{
  "text": "这个本地AI工具有点离谱\n它能实时语音对话\n哈喽 你能帮我做什么\n你好 我可以帮你回答问题\n提供信息 安排日程\n还能声音克隆\n大家好呀 要站在光里\n不要光站在那里\n不仅能声称AI音乐\n还能换词翻唱\n这一整套服务\n全都跑在这台8G显存的笔记本上\n不用云端语音平台\n不需要联网语音服务\n这个工具叫Audio CPP\n你可以把它理解成音乐领域的Nama CPP\n它想解决的不是某一个语音功能\n而是把本地音频模型\n统一到同一套云形方式里\n以前我们想在本地部署语音模型\n经常是一个模型一套环境\n我自己这台机器上\n就装过声音克龙\n谈文本合成\nAI翻唱\n实施语音好几套环境\n每一套都有不同的配声版本\n不同的库大依赖\n最后不是模型跑不动\n是环境先把人劝退了\nAudio CPP解决的就是这个问题\n它把TTS,ASR,音乐生存,声音转换\n都接近同一个后台\n通过界面可以随时切换模型\n就像你用Olama在本地跑大语音模型一样\nAudio CPP能让你更简单的在本地跑各种语音模型\n我这次演示的大部分核心功能\n8G显存就能跑起来\n有些小模型甚至不需要显卡\n用CPU也能运行\n它现在已经接入的模型家族\n有21个可以直接下载使用\n还有一批在测试当中\n统一底座带来的另一个直接好处是\n运行效率更高\n这是官方在RTX5090上的速度对比\n大部分模型在AudioCPP下运行\n比他们用自己的推理框架快了1-3倍\n有的甚至快了8-10倍\n我也在本地用RTX4060做了测试\n拿微软开源的Vive Voice对比\n用官方自己的程序\n合成1万字的小说朗读\n耗时110分钟\n而同样的文本\n拿到AudioCPP下合成\n仅用了29分钟\n只有原来的四分之一\n除了速度\n这期还有我最想让你看的\n它把一堆分散的语音能力\n变成一套本地系统\n前面那个实时语音系统\n等会我会完整演示\n后面还有一个更好玩的音乐生成能力\n歌曲旋律不变\n直接换词翻唱\n最后过一遍声音转换\n设计等扩展功能\n以及整合包的下载和使用\n先看TTS\n也就是文本转语音\n这里列的模型\n整合包里自带了Pocket TTS和Chemun3 TTS 0.6B\n它们都支持参考音频\n也就意味着可以做声音克隆\n基本用法非常简单\n选择模型\n点击加载\n加载完会有提示\n然后选一段参考音频\n再输入合成文本\n点击生成\n这种一句话的声音克隆\n基本上是秒生成\nOpenAI launched GPT Live\nA full duplex voice system for chat GPT\n不过要注意\nPocket TTS不支持中文\n要生成中文语音\n可以换成千问3 TTS 0.6B\n它的参数虽然不大\n但合成效果相当不错\n全双功语音系统\n支持实时打断联网搜索和视觉理解\n这两款模型显存占用都很小\n非常适合低配显卡\n以及没有独显的机器\n刚上手\n合成设置里的参数都可以不用管\n保持默认\n先把模型跑起来\n听一下技术效果\n熟悉之后再去调温度\n采样这些参数\n列表里还有两个模型比较有特点\n一个是Vox CPM2\n它好玩的地方是能生成方言\n给一段方言参考音频\n我们每个人都在做小丑\n就可以合成相应的方言语音\n错了你又不认\n认了你又不改\n它支持国内大部分方言\n还能克隆语气\n想深入了解这个模型的能力\n可以翻译一下我之前专门做的Vox CPM那期视频\n另一个有特点的是Voic Voice\n就是开头速度对比的那个模型\n那段近一万字的小说\n合成期间显存占用没超过7G\n生存出来的音频我全程听了一遍\n没有明显的漏读和错读\n忽然山腰里传来群马奔驰之声\n提声越来越响\n不久四面黄布大旗从山崖边升起\n这里有个注意点\n不管你用哪款模型\n参考音频都不要太长\n建议控制在10秒以内\n否则会明显拖慢合成速度\n另外还有个小技巧\n如果有些参考音频你经常用\n可以把它们放到WebUI下的Voice目标里\n然后把文件名和对应的文本写到这个文件里\n刷新列表就可以直接选了\n临时上传的参考音频\n下面对应的文本也不用手动输入\n可以先到ASR标签里\n把音频内容转写出来\n这就引出了AudioCPP第二个核心能力\nASR语音转文字\n我个人常用的模型是千问3ASR\n等会要演示的实时语音\n我的声音转文本用的就是它\n中英文都支持\n速度也很快\n这段音频两分多钟\n转录只用了21秒\n如果转录的是多人对话\n要勾选对话模式\n转出来的就是这种带说话人序列的文本\n不过对话模式有长度限制\n目前更适合较短的对话片段\n单人语音转写没有长度限制\n歌词也能识别\n等下做AI翻唱要拿原曲歌词\n就可以先到这里做一遍初步识别\n再手动校对\n有了TTS和ASR\n现在我们可以把它们拼起来用\n这是我认为Audio CPP最有价值的地方\n本地实时语音系统\n我正在测试一个实时语音系统\n好的\n我正在实时语音系统里和你对话\n如果你想测试它的稳定性\n延迟或其他功能\n请随时提问或下达指令\n语音识别和合成都是跑在本地的\n明白了\n本地运行意味着响应速度更快\n同时数据也更安全\n这是一个很注重隐私和效率的设计思路\n大语言模型是云端\n你觉得这个架构是否合理\n这个架构是合理的\n从我说完话\n到它开始回答\n中间大概一秒\n如果是一问对话\n这个颜色会更低\nWhat can you do for me\nI can answer questions\nSummarize information\n整个流程就是\n麦克风输入\nASR转文本\n大语言模型声称回答\nTTS再把它朗读出来\n这就是一个\n本地语音交互系统的\n基本形态\n设置里\n大语言模型这一步是活的\n本地化的是语音这一层\n你可以接本地\n欧拉玛跑到语言模型\n也可以接自己订阅的API\nAudio CPP负责的是音频这一层\n听和说\n所以我说它不是一个声音克隆玩具\n而是消费级显卡上的\n本地AI语音底座\n而且它用的是OpenAI兼容接口\n你本地的其他应用\n只要支持这种接口形式\n就能直接调用服务\n就像设置里这样写\nTTS的地址\n模型名称\nASR的地址\n模型名称\n朗读的声音也可以自定义\n比如我在这个OpenMagic AI应用里\n直接调用AudioCPP的TTS服务\n填写UIL\n还有模型名称\n由于这个应用没法填参考音频\n所以我后台单独起一个AudioCPP服务\n启动时带上参考音频\n回到OpenMagic测试\n你好 这是一段测试语音\n能正确返回语音\n这样就接好了\n这就是一个完全免费的本地TTS服务\n如果你在做本地Agent\n或者想给自己的应用\n加上语音输入输出\nAudioCPP提供的接口就非常有用\n接下来我们看音乐生成\n这部分也是我觉得AudioCPP很强的地方\n它不只是把TTS和ASR放到一起\n连音乐生成换词翻唱\n这些原本比较重的模型\n也开始接近同一个底座里\n这个ACE STEP我之前介绍过\n当时我在本地8G显存上跑不动翻唱\n最后是在Club上演示的\n现在用AudioCPP\n本地就能跑起来\n翻唱可以按下面步骤操作\n先上传要翻唱的歌曲\n然后点分析\n这个过程要等几十秒到几分钟\n具体看歌曲长度\n分析完成后\n合成翻唱的时长改成-1\n点开高级参数\n操作类型选Remix\n歌曲风格\n曲谱信息\n分析完以后就会自动填好\n不用动\n你只需要填上圆曲歌词\n提示词直接复制上面的圆曲描述\n最后填上你的翻唱歌词\n可以看一下显存占用\n虽然看上去超过8G\n但不会爆显存\n最后能顺利跑完\n换词翻唱的随机性比较大\n需要多试几次\n如果音质不好\n可以适当增加生存部署\n要是新词唱不准或者唱不出来\n可以调这个Flowedit参数\n从0.7到0.9去尝试\n如果你只想生成背景音乐\n我推荐用Stable Audio\n比ACE Stable更稳\n声音转换相当于音色迁移\n保持说话内容不变\n语气不变\n只把音色换掉\n马匪\n任何时候都要搅\n不搅不行\n马匪\n任何时候都要搅\n不搅不行\n歌声转换也是类似逻辑\n但我实测\n如果你追求追踪质量\n我更推荐之前那套RVC流程\n做歌声转换之前\n最好先做人声分离\n音频分离标签下\n就有相应的模型\n再往下是音频分析这一组\n它们都是工具类模型\n比如这个是实时语音检测\n下面这个是多人对话识别\n最后一个是声音设计\n这个挺有意思\n它其实类似声音克隆\n只是不需要参考音频\n你直接用文字描述一个声音\n比如磁性的中年男生\n语速偏慢\n波音枪\n模型就会按描述生成\n这个功能做临时配音比较好用\n最后说一下整合包\nAudio CPP本身是个C++项目\n官方目前主要靠命令行来使用\n对有经验的朋友来说问题不大\n但对没技术基础的\n这一步确实比较劝退\n所以我做了带界面的整合包\n解压就能用\n支持16-50系列的N卡\n目前A卡以及没有独显的机器\n会自动走CPU模式\n速度会慢一些\n但跑几个轻量模型没问题\n整合包里\nRun Web UI是启动网页界面\n想用哪个模型\n界面里就能直接下载\n想体验实时语音\n先在Web UI里加载模型\n再启动这个ASR服务\n最后运行Run Real Time\n配置里把接入的大模型改一下\n如果你也用Deep Seek\nAPI Key换成自己的就可以了\n这一整套原代码我已经开完了\n这是GitHub地址\n使用中如果遇到问题\n或者希望增加功能\n可以在评论区告诉我\n我会收集整理定期更新版本\n总结一下\nAudio CPP最大的意义\n不是又多了一个声音客户工具\n而是本地语音模型\n开始变成一套统一的系统\n大语言模型有Lama CPP\n图片视频生成有Conf UI\n语音模型现在也开始有了\n自己的本地运行中心\n对于普通用户来说\n使用门槛更低了\n对开发者来说\n本地语音服务接入更容易\n这也是我这期最想讲清楚的地方\n本期视频用到的所有链接\n我都放在评论区置顶\n这里是AI探索已发现\n感谢观看\n下期见",
  "language": "zh",
  "duration": 686.016,
  "segments": [
    {
      "start": 0.0,
      "end": 1.78,
      "text": "这个本地AI工具有点离谱",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.13889033735291031,
      "no_speech_prob": 1.6448698306192888e-11,
      "compression_ratio": 1.2008928571428572
    },
    {
      "start": 1.78,
      "end": 3.4,
      "text": "它能实时语音对话",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.13889033735291031,
      "no_speech_prob": 1.6448698306192888e-11,
      "compression_ratio": 1.2008928571428572
    },
    {
      "start": 3.4,
      "end": 5.32,
      "text": "哈喽 你能帮我做什么",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.13889033735291031,
      "no_speech_prob": 1.6448698306192888e-11,
      "compression_ratio": 1.2008928571428572
    },
    {
      "start": 5.32,
      "end": 7.56,
      "text": "你好 我可以帮你回答问题",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.13889033735291031,
      "no_speech_prob": 1.6448698306192888e-11,
      "compression_ratio": 1.2008928571428572
    },
    {
      "start": 7.56,
      "end": 9.44,
      "text": "提供信息 安排日程",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.13889033735291031,
      "no_speech_prob": 1.6448698306192888e-11,
      "compression_ratio": 1.2008928571428572
    },
    {
      "start": 9.44,
      "end": 11.0,
      "text": "还能声音克隆",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.13889033735291031,
      "no_speech_prob": 1.6448698306192888e-11,
      "compression_ratio": 1.2008928571428572
    },
    {
      "start": 11.0,
      "end": 13.12,
      "text": "大家好呀 要站在光里",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.13889033735291031,
      "no_speech_prob": 1.6448698306192888e-11,
      "compression_ratio": 1.2008928571428572
    },
    {
      "start": 13.12,
      "end": 14.72,
      "text": "不要光站在那里",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.13889033735291031,
      "no_speech_prob": 1.6448698306192888e-11,
      "compression_ratio": 1.2008928571428572
    },
    {
      "start": 14.72,
      "end": 16.24,
      "text": "不仅能声称AI音乐",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.13889033735291031,
      "no_speech_prob": 1.6448698306192888e-11,
      "compression_ratio": 1.2008928571428572
    },
    {
      "start": 16.24,
      "end": 21.04,
      "text": "还能换词翻唱",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.13889033735291031,
      "no_speech_prob": 1.6448698306192888e-11,
      "compression_ratio": 1.2008928571428572
    },
    {
      "start": 21.04,
      "end": 28.14,
      "text": "这一整套服务",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.13889033735291031,
      "no_speech_prob": 1.6448698306192888e-11,
      "compression_ratio": 1.2008928571428572
    },
    {
      "start": 28.14,
      "end": 30.62,
