "start": 401.897,
      "end": 404.677,
      "text": "运行它的最低GPU要求约为4GB",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 12,
      "language_run": 2,
      "source": "whisper"
    },
    {
      "start": 404.677,
      "end": 408.357,
      "text": "所以它实际上可以在许多小型GPU上运行",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 12,
      "language_run": 2,
      "source": "whisper"
    },
    {
      "start": 408.357,
      "end": 411.637,
      "text": "而且应该在笔记本电脑的RTX显卡上",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 12,
      "language_run": 2,
      "source": "whisper"
    },
    {
      "start": 411.637,
      "end": 412.997,
      "text": "也能跑得相当好",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 12,
      "language_run": 2,
      "source": "whisper"
    },
    {
      "start": 412.997,
      "end": 416.097,
      "text": "我之前提到过它支持多达8位说话人",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 12,
      "language_run": 2,
      "source": "whisper"
    },
    {
      "start": 416.097,
      "end": 418.297,
      "text": "另一点是它是语言无关的",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 12,
      "language_run": 2,
      "source": "whisper"
    },
    {
      "start": 418.297,
      "end": 420.757,
      "text": "并且在流逝和离线模式下都能工作",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 12,
      "language_run": 2,
      "source": "whisper"
    },
    {
      "start": 420.757,
      "end": 424.417,
      "text": "所以我认为最好的办法是直接进入演示",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 12,
      "language_run": 2,
      "source": "whisper"
    },
    {
      "start": 424.417,
      "end": 426.817,
      "text": "看看它实际上是如何运作的",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 12,
      "language_run": 2,
      "source": "whisper"
    },
    {
      "start": 426.817,
      "end": 427.657,
      "text": "好的",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 12,
      "language_run": 2,
      "source": "whisper"
    },
    {
      "start": 427.657,
      "end": 429.017,
      "text": "进入演示",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 12,
      "language_run": 2,
      "source": "whisper"
    },
    {
      "start": 429.017,
      "end": 431.057,
      "text": "让我先展示一下设置方式",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 12,
      "language_run": 2,
      "source": "whisper"
    },
    {
      "start": 431.057,
      "end": 434.577,
      "text": "首先感谢NVIDIA赞助这里的计算资源",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 12,
      "language_run": 2,
      "source": "whisper"
    },
    {
      "start": 434.577,
      "end": 439.577,
      "text": "我使用DGX Spark来实际运行这里的三个语音模型",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 12,
      "language_run": 2,
      "source": "whisper"
    },
    {
      "start": 439.577,
      "end": 442.497,
      "text": "让我带大家过一遍设置架构图",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 12,
      "language_run": 2,
      "source": "whisper"
    },
    {
      "start": 443.69,
      "end": 448.78,
      "text": "你在这里看到的是Next.js应用,它在我的Mac上本地运行。",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 13,
      "language_run": 2,
      "source": "whisper"
    },
    {
      "start": 449.68,
      "end": 453.365,
      "text": "然后通过Talescale调用后端到DGX。",
      "language": "zh",
      "chunk": 0,
      "local_language_run": 13,
      "language_run": 2,
      "source": "whisper"
    },
    {
      "start": 453.365,
      "end": 453.975,
      "text": "X Spark.",
      "language": "en",
      "chunk": 0,
      "local_language_run": 14,
      "language_run": 3,
      "source": "whisper"
    },
    {
      "start": 455.929,
      "end": 461.019,
      "text": "你可以看到,我在上面有一个包含NVIDIA NEMO的Docker容器。",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 2,
      "language_run": 4,
      "source": "whisper"
    },
    {
      "start": 461.659,
      "end": 466.119,
      "text": "VDI NEMO既提供我们正在讨论的说话人分离模型,",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 2,
      "language_run": 4,
      "source": "whisper"
    },
    {
      "start": 466.299,
      "end": 472.919,
      "text": "也提供ParaKit ASR模型和NEMOTRON 3.5多语言 ASR模型。",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 2,
      "language_run": 4,
      "source": "whisper"
    },
    {
      "start": 473.599,
      "end": 477.359,
      "text": "为了使用这些,我设置了一个Fast API应用。",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 2,
      "language_run": 4,
      "source": "whisper"
    },
    {
      "start": 477.519,
      "end": 479.919,
      "text": "这是一种实现此类功能的简单方法。",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 2,
      "language_run": 4,
      "source": "whisper"
    },
    {
      "start": 479.919,
      "end": 481.899,
      "text": "如果我要大规模部署",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 2,
      "language_run": 4,
      "source": "whisper"
    },
    {
      "start": 481.899,
      "end": 485.659,
      "text": "我可能会选择Dynamo或Triton来提供服务",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 2,
      "language_run": 4,
      "source": "whisper"
    },
    {
      "start": 485.659,
      "end": 487.439,
      "text": "但对于我想要的功能",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 2,
      "language_run": 4,
      "source": "whisper"
    },
    {
      "start": 487.439,
      "end": 489.959,
      "text": "目前的设置运行得非常完美",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 2,
      "language_run": 4,
      "source": "whisper"
    },
    {
      "start": 489.959,
      "end": 491.759,
      "text": "如果我们进入这个应用",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 2,
      "language_run": 4,
      "source": "whisper"
    },
    {
      "start": 491.759,
      "end": 495.099,
      "text": "你可以看到我基本上可以直接使用它们的演示",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 2,
      "language_run": 4,
      "source": "whisper"
    },
    {
      "start": 495.099,
      "end": 498.619,
      "text": "我可以选择想要使用哪个AS2模型",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 2,
      "language_run": 4,
      "source": "whisper"
    },
    {
      "start": 498.619,
      "end": 500.279,
      "text": "我打算用ParaKit",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 2,
      "language_run": 4,
      "source": "whisper"
    },
    {
      "start": 500.279,
      "end": 503.839,
      "text": "因为它在处理多说话人方面通常表现更好",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 2,
      "language_run": 4,
      "source": "whisper"
    },
    {