      "text": "全都跑在这台8G显存的笔记本上",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10266796672513699,
      "no_speech_prob": 1.5881304951670394e-11,
      "compression_ratio": 1.3457142857142856
    },
    {
      "start": 30.62,
      "end": 32.42,
      "text": "不用云端语音平台",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10266796672513699,
      "no_speech_prob": 1.5881304951670394e-11,
      "compression_ratio": 1.3457142857142856
    },
    {
      "start": 32.42,
      "end": 33.92,
      "text": "不需要联网语音服务",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10266796672513699,
      "no_speech_prob": 1.5881304951670394e-11,
      "compression_ratio": 1.3457142857142856
    },
    {
      "start": 33.92,
      "end": 35.94,
      "text": "这个工具叫Audio CPP",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10266796672513699,
      "no_speech_prob": 1.5881304951670394e-11,
      "compression_ratio": 1.3457142857142856
    },
    {
      "start": 35.94,
      "end": 39.14,
      "text": "你可以把它理解成音乐领域的Nama CPP",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10266796672513699,
      "no_speech_prob": 1.5881304951670394e-11,
      "compression_ratio": 1.3457142857142856
    },
    {
      "start": 39.14,
      "end": 41.6,
      "text": "它想解决的不是某一个语音功能",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10266796672513699,
      "no_speech_prob": 1.5881304951670394e-11,
      "compression_ratio": 1.3457142857142856
    },
    {
      "start": 41.6,
      "end": 43.3,
      "text": "而是把本地音频模型",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10266796672513699,
      "no_speech_prob": 1.5881304951670394e-11,
      "compression_ratio": 1.3457142857142856
    },
    {
      "start": 43.3,
      "end": 45.4,
      "text": "统一到同一套云形方式里",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10266796672513699,
      "no_speech_prob": 1.5881304951670394e-11,
      "compression_ratio": 1.3457142857142856
    },
    {
      "start": 45.4,
      "end": 47.98,
      "text": "以前我们想在本地部署语音模型",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10266796672513699,
      "no_speech_prob": 1.5881304951670394e-11,
      "compression_ratio": 1.3457142857142856
    },
    {
      "start": 47.98,
      "end": 49.86,
      "text": "经常是一个模型一套环境",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10266796672513699,
      "no_speech_prob": 1.5881304951670394e-11,
      "compression_ratio": 1.3457142857142856
    },
    {
      "start": 49.86,
      "end": 51.44,
      "text": "我自己这台机器上",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10266796672513699,
      "no_speech_prob": 1.5881304951670394e-11,
      "compression_ratio": 1.3457142857142856
    },
    {
      "start": 51.44,
      "end": 52.74,
      "text": "就装过声音克龙",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10266796672513699,
      "no_speech_prob": 1.5881304951670394e-11,
      "compression_ratio": 1.3457142857142856
    },
    {
      "start": 52.74,
      "end": 53.66,
      "text": "谈文本合成",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10266796672513699,
      "no_speech_prob": 1.5881304951670394e-11,
      "compression_ratio": 1.3457142857142856
    },
    {
      "start": 53.66,
      "end": 54.48,
      "text": "AI翻唱",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10266796672513699,
      "no_speech_prob": 1.5881304951670394e-11,
      "compression_ratio": 1.3457142857142856
    },
    {
      "start": 54.48,
      "end": 56.02,
      "text": "实施语音好几套环境",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10266796672513699,
      "no_speech_prob": 1.5881304951670394e-11,
      "compression_ratio": 1.3457142857142856
    },
    {
      "start": 56.02,
      "end": 57.96,
      "text": "每一套都有不同的配声版本",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10266796672513699,
      "no_speech_prob": 1.5881304951670394e-11,
      "compression_ratio": 1.3457142857142856
    },
    {
      "start": 57.96,
      "end": 59.0,
      "text": "不同的库大依赖",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10243697919343647,
      "no_speech_prob": 1.6444850689523172e-11,
      "compression_ratio": 1.2564841498559078
    },
    {
      "start": 59.0,
      "end": 60.64,
      "text": "最后不是模型跑不动",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10243697919343647,
      "no_speech_prob": 1.6444850689523172e-11,
      "compression_ratio": 1.2564841498559078
    },
    {
      "start": 60.64,
      "end": 62.36,
      "text": "是环境先把人劝退了",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10243697919343647,
      "no_speech_prob": 1.6444850689523172e-11,
      "compression_ratio": 1.2564841498559078
    },
    {
      "start": 62.36,
      "end": 64.74,
      "text": "Audio CPP解决的就是这个问题",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10243697919343647,
      "no_speech_prob": 1.6444850689523172e-11,
      "compression_ratio": 1.2564841498559078
    },
    {
      "start": 64.74,
      "end": 68.16,
      "text": "它把TTS,ASR,音乐生存,声音转换",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10243697919343647,
      "no_speech_prob": 1.6444850689523172e-11,
      "compression_ratio": 1.2564841498559078
    },
    {
      "start": 68.16,
      "end": 69.7,
      "text": "都接近同一个后台",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10243697919343647,
      "no_speech_prob": 1.6444850689523172e-11,
      "compression_ratio": 1.2564841498559078
    },
    {
      "start": 69.7,
      "end": 72.06,
      "text": "通过界面可以随时切换模型",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10243697919343647,
      "no_speech_prob": 1.6444850689523172e-11,
      "compression_ratio": 1.2564841498559078
    },
    {
      "start": 72.06,
      "end": 75.46,
      "text": "就像你用Olama在本地跑大语音模型一样",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10243697919343647,
      "no_speech_prob": 1.6444850689523172e-11,
      "compression_ratio": 1.2564841498559078
    },
    {
      "start": 75.46,
      "end": 79.64,
      "text": "Audio CPP能让你更简单的在本地跑各种语音模型",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10243697919343647,
      "no_speech_prob": 1.6444850689523172e-11,
      "compression_ratio": 1.2564841498559078
    },
    {
      "start": 79.64,
      "end": 82.04,
      "text": "我这次演示的大部分核心功能",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10243697919343647,
      "no_speech_prob": 1.6444850689523172e-11,
      "compression_ratio": 1.2564841498559078
    },
    {
      "start": 82.04,
      "end": 83.48,
      "text": "8G显存就能跑起来",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10243697919343647,
      "no_speech_prob": 1.6444850689523172e-11,
      "compression_ratio": 1.2564841498559078
    },
    {
      "start": 83.48,
      "end": 85.66,
      "text": "有些小模型甚至不需要显卡",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10243697919343647,
      "no_speech_prob": 1.6444850689523172e-11,
      "compression_ratio": 1.2564841498559078
    },
    {
      "start": 85.66,
      "end": 87.08,
      "text": "用CPU也能运行",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10243697919343647,
      "no_speech_prob": 1.6444850689523172e-11,
      "compression_ratio": 1.2564841498559078
    },
    {
      "start": 87.08,