      "start": 503.839,
      "end": 505.019,
      "text": "我可以开始转录",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 2,
      "language_run": 4,
      "source": "whisper"
    },
    {
      "start": 505.019,
      "end": 508.439,
      "text": "你可以看到字面上只需要三四秒钟",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 2,
      "language_run": 4,
      "source": "whisper"
    },
    {
      "start": 508.439,
      "end": 512.579,
      "text": "他就处理并转录了这里一分三十多分钟的音频",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 2,
      "language_run": 4,
      "source": "whisper"
    },
    {
      "start": 516.257,
      "end": 519.447,
      "text": "由于我之前在摆弄这个并设置了一些名字",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 3,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 519.447,
      "end": 521.939,
      "text": "他实际上记住了这里这个人的名字",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 3,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 523.694,
      "end": 524.784,
      "text": "让我们听一下",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 4,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 524.784,
      "end": 528.544,
      "text": "看看当多个说话人同时说话时会发生什么",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 4,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 529.775,
      "end": 532.525,
      "text": "为了庆祝Nemotron三说话人分离的发布",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 5,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 532.525,
      "end": 535.885,
      "text": "我们邀请了另外七位开放文本转语音声音加入派对",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 5,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 535.885,
      "end": 538.465,
      "text": "我是Chris,NVDI的Nemotron声音",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 5,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 538.465,
      "end": 540.205,
      "text": "大家好,谢谢邀请",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 5,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 540.205,
      "end": 543.325,
      "text": "我是Jane,来自QTI的Pocket声音",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 5,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 543.325,
      "end": 545.905,
      "text": "很高兴见到你们,Chris和Jane",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 5,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 545.905,
      "end": 548.025,
      "text": "我是Serena,Queen三的声音",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 5,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 548.025,
      "end": 549.625,
      "text": "我是来自Cocoro的Adam",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 5,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 549.625,
      "end": 551.385,
      "text": "只有8200万参数量",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 5,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 551.385,
      "end": 552.805,
      "text": "我还是入选了嘉宾名单",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 5,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 552.805,
      "end": 553.685,
      "text": "你好,你好",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 5,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 553.685,
      "end": 555.825,
      "text": "我是来自Resemble AI的Chatterbox",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 5,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 555.825,
      "end": 557.825,
      "text": "我带来了富有表现力的能量",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 5,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 557.825,
      "end": 559.405,
      "text": "嗨,我是来自Nari Labs",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 5,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 559.405,
      "end": 560.445,
      "text": "的John Dyer声音",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 5,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 560.445,
      "end": 561.485,
      "text": "我们这里已经聚集了",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 5,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 561.485,
      "end": 562.185,
      "text": "很棒的人群",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 5,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 562.185,
      "end": 562.805,
      "text": "大家好",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 5,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 562.805,
      "end": 563.605,
      "text": "最近怎么样",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 5,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 563.605,
      "end": 565.505,
      "text": "我是来自Maya Research的",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 5,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 565.505,
      "end": 566.185,
      "text": "Maya E声音",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 5,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 566.185,
      "end": 566.965,
      "text": "好了",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 5,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 566.965,
      "end": 568.325,
      "text": "现在我们有八位合成",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 5,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 568.325,
      "end": 568.965,
      "text": "开放声音",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 5,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 568.965,
      "end": 569.865,
      "text": "做满了场子",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 5,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 569.865,
      "end": 570.785,
      "text": "让我们来测试一下",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 5,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 570.785,
      "end": 572.045,
      "text": "Nemotron三说话人分离",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 5,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 572.045,
      "end": 572.265,
      "text": "好的",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 5,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 572.265,
      "end": 573.865,
      "text": "新的说话人分离模型",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 5,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 573.865,
      "end": 574.885,
      "text": "可以在转路时",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 5,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 574.885,
      "end": 575.925,
      "text": "标记我们的每个声音",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 5,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 575.925,
      "end": 577.425,
      "text": "但它能在我们所有人",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 5,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 577.425,
      "end": 578.065,
      "text": "一起说话时",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 5,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 578.065,
      "end": 578.765,
      "text": "做到这一点吗",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 5,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 578.765,
      "end": 579.525,
      "text": "让我们来看看",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 5,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 579.525,
      "end": 580.925,
      "text": "即使我们互相打断",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 5,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 580.925,
      "end": 581.905,
      "text": "它也能叫出",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 5,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 581.905,
      "end": 582.805,
      "text": "我们俩的名字",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 5,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 582.805,
      "end": 583.785,
      "text": "竹针标记",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 5,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 583.785,
      "end": 585.525,
      "text": "谁在何时说话",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 5,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 585.525,
      "end": 586.485,
      "text": "好的",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 5,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 586.485,
      "end": 588.725,
      "text": "正如你在这里看到的",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 5,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 588.725,
      "end": 591.925,
      "text": "它在转录每一个声音方面做得很好",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 5,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 591.925,
      "end": 593.365,
      "text": "如果我们往下看看",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 5,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 593.365,
      "end": 595.545,
      "text": "可以看到每个声音都被转录了",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 5,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 595.545,
      "end": 598.325,
      "text": "我想我的UI在这方面稍微有点滞后",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 5,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 598.325,
      "end": 601.645,
      "text": "但你完全可以感受到它是如何组装起来的",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 5,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 601.645,
      "end": 604.765,
      "text": "实际上如果我们放大一点来自QTI的声音",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 5,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 604.765,
      "end": 605.925,
      "text": "很高兴见到你们",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 5,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 605.925,
      "end": 607.185,
      "text": "Chris和Jen",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 5,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 607.185,
      "end": 607.965,
      "text": "我是Serena",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 5,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 607.965,
      "end": 609.325,
      "text": "Queen3的声音",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 5,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 609.325,
      "end": 610.765,
      "text": "你可以很快看到",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 5,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 610.765,
      "end": 615.525,
      "text": "我们基本上可以组装出一个能转录这些声音的东西",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 5,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 615.525,
      "end": 617.765,
      "text": "如果我们想在这里输入它们的名字",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 5,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 617.765,
      "end": 620.805,
      "text": "它会便利并为我们重命名所有人",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 5,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 620.805,
      "end": 625.485,
      "text": "现在,如果我们想看看实际模型返回的是什么",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 5,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 625.485,
      "end": 628.945,
      "text": "你可以看到这里是模型返回的原始JSON",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 5,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 628.945,
      "end": 632.645,
      "text": "我们可以看到这里实际使用的是什么",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 5,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 632.645,
      "end": 635.465,
      "text": "我们可以看到不同的说话人被标记等",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 5,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 635.465,
      "end": 641.845,
      "text": "然后这个软件基本上将Speaker三重写为ADAM等便利这个过程",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 5,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 641.845,
      "end": 644.945,
      "text": "做这样的事情很酷的一点是",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 5,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 644.945,
      "end": 647.045,
      "text": "我们可以很快获得统计数据",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 5,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 647.045,
      "end": 650.056,
      "text": "比如每个人在那里说了多久",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 5,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 651.176,
      "end": 652.346,
      "text": "说了多少次",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 6,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 652.346,
      "end": 653.706,
      "text": "有多少次发言",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 6,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 653.706,
      "end": 657.146,
      "text": "如果我们想在这里放入更长的内容",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 6,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 657.146,
      "end": 658.326,
      "text": "这会非常有用",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 6,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 658.326,
      "end": 658.806,
      "text": "好的",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 6,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 658.806,