      "end": 89.28,
      "text": "它现在已经接入的模型家族",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10777023641820897,
      "no_speech_prob": 1.4015149284174466e-11,
      "compression_ratio": 1.2383720930232558
    },
    {
      "start": 89.28,
      "end": 91.44,
      "text": "有21个可以直接下载使用",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10777023641820897,
      "no_speech_prob": 1.4015149284174466e-11,
      "compression_ratio": 1.2383720930232558
    },
    {
      "start": 91.44,
      "end": 93.12,
      "text": "还有一批在测试当中",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10777023641820897,
      "no_speech_prob": 1.4015149284174466e-11,
      "compression_ratio": 1.2383720930232558
    },
    {
      "start": 93.12,
      "end": 96.08,
      "text": "统一底座带来的另一个直接好处是",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10777023641820897,
      "no_speech_prob": 1.4015149284174466e-11,
      "compression_ratio": 1.2383720930232558
    },
    {
      "start": 96.08,
      "end": 97.18,
      "text": "运行效率更高",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10777023641820897,
      "no_speech_prob": 1.4015149284174466e-11,
      "compression_ratio": 1.2383720930232558
    },
    {
      "start": 97.18,
      "end": 100.48,
      "text": "这是官方在RTX5090上的速度对比",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10777023641820897,
      "no_speech_prob": 1.4015149284174466e-11,
      "compression_ratio": 1.2383720930232558
    },
    {
      "start": 100.48,
      "end": 103.12,
      "text": "大部分模型在AudioCPP下运行",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10777023641820897,
      "no_speech_prob": 1.4015149284174466e-11,
      "compression_ratio": 1.2383720930232558
    },
    {
      "start": 103.12,
      "end": 105.98,
      "text": "比他们用自己的推理框架快了1-3倍",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10777023641820897,
      "no_speech_prob": 1.4015149284174466e-11,
      "compression_ratio": 1.2383720930232558
    },
    {
      "start": 105.98,
      "end": 107.78,
      "text": "有的甚至快了8-10倍",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10777023641820897,
      "no_speech_prob": 1.4015149284174466e-11,
      "compression_ratio": 1.2383720930232558
    },
    {
      "start": 107.78,
      "end": 111.14,
      "text": "我也在本地用RTX4060做了测试",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10777023641820897,
      "no_speech_prob": 1.4015149284174466e-11,
      "compression_ratio": 1.2383720930232558
    },
    {
      "start": 111.14,
      "end": 113.12,
      "text": "拿微软开源的Vive Voice对比",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10777023641820897,
      "no_speech_prob": 1.4015149284174466e-11,
      "compression_ratio": 1.2383720930232558
    },
    {
      "start": 113.12,
      "end": 114.64,
      "text": "用官方自己的程序",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10777023641820897,
      "no_speech_prob": 1.4015149284174466e-11,
      "compression_ratio": 1.2383720930232558
    },
    {
      "start": 114.64,
      "end": 116.44,
      "text": "合成1万字的小说朗读",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10777023641820897,
      "no_speech_prob": 1.4015149284174466e-11,
      "compression_ratio": 1.2383720930232558
    },
    {
      "start": 116.44,
      "end": 118.0,
      "text": "耗时110分钟",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.1118585869456022,
      "no_speech_prob": 1.8711828961293087e-11,
      "compression_ratio": 1.3008595988538683
    },
    {
      "start": 118.0,
      "end": 119.1,
      "text": "而同样的文本",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.1118585869456022,
      "no_speech_prob": 1.8711828961293087e-11,
      "compression_ratio": 1.3008595988538683
    },
    {
      "start": 119.1,
      "end": 120.86,
      "text": "拿到AudioCPP下合成",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.1118585869456022,
      "no_speech_prob": 1.8711828961293087e-11,
      "compression_ratio": 1.3008595988538683
    },
    {
      "start": 120.86,
      "end": 122.12,
      "text": "仅用了29分钟",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.1118585869456022,
      "no_speech_prob": 1.8711828961293087e-11,
      "compression_ratio": 1.3008595988538683
    },
    {
      "start": 122.12,
      "end": 123.72,
      "text": "只有原来的四分之一",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.1118585869456022,
      "no_speech_prob": 1.8711828961293087e-11,
      "compression_ratio": 1.3008595988538683
    },
    {
      "start": 123.72,
      "end": 124.84,
      "text": "除了速度",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.1118585869456022,
      "no_speech_prob": 1.8711828961293087e-11,
      "compression_ratio": 1.3008595988538683
    },
    {
      "start": 124.84,
      "end": 126.72,
      "text": "这期还有我最想让你看的",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.1118585869456022,
      "no_speech_prob": 1.8711828961293087e-11,
      "compression_ratio": 1.3008595988538683
    },
    {
      "start": 126.72,
      "end": 128.34,
      "text": "它把一堆分散的语音能力",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.1118585869456022,
      "no_speech_prob": 1.8711828961293087e-11,
      "compression_ratio": 1.3008595988538683
    },
    {
      "start": 128.34,
      "end": 129.72,
      "text": "变成一套本地系统",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.1118585869456022,
      "no_speech_prob": 1.8711828961293087e-11,
      "compression_ratio": 1.3008595988538683
    },
    {
      "start": 129.72,
      "end": 131.6,
      "text": "前面那个实时语音系统",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.1118585869456022,
      "no_speech_prob": 1.8711828961293087e-11,
      "compression_ratio": 1.3008595988538683
    },
    {
      "start": 131.6,
      "end": 133.1,
      "text": "等会我会完整演示",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.1118585869456022,
      "no_speech_prob": 1.8711828961293087e-11,
      "compression_ratio": 1.3008595988538683
    },
    {
      "start": 133.1,
      "end": 135.8,
      "text": "后面还有一个更好玩的音乐生成能力",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.1118585869456022,
      "no_speech_prob": 1.8711828961293087e-11,
      "compression_ratio": 1.3008595988538683
    },
    {
      "start": 135.8,
      "end": 137.16,
      "text": "歌曲旋律不变",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.1118585869456022,
      "no_speech_prob": 1.8711828961293087e-11,
      "compression_ratio": 1.3008595988538683
    },
    {
      "start": 137.16,
      "end": 138.44,
      "text": "直接换词翻唱",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.1118585869456022,
      "no_speech_prob": 1.8711828961293087e-11,
      "compression_ratio": 1.3008595988538683
    },
    {
      "start": 138.44,
      "end": 140.18,
      "text": "最后过一遍声音转换",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.1118585869456022,
      "no_speech_prob": 1.8711828961293087e-11,
      "compression_ratio": 1.3008595988538683
    },
    {
      "start": 140.18,
      "end": 141.58,
      "text": "设计等扩展功能",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.1118585869456022,
      "no_speech_prob": 1.8711828961293087e-11,