      "end": 661.866,
      "text": "这里酷的是我们可以处理长文件",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 6,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 661.866,
      "end": 663.926,
      "text": "所以我基本上放入了",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 6,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 663.926,
      "end": 666.206,
      "text": "All In 播客的一集完整节目",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 6,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 666.206,
      "end": 667.466,
      "text": "他们采访了黄仁勋",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 6,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 667.466,
      "end": 669.786,
      "text": "你可以看到他基本上已经处理完了",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 6,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 671.01,
      "end": 672.1,
      "text": "我们得到了输出",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 7,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 672.1,
      "end": 674.44,
      "text": "我命名了这里的实际说话人",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 7,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 674.44,
      "end": 675.78,
      "text": "所以我们可以看到",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 7,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 675.78,
      "end": 677.56,
      "text": "说话人依次出现",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 7,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 677.56,
      "end": 679.1,
      "text": "现在我们有了每个人",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 7,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 679.1,
      "end": 681.08,
      "text": "说了什么的完整转录文本",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 7,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 681.08,
      "end": 683.92,
      "text": "这里酷的是我们可以查看统计数据",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 7,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 683.92,
      "end": 686.3,
      "text": "我们可以看到他们各自的说话时长",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 7,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 687.651,
      "end": 701.781,
      "text": "我们可以看到,果然,显然,黄仁勋是这里的嘉宾,所以他是说话最多的,而这一切都在148秒内完成,所以对于这个超过一小时的长文件,这大约花了两分半钟。",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 8,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 701.781,
      "end": 704.401,
      "text": "这让我们基本上可以便利",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 8,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 704.401,
      "end": 705.621,
      "text": "你可以看到",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 8,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 705.621,
      "end": 706.761,
      "text": "现在我们基本上",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 8,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 706.761,
      "end": 708.601,
      "text": "可以直接阅读转录文本",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 8,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 708.601,
      "end": 711.041,
      "text": "我可以将其导出为文本文件",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 8,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 711.041,
      "end": 712.921,
      "text": "或者在这里导出为srt",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 8,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 712.921,
      "end": 714.141,
      "text": "这里酷的是",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 8,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 714.141,
      "end": 716.601,
      "text": "如果我确实想做多语言",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 8,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 716.601,
      "end": 718.061,
      "text": "或其他类似的事情",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 8,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 718.061,
      "end": 719.681,
      "text": "我可以切换模型",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 8,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 719.681,
      "end": 721.021,
      "text": "使用这个模型",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 8,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 721.021,
      "end": 723.341,
      "text": "而不是这里的paraket模型",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 8,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 723.341,
      "end": 724.341,
      "text": "你可以看到",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 8,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 724.341,
      "end": 725.521,
      "text": "用这个模型",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 8,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 725.521,
      "end": 727.781,
      "text": "我们支持很多不同的语言",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 8,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 727.781,
      "end": 729.181,
      "text": "如果你记得我做过",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 8,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 729.181,
      "end": 730.101,
      "text": "关于这个的视频",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 8,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 730.101,
      "end": 732.441,
      "text": "你实际上可以微调这个模型",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 8,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 732.441,
      "end": 735.541,
      "text": "使其在其他语言上表现非常好",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 8,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 735.541,
      "end": 737.741,
      "text": "他们在这里称之为适配就绪",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 8,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 737.741,
      "end": 740.701,
      "text": "现在我还没有真正设置这个应用去做",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 8,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 740.701,
      "end": 742.561,
      "text": "但模型能做的另一件事",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 8,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 742.561,
      "end": 744.601,
      "text": "是你也可以进行流逝响应",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 8,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 744.601,
      "end": 744.961,
      "text": "当然",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 8,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 744.961,
      "end": 747.541,
      "text": "那样你的准确率可能会下降一点",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 8,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 747.541,
      "end": 748.881,
      "text": "但你在那里",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 8,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 748.881,
      "end": 751.281,
      "text": "仍然应该能获得相当好的结果",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 8,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 751.281,
      "end": 752.241,
      "text": "所以",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 8,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 752.241,
      "end": 754.741,
      "text": "如果你有兴趣做任何语音转录",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 8,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 754.741,
      "end": 757.521,
      "text": "并且希望其中包含说话人分离等",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 8,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 757.521,
      "end": 759.681,
      "text": "这是一个很好的设置方法",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 8,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 759.681,
      "end": 762.141,
      "text": "你可以批量完成所有转路等",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 8,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 762.141,
      "end": 763.041,
      "text": "最重要的是",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 8,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 763.041,
      "end": 765.301,
      "text": "能够完全在本地完成所有这一切",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 8,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 765.301,
      "end": 766.121,
      "text": "所以",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 8,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 766.121,
      "end": 767.981,
      "text": "如果你对这方面有任何问题",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 8,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 767.981,
      "end": 769.361,
      "text": "请在评论区告诉我",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 8,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 769.361,
      "end": 770.941,
      "text": "我认为看到NVIDIA",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 8,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 770.941,
      "end": 773.001,
      "text": "接下来在这些语音模型上",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 8,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 773.001,
      "end": 774.541,
      "text": "会做什么将会非常酷",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 8,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 774.541,
      "end": 775.881,
      "text": "就像我说的",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 8,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 775.881,
      "end": 777.081,
      "text": "从我看到的来看",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 8,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 777.081,
      "end": 778.001,
      "text": "它们不断迭代",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 8,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 778.001,
      "end": 779.141,
      "text": "以改进这些模型",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 8,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 779.141,
      "end": 780.701,
      "text": "并为不同的用力",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 8,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 780.701,
      "end": 782.821,
      "text": "提供不同类型的模型",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 8,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 782.821,
      "end": 783.461,
      "text": "总之",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 8,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 783.461,
      "end": 784.521,
      "text": "一如既往",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 8,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 784.521,
      "end": 785.981,
      "text": "如果你觉得视频有用",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 8,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 785.981,
      "end": 787.341,
      "text": "请点赞并订阅",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 8,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 787.341,
      "end": 788.521,
      "text": "我们下个视频见",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 8,
      "language_run": 5,
      "source": "whisper"
    },
    {
      "start": 788.521,
      "end": 789.181,
      "text": "再见",
      "language": "zh",
      "chunk": 1,
      "local_language_run": 8,
      "language_run": 5,
      "source": "whisper"
    }
  ],
  "transcript_selection": {
    "schema": "videonote.transcript_selection.v1",
    "status": "SELECTED",
    "requested_language": "zh",
    "duration": 789.24625,
    "reference_path": null,
    "reference_kind": null,
    "reference_language": null,
    "selected_source": "whisper_segmented_multilingual",
    "decision_reason": "multilingual_segmented_whisper_passed",
    "whisper_rejected": false,
    "source_profile": {
      "segment_count": 0,
      "duration_coverage": 0.0,
      "usable_full": false,
      "usable_partial": false,
      "hallucination_region_count": 0,
      "hallucination_regions": [],
      "language": {
        "requested": "zh",
        "matched": false
      }
    },
    "whisper_profile": {
      "segment_count": 291,
      "multilingual": true,
      "first_start": 0.0,
      "final_end": 789.181,
      "duration_coverage": 0.999917,
      "maximum_gap_seconds": 3.678,
      "inverted_count": 0,
      "language": {
        "requested": "zh",
        "matched": true,
        "latin": 496,
        "han": 3408,
        "kana": 0,
        "hangul": 0
      },
      "hallucination_region_count": 0,
      "hallucination_regions": [],
      "usable_full": true,
      "usable_partial": true
    },
    "comparison": {
      "window_seconds": 30.0,
      "compared_window_count": 0,
      "median_score": 0.0,
      "divergent_window_count": 0,
      "divergent_ratio": 0.0,
      "divergent_windows": []
    },
    "timestamp_alignment": {
      "applied": false,
      "estimated_offset_seconds": 0.0,
      "matched_segment_count": 0,
      "match_scores": []
    },
    "caption_timing_repair": null,
    "filled_source_profile": null,
    "whisper_duration_guard": {
      "media_duration": 789.246,
      "rejected_segment_count": 0,
      "rejected_segments": []
    }
  }
}