      "compression_ratio": 1.3008595988538683
    },
    {
      "start": 141.58,
      "end": 143.34,
      "text": "以及整合包的下载和使用",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.1118585869456022,
      "no_speech_prob": 1.8711828961293087e-11,
      "compression_ratio": 1.3008595988538683
    },
    {
      "start": 143.34,
      "end": 144.32,
      "text": "先看TTS",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.1118585869456022,
      "no_speech_prob": 1.8711828961293087e-11,
      "compression_ratio": 1.3008595988538683
    },
    {
      "start": 144.32,
      "end": 145.8,
      "text": "也就是文本转语音",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.1118585869456022,
      "no_speech_prob": 1.8711828961293087e-11,
      "compression_ratio": 1.3008595988538683
    },
    {
      "start": 145.8,
      "end": 146.86,
      "text": "这里列的模型",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.165881628148696,
      "no_speech_prob": 1.5545955148188462e-11,
      "compression_ratio": 1.1474926253687316
    },
    {
      "start": 146.86,
      "end": 151.16,
      "text": "整合包里自带了Pocket TTS和Chemun3 TTS 0.6B",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.165881628148696,
      "no_speech_prob": 1.5545955148188462e-11,
      "compression_ratio": 1.1474926253687316
    },
    {
      "start": 151.16,
      "end": 152.94,
      "text": "它们都支持参考音频",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.165881628148696,
      "no_speech_prob": 1.5545955148188462e-11,
      "compression_ratio": 1.1474926253687316
    },
    {
      "start": 152.94,
      "end": 154.96,
      "text": "也就意味着可以做声音克隆",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.165881628148696,
      "no_speech_prob": 1.5545955148188462e-11,
      "compression_ratio": 1.1474926253687316
    },
    {
      "start": 154.96,
      "end": 156.56,
      "text": "基本用法非常简单",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.165881628148696,
      "no_speech_prob": 1.5545955148188462e-11,
      "compression_ratio": 1.1474926253687316
    },
    {
      "start": 156.56,
      "end": 157.62,
      "text": "选择模型",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.165881628148696,
      "no_speech_prob": 1.5545955148188462e-11,
      "compression_ratio": 1.1474926253687316
    },
    {
      "start": 157.62,
      "end": 159.14,
      "text": "点击加载",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.165881628148696,
      "no_speech_prob": 1.5545955148188462e-11,
      "compression_ratio": 1.1474926253687316
    },
    {
      "start": 159.14,
      "end": 161.06,
      "text": "加载完会有提示",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.165881628148696,
      "no_speech_prob": 1.5545955148188462e-11,
      "compression_ratio": 1.1474926253687316
    },
    {
      "start": 161.06,
      "end": 163.06,
      "text": "然后选一段参考音频",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.165881628148696,
      "no_speech_prob": 1.5545955148188462e-11,
      "compression_ratio": 1.1474926253687316
    },
    {
      "start": 163.06,
      "end": 165.12,
      "text": "再输入合成文本",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.165881628148696,
      "no_speech_prob": 1.5545955148188462e-11,
      "compression_ratio": 1.1474926253687316
    },
    {
      "start": 165.12,
      "end": 166.4,
      "text": "点击生成",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.165881628148696,
      "no_speech_prob": 1.5545955148188462e-11,
      "compression_ratio": 1.1474926253687316
    },
    {
      "start": 166.4,
      "end": 168.34,
      "text": "这种一句话的声音克隆",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.165881628148696,
      "no_speech_prob": 1.5545955148188462e-11,
      "compression_ratio": 1.1474926253687316
    },
    {
      "start": 168.34,
      "end": 169.62,
      "text": "基本上是秒生成",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.165881628148696,
      "no_speech_prob": 1.5545955148188462e-11,
      "compression_ratio": 1.1474926253687316
    },
    {
      "start": 169.62,
      "end": 171.96,
      "text": "OpenAI launched GPT Live",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.165881628148696,
      "no_speech_prob": 1.5545955148188462e-11,
      "compression_ratio": 1.1474926253687316
    },
    {
      "start": 171.96,
      "end": 174.68,
      "text": "A full duplex voice system for chat GPT",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.165881628148696,
      "no_speech_prob": 1.5545955148188462e-11,
      "compression_ratio": 1.1474926253687316
    },
    {
      "start": 174.68,
      "end": 175.56,
      "text": "不过要注意",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.165881628148696,
      "no_speech_prob": 1.5545955148188462e-11,
      "compression_ratio": 1.1474926253687316
    },
    {
      "start": 175.56,
      "end": 177.36,
      "text": "Pocket TTS不支持中文",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.0987540330037032,
      "no_speech_prob": 1.619115778950242e-11,
      "compression_ratio": 1.2369942196531791
    },
    {
      "start": 177.36,
      "end": 178.88,
      "text": "要生成中文语音",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.0987540330037032,
      "no_speech_prob": 1.619115778950242e-11,
      "compression_ratio": 1.2369942196531791
    },
    {
      "start": 178.88,
      "end": 181.6,
      "text": "可以换成千问3 TTS 0.6B",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.0987540330037032,
      "no_speech_prob": 1.619115778950242e-11,
      "compression_ratio": 1.2369942196531791
    },
    {
      "start": 181.6,
      "end": 183.4,
      "text": "它的参数虽然不大",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.0987540330037032,
      "no_speech_prob": 1.619115778950242e-11,
      "compression_ratio": 1.2369942196531791
    },
    {
      "start": 183.4,
      "end": 185.02,
      "text": "但合成效果相当不错",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.0987540330037032,
      "no_speech_prob": 1.619115778950242e-11,
      "compression_ratio": 1.2369942196531791
    },
    {
      "start": 185.02,
      "end": 187.18,
      "text": "全双功语音系统",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.0987540330037032,
      "no_speech_prob": 1.619115778950242e-11,
      "compression_ratio": 1.2369942196531791
    },
    {
      "start": 187.18,
      "end": 190.52,
      "text": "支持实时打断联网搜索和视觉理解",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.0987540330037032,
      "no_speech_prob": 1.619115778950242e-11,
      "compression_ratio": 1.2369942196531791
    },
    {
      "start": 190.52,
      "end": 192.76,
      "text": "这两款模型显存占用都很小",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.0987540330037032,
      "no_speech_prob": 1.619115778950242e-11,
      "compression_ratio": 1.2369942196531791
    },
    {
      "start": 192.76,
      "end": 194.3,
      "text": "非常适合低配显卡",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.0987540330037032,
      "no_speech_prob": 1.619115778950242e-11,
      "compression_ratio": 1.2369942196531791
    },
    {
      "start": 194.3,
      "end": 195.86,
      "text": "以及没有独显的机器",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.0987540330037032,
      "no_speech_prob": 1.619115778950242e-11,
      "compression_ratio": 1.2369942196531791
    },
    {
      "start": 195.86,
      "end": 196.98,
      "text": "刚上手",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.0987540330037032,
      "no_speech_prob": 1.619115778950242e-11,
      "compression_ratio": 1.2369942196531791
    },
    {
      "start": 196.98,
      "end": 199.08,
      "text": "合成设置里的参数都可以不用管",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.0987540330037032,
      "no_speech_prob": 1.619115778950242e-11,
      "compression_ratio": 1.2369942196531791
    },
    {
      "start": 199.08,
      "end": 199.88,
      "text": "保持默认",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.0987540330037032,
      "no_speech_prob": 1.619115778950242e-11,
      "compression_ratio": 1.2369942196531791
    },
    {
      "start": 199.88,
      "end": 201.26,
      "text": "先把模型跑起来",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.0987540330037032,
      "no_speech_prob": 1.619115778950242e-11,
      "compression_ratio": 1.2369942196531791
    },
    {
      "start": 201.26,
      "end": 202.52,
      "text": "听一下技术效果",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.0987540330037032,
      "no_speech_prob": 1.619115778950242e-11,
      "compression_ratio": 1.2369942196531791
    },
    {
      "start": 202.52,
      "end": 204.02,
      "text": "熟悉之后再去调温度",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.0987540330037032,
      "no_speech_prob": 1.619115778950242e-11,
      "compression_ratio": 1.2369942196531791
    },
    {
      "start": 204.02,
      "end": 205.02,
      "text": "采样这些参数",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.0987540330037032,
      "no_speech_prob": 1.619115778950242e-11,
      "compression_ratio": 1.2369942196531791
    },
    {
      "start": 205.02,
      "end": 207.72,
      "text": "列表里还有两个模型比较有特点",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.15548402612859552,
      "no_speech_prob": 1.854451141258817e-11,
      "compression_ratio": 1.342507645259939
    },
    {
      "start": 207.72,
      "end": 209.46,
      "text": "一个是Vox CPM2",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.15548402612859552,
      "no_speech_prob": 1.854451141258817e-11,
      "compression_ratio": 1.342507645259939
    },
    {
      "start": 209.46,
      "end": 211.68,
      "text": "它好玩的地方是能生成方言",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.15548402612859552,
      "no_speech_prob": 1.854451141258817e-11,
      "compression_ratio": 1.342507645259939
    },
    {
      "start": 211.68,
      "end": 213.66,
      "text": "给一段方言参考音频",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.15548402612859552,
      "no_speech_prob": 1.854451141258817e-11,
      "compression_ratio": 1.342507645259939
    },
    {
      "start": 213.66,
      "end": 215.56,
      "text": "我们每个人都在做小丑",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.15548402612859552,
      "no_speech_prob": 1.854451141258817e-11,
      "compression_ratio": 1.342507645259939
    },
    {
      "start": 215.56,
      "end": 217.86,
      "text": "就可以合成相应的方言语音",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.15548402612859552,
      "no_speech_prob": 1.854451141258817e-11,
      "compression_ratio": 1.342507645259939
    },
    {
      "start": 217.86,
      "end": 218.94,
      "text": "错了你又不认",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.15548402612859552,
      "no_speech_prob": 1.854451141258817e-11,
      "compression_ratio": 1.342507645259939
    },
    {
      "start": 218.94,
      "end": 220.04,
      "text": "认了你又不改",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.15548402612859552,
      "no_speech_prob": 1.854451141258817e-11,
      "compression_ratio": 1.342507645259939
    },
    {
      "start": 220.04,
      "end": 222.36,
      "text": "它支持国内大部分方言",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.15548402612859552,
      "no_speech_prob": 1.854451141258817e-11,
      "compression_ratio": 1.342507645259939
    },
    {
      "start": 222.36,
      "end": 223.66,
      "text": "还能克隆语气",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.15548402612859552,
      "no_speech_prob": 1.854451141258817e-11,
      "compression_ratio": 1.342507645259939
    },
    {
      "start": 223.66,
      "end": 225.62,
      "text": "想深入了解这个模型的能力",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.15548402612859552,
      "no_speech_prob": 1.854451141258817e-11,
      "compression_ratio": 1.342507645259939
    },
    {
      "start": 225.62,
      "end": 228.96,
      "text": "可以翻译一下我之前专门做的Vox CPM那期视频",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.15548402612859552,
      "no_speech_prob": 1.854451141258817e-11,
      "compression_ratio": 1.342507645259939
    },
    {
      "start": 228.96,
      "end": 231.36,
      "text": "另一个有特点的是Voic Voice",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.15548402612859552,
      "no_speech_prob": 1.854451141258817e-11,
      "compression_ratio": 1.342507645259939
    },
    {
      "start": 231.36,
      "end": 233.76,
      "text": "就是开头速度对比的那个模型",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.15548402612859552,
      "no_speech_prob": 1.854451141258817e-11,
      "compression_ratio": 1.342507645259939
    },
    {
      "start": 233.76,
      "end": 235.66,
      "text": "那段近一万字的小说",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.11233290504006778,
      "no_speech_prob": 1.6636815189374765e-11,
      "compression_ratio": 1.25
    },
    {
      "start": 235.66,
      "end": 238.24,
      "text": "合成期间显存占用没超过7G",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.11233290504006778,
      "no_speech_prob": 1.6636815189374765e-11,
      "compression_ratio": 1.25
    },
    {
      "start": 238.24,
      "end": 240.82,
      "text": "生存出来的音频我全程听了一遍",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.11233290504006778,
      "no_speech_prob": 1.6636815189374765e-11,
      "compression_ratio": 1.25
    },
    {
      "start": 240.82,
      "end": 242.64,
      "text": "没有明显的漏读和错读",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.11233290504006778,
      "no_speech_prob": 1.6636815189374765e-11,
      "compression_ratio": 1.25
    },
    {
      "start": 242.64,
      "end": 245.04,
      "text": "忽然山腰里传来群马奔驰之声",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.11233290504006778,
      "no_speech_prob": 1.6636815189374765e-11,
      "compression_ratio": 1.25
    },
    {
      "start": 245.04,
      "end": 246.28,
      "text": "提声越来越响",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.11233290504006778,
      "no_speech_prob": 1.6636815189374765e-11,
      "compression_ratio": 1.25
    },
    {
      "start": 246.28,
      "end": 249.1,
      "text": "不久四面黄布大旗从山崖边升起",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.11233290504006778,
      "no_speech_prob": 1.6636815189374765e-11,
      "compression_ratio": 1.25
    },
    {
      "start": 249.1,
      "end": 250.2,
      "text": "这里有个注意点",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.11233290504006778,
      "no_speech_prob": 1.6636815189374765e-11,
      "compression_ratio": 1.25
    },
    {
      "start": 250.2,
      "end": 251.7,
      "text": "不管你用哪款模型",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.11233290504006778,
      "no_speech_prob": 1.6636815189374765e-11,
      "compression_ratio": 1.25
    },
    {
      "start": 251.7,
      "end": 253.44,
      "text": "参考音频都不要太长",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.11233290504006778,
      "no_speech_prob": 1.6636815189374765e-11,
      "compression_ratio": 1.25
    },
    {
      "start": 253.44,
      "end": 255.0,
      "text": "建议控制在10秒以内",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.11233290504006778,
      "no_speech_prob": 1.6636815189374765e-11,
      "compression_ratio": 1.25
    },
    {
      "start": 255.0,
      "end": 257.7,
      "text": "否则会明显拖慢合成速度",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.11233290504006778,
      "no_speech_prob": 1.6636815189374765e-11,
      "compression_ratio": 1.25
    },
    {
      "start": 257.7,
      "end": 258.78,
      "text": "另外还有个小技巧",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.11233290504006778,
      "no_speech_prob": 1.6636815189374765e-11,
      "compression_ratio": 1.25
    },
    {
      "start": 258.78,
      "end": 260.86,
      "text": "如果有些参考音频你经常用",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.11233290504006778,
      "no_speech_prob": 1.6636815189374765e-11,
      "compression_ratio": 1.25
    },
    {
      "start": 260.86,
      "end": 263.6,
      "text": "可以把它们放到WebUI下的Voice目标里",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.11233290504006778,
      "no_speech_prob": 1.6636815189374765e-11,
      "compression_ratio": 1.25
    },
    {
      "start": 263.6,
      "end": 267.16,
      "text": "然后把文件名和对应的文本写到这个文件里",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.11582468892191793,
      "no_speech_prob": 1.301001320563877e-11,
      "compression_ratio": 1.3211009174311927
    },
    {
      "start": 267.16,
      "end": 270.24,
      "text": "刷新列表就可以直接选了",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.11582468892191793,
      "no_speech_prob": 1.301001320563877e-11,
      "compression_ratio": 1.3211009174311927
    },
    {
      "start": 270.24,
      "end": 272.02,
      "text": "临时上传的参考音频",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.11582468892191793,
      "no_speech_prob": 1.301001320563877e-11,
      "compression_ratio": 1.3211009174311927
    },
    {
      "start": 272.02,
      "end": 274.3,
      "text": "下面对应的文本也不用手动输入",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.11582468892191793,
      "no_speech_prob": 1.301001320563877e-11,
      "compression_ratio": 1.3211009174311927
    },
    {
      "start": 274.3,
      "end": 276.14,
      "text": "可以先到ASR标签里",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.11582468892191793,
      "no_speech_prob": 1.301001320563877e-11,
      "compression_ratio": 1.3211009174311927
    },
    {
      "start": 276.14,
      "end": 278.02,
      "text": "把音频内容转写出来",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.11582468892191793,
      "no_speech_prob": 1.301001320563877e-11,
      "compression_ratio": 1.3211009174311927
    },
    {
      "start": 278.02,
      "end": 280.92,
      "text": "这就引出了AudioCPP第二个核心能力",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.11582468892191793,
      "no_speech_prob": 1.301001320563877e-11,
      "compression_ratio": 1.3211009174311927
    },
    {
      "start": 280.92,
      "end": 282.56,
      "text": "ASR语音转文字",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.11582468892191793,
      "no_speech_prob": 1.301001320563877e-11,
      "compression_ratio": 1.3211009174311927
    },
    {
      "start": 282.56,
      "end": 285.36,
      "text": "我个人常用的模型是千问3ASR",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.11582468892191793,
      "no_speech_prob": 1.301001320563877e-11,
      "compression_ratio": 1.3211009174311927
    },
    {
      "start": 285.36,
      "end": 287.12,
      "text": "等会要演示的实时语音",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.11582468892191793,
      "no_speech_prob": 1.301001320563877e-11,
      "compression_ratio": 1.3211009174311927
    },
    {
      "start": 287.12,
      "end": 288.86,
      "text": "我的声音转文本用的就是它",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.11582468892191793,
      "no_speech_prob": 1.301001320563877e-11,
      "compression_ratio": 1.3211009174311927
    },
    {
      "start": 288.86,
      "end": 290.22,
      "text": "中英文都支持",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.11582468892191793,
      "no_speech_prob": 1.301001320563877e-11,
      "compression_ratio": 1.3211009174311927
    },
    {
      "start": 290.22,
      "end": 291.3,
      "text": "速度也很快",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.11582468892191793,
      "no_speech_prob": 1.301001320563877e-11,
      "compression_ratio": 1.3211009174311927
    },
    {
      "start": 291.3,
      "end": 292.64,
      "text": "这段音频两分多钟",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.11582468892191793,
      "no_speech_prob": 1.301001320563877e-11,
      "compression_ratio": 1.3211009174311927
    },
    {
      "start": 292.64,
      "end": 294.32,
      "text": "转录只用了21秒",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10616024037733435,
      "no_speech_prob": 1.4347900471889385e-11,
      "compression_ratio": 1.2861356932153392
    },
    {
      "start": 294.32,
      "end": 296.42,
      "text": "如果转录的是多人对话",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10616024037733435,
      "no_speech_prob": 1.4347900471889385e-11,
      "compression_ratio": 1.2861356932153392
    },
    {
      "start": 296.42,
      "end": 298.54,
      "text": "要勾选对话模式",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10616024037733435,
      "no_speech_prob": 1.4347900471889385e-11,
      "compression_ratio": 1.2861356932153392
    },
    {
      "start": 298.54,
      "end": 301.6,
      "text": "转出来的就是这种带说话人序列的文本",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10616024037733435,
      "no_speech_prob": 1.4347900471889385e-11,
      "compression_ratio": 1.2861356932153392
    },
    {
      "start": 301.6,
      "end": 303.46,
      "text": "不过对话模式有长度限制",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10616024037733435,
      "no_speech_prob": 1.4347900471889385e-11,
      "compression_ratio": 1.2861356932153392
    },
    {
      "start": 303.46,
      "end": 305.84,
      "text": "目前更适合较短的对话片段",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10616024037733435,
      "no_speech_prob": 1.4347900471889385e-11,
      "compression_ratio": 1.2861356932153392
    },
    {
      "start": 305.84,
      "end": 308.3,
      "text": "单人语音转写没有长度限制",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10616024037733435,
      "no_speech_prob": 1.4347900471889385e-11,
      "compression_ratio": 1.2861356932153392
    },
    {
      "start": 308.3,
      "end": 309.22,
      "text": "歌词也能识别",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10616024037733435,
      "no_speech_prob": 1.4347900471889385e-11,
      "compression_ratio": 1.2861356932153392
    },
    {
      "start": 309.22,
      "end": 311.86,
      "text": "等下做AI翻唱要拿原曲歌词",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10616024037733435,
      "no_speech_prob": 1.4347900471889385e-11,
      "compression_ratio": 1.2861356932153392
    },
    {
      "start": 311.86,
      "end": 314.12,
      "text": "就可以先到这里做一遍初步识别",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10616024037733435,
      "no_speech_prob": 1.4347900471889385e-11,
      "compression_ratio": 1.2861356932153392
    },
    {
      "start": 314.12,
      "end": 315.22,
      "text": "再手动校对",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10616024037733435,
      "no_speech_prob": 1.4347900471889385e-11,
      "compression_ratio": 1.2861356932153392
    },
    {
      "start": 315.22,
      "end": 317.66,
      "text": "有了TTS和ASR",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10616024037733435,
      "no_speech_prob": 1.4347900471889385e-11,
      "compression_ratio": 1.2861356932153392
    },
    {
      "start": 317.66,
      "end": 319.5,
      "text": "现在我们可以把它们拼起来用",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10616024037733435,
      "no_speech_prob": 1.4347900471889385e-11,
      "compression_ratio": 1.2861356932153392
    },
    {
      "start": 319.5,
      "end": 322.26,
      "text": "这是我认为Audio CPP最有价值的地方",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10616024037733435,
      "no_speech_prob": 1.4347900471889385e-11,
      "compression_ratio": 1.2861356932153392
    },
    {
      "start": 322.26,
      "end": 323.84,
      "text": "本地实时语音系统",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10756157539986275,
      "no_speech_prob": 1.9528620040509814e-11,
      "compression_ratio": 1.404040404040404
    },
    {
      "start": 323.84,
      "end": 326.52,
      "text": "我正在测试一个实时语音系统",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10756157539986275,
      "no_speech_prob": 1.9528620040509814e-11,
      "compression_ratio": 1.404040404040404
    },
    {
      "start": 326.52,
      "end": 327.68,
      "text": "好的",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10756157539986275,
      "no_speech_prob": 1.9528620040509814e-11,
      "compression_ratio": 1.404040404040404
    },
    {
      "start": 327.68,
      "end": 330.28,
      "text": "我正在实时语音系统里和你对话",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10756157539986275,
      "no_speech_prob": 1.9528620040509814e-11,
      "compression_ratio": 1.404040404040404
    },
    {
      "start": 330.28,
      "end": 332.14,
      "text": "如果你想测试它的稳定性",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10756157539986275,
      "no_speech_prob": 1.9528620040509814e-11,
      "compression_ratio": 1.404040404040404
    },
    {
      "start": 332.14,
      "end": 333.54,
      "text": "延迟或其他功能",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10756157539986275,
      "no_speech_prob": 1.9528620040509814e-11,
      "compression_ratio": 1.404040404040404
    },
    {
      "start": 333.54,
      "end": 335.44,
      "text": "请随时提问或下达指令",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10756157539986275,
      "no_speech_prob": 1.9528620040509814e-11,
      "compression_ratio": 1.404040404040404
    },
    {
      "start": 335.44,
      "end": 338.56,
      "text": "语音识别和合成都是跑在本地的",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10756157539986275,
      "no_speech_prob": 1.9528620040509814e-11,
      "compression_ratio": 1.404040404040404
    },
    {
      "start": 338.56,
      "end": 339.96,
      "text": "明白了",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10756157539986275,
      "no_speech_prob": 1.9528620040509814e-11,
      "compression_ratio": 1.404040404040404
    },
    {
      "start": 339.96,
      "end": 342.24,
      "text": "本地运行意味着响应速度更快",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10756157539986275,
      "no_speech_prob": 1.9528620040509814e-11,
      "compression_ratio": 1.404040404040404
    },
    {
      "start": 342.24,
      "end": 343.98,
      "text": "同时数据也更安全",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10756157539986275,
      "no_speech_prob": 1.9528620040509814e-11,
      "compression_ratio": 1.404040404040404
    },
    {
      "start": 343.98,
      "end": 346.94,
      "text": "这是一个很注重隐私和效率的设计思路",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10756157539986275,
      "no_speech_prob": 1.9528620040509814e-11,
      "compression_ratio": 1.404040404040404
    },
    {
      "start": 346.94,
      "end": 348.64,
      "text": "大语言模型是云端",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10756157539986275,
      "no_speech_prob": 1.9528620040509814e-11,
      "compression_ratio": 1.404040404040404
    },
    {
      "start": 348.64,
      "end": 351.38,
      "text": "你觉得这个架构是否合理",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10756157539986275,
      "no_speech_prob": 1.9528620040509814e-11,
      "compression_ratio": 1.404040404040404
    },
    {
      "start": 351.38,
      "end": 353.84,
      "text": "这个架构是合理的",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.14488825590714163,
      "no_speech_prob": 1.6369456137810268e-11,
      "compression_ratio": 1.2267441860465116
    },
    {
      "start": 353.84,
      "end": 354.7,
      "text": "从我说完话",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.14488825590714163,
      "no_speech_prob": 1.6369456137810268e-11,
      "compression_ratio": 1.2267441860465116
    },
    {
      "start": 354.7,
      "end": 355.78,
      "text": "到它开始回答",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.14488825590714163,
      "no_speech_prob": 1.6369456137810268e-11,
      "compression_ratio": 1.2267441860465116
    },
    {
      "start": 355.78,
      "end": 356.94,
      "text": "中间大概一秒",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.14488825590714163,
      "no_speech_prob": 1.6369456137810268e-11,
      "compression_ratio": 1.2267441860465116
    },
    {
      "start": 356.94,
      "end": 358.52,
      "text": "如果是一问对话",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.14488825590714163,
      "no_speech_prob": 1.6369456137810268e-11,
      "compression_ratio": 1.2267441860465116
    },
    {
      "start": 358.52,
      "end": 359.8,
      "text": "这个颜色会更低",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.14488825590714163,
      "no_speech_prob": 1.6369456137810268e-11,
      "compression_ratio": 1.2267441860465116
    },
    {
      "start": 359.8,
      "end": 361.22,
      "text": "What can you do for me",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.14488825590714163,
      "no_speech_prob": 1.6369456137810268e-11,
      "compression_ratio": 1.2267441860465116
    },
    {
      "start": 361.22,
      "end": 363.1,
      "text": "I can answer questions",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.14488825590714163,
      "no_speech_prob": 1.6369456137810268e-11,
      "compression_ratio": 1.2267441860465116
    },
    {
      "start": 363.1,
      "end": 364.56,
      "text": "Summarize information",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.14488825590714163,
      "no_speech_prob": 1.6369456137810268e-11,
      "compression_ratio": 1.2267441860465116
    },
    {
      "start": 364.56,
      "end": 365.92,
      "text": "整个流程就是",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.14488825590714163,
      "no_speech_prob": 1.6369456137810268e-11,
      "compression_ratio": 1.2267441860465116
    },
    {
      "start": 365.92,
      "end": 366.98,
      "text": "麦克风输入",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.14488825590714163,
      "no_speech_prob": 1.6369456137810268e-11,
      "compression_ratio": 1.2267441860465116
    },
    {
      "start": 366.98,
      "end": 368.34,
      "text": "ASR转文本",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.14488825590714163,
      "no_speech_prob": 1.6369456137810268e-11,
      "compression_ratio": 1.2267441860465116
    },
    {
      "start": 368.34,
      "end": 369.84,
      "text": "大语言模型声称回答",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.14488825590714163,
      "no_speech_prob": 1.6369456137810268e-11,
      "compression_ratio": 1.2267441860465116
    },
    {
      "start": 369.84,
      "end": 371.44,
      "text": "TTS再把它朗读出来",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.14488825590714163,
      "no_speech_prob": 1.6369456137810268e-11,
      "compression_ratio": 1.2267441860465116
    },
    {
      "start": 371.44,
      "end": 372.3,
      "text": "这就是一个",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.14488825590714163,
      "no_speech_prob": 1.6369456137810268e-11,
      "compression_ratio": 1.2267441860465116
    },
    {
      "start": 372.3,
      "end": 373.7,
      "text": "本地语音交互系统的",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.14488825590714163,
      "no_speech_prob": 1.6369456137810268e-11,
      "compression_ratio": 1.2267441860465116
    },
    {
      "start": 373.7,
      "end": 374.7,
      "text": "基本形态",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.14488825590714163,
      "no_speech_prob": 1.6369456137810268e-11,
      "compression_ratio": 1.2267441860465116
    },
    {
      "start": 374.7,
      "end": 375.22,
      "text": "设置里",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.14488825590714163,
      "no_speech_prob": 1.6369456137810268e-11,
      "compression_ratio": 1.2267441860465116
    },
    {
      "start": 375.22,
      "end": 377.2,
      "text": "大语言模型这一步是活的",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.14488825590714163,
      "no_speech_prob": 1.6369456137810268e-11,
      "compression_ratio": 1.2267441860465116
    },
    {
      "start": 377.2,
      "end": 378.82,
      "text": "本地化的是语音这一层",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.14488825590714163,
      "no_speech_prob": 1.6369456137810268e-11,
      "compression_ratio": 1.2267441860465116
    },
    {
      "start": 378.82,
      "end": 380.12,
      "text": "你可以接本地",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.14488825590714163,
      "no_speech_prob": 1.6369456137810268e-11,
      "compression_ratio": 1.2267441860465116
    },
    {
      "start": 380.12,
      "end": 381.42,
      "text": "欧拉玛跑到语言模型",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.11342611814800062,
      "no_speech_prob": 2.4398949874382403e-11,
      "compression_ratio": 1.2826086956521738
    },
    {
      "start": 381.42,
      "end": 383.58,
      "text": "也可以接自己订阅的API",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.11342611814800062,
      "no_speech_prob": 2.4398949874382403e-11,
      "compression_ratio": 1.2826086956521738
    },
    {
      "start": 383.58,
      "end": 386.06,
      "text": "Audio CPP负责的是音频这一层",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.11342611814800062,
      "no_speech_prob": 2.4398949874382403e-11,
      "compression_ratio": 1.2826086956521738
    },
    {
      "start": 386.06,
      "end": 386.74,
      "text": "听和说",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.11342611814800062,
      "no_speech_prob": 2.4398949874382403e-11,
      "compression_ratio": 1.2826086956521738
    },
    {
      "start": 386.74,
      "end": 389.12,
      "text": "所以我说它不是一个声音克隆玩具",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.11342611814800062,
      "no_speech_prob": 2.4398949874382403e-11,
      "compression_ratio": 1.2826086956521738
    },
    {
      "start": 389.12,
      "end": 390.74,
      "text": "而是消费级显卡上的",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.11342611814800062,
      "no_speech_prob": 2.4398949874382403e-11,
      "compression_ratio": 1.2826086956521738
    },
    {
      "start": 390.74,
      "end": 392.12,
      "text": "本地AI语音底座",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.11342611814800062,
      "no_speech_prob": 2.4398949874382403e-11,
      "compression_ratio": 1.2826086956521738
    },
    {
      "start": 392.12,
      "end": 394.66,
      "text": "而且它用的是OpenAI兼容接口",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.11342611814800062,
      "no_speech_prob": 2.4398949874382403e-11,
      "compression_ratio": 1.2826086956521738
    },
    {
      "start": 394.66,
      "end": 396.28,
      "text": "你本地的其他应用",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.11342611814800062,
      "no_speech_prob": 2.4398949874382403e-11,
      "compression_ratio": 1.2826086956521738
    },
    {
      "start": 396.28,
      "end": 398.08,
      "text": "只要支持这种接口形式",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.11342611814800062,
      "no_speech_prob": 2.4398949874382403e-11,
      "compression_ratio": 1.2826086956521738
    },
    {
      "start": 398.08,
      "end": 399.56,
      "text": "就能直接调用服务",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.11342611814800062,
      "no_speech_prob": 2.4398949874382403e-11,
      "compression_ratio": 1.2826086956521738
    },
    {
      "start": 399.56,
      "end": 401.02,
      "text": "就像设置里这样写",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.11342611814800062,
      "no_speech_prob": 2.4398949874382403e-11,
      "compression_ratio": 1.2826086956521738
    },
    {
      "start": 401.02,
      "end": 402.32,
      "text": "TTS的地址",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.11342611814800062,
      "no_speech_prob": 2.4398949874382403e-11,
      "compression_ratio": 1.2826086956521738
    },
    {
      "start": 402.32,
      "end": 403.44,
      "text": "模型名称",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.11342611814800062,
      "no_speech_prob": 2.4398949874382403e-11,
      "compression_ratio": 1.2826086956521738
    },
    {
      "start": 403.44,
      "end": 405.08,
      "text": "ASR的地址",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.11342611814800062,
      "no_speech_prob": 2.4398949874382403e-11,
      "compression_ratio": 1.2826086956521738
    },
    {
      "start": 405.08,
      "end": 406.46,
      "text": "模型名称",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.11342611814800062,
      "no_speech_prob": 2.4398949874382403e-11,
      "compression_ratio": 1.2826086956521738
    },
    {
      "start": 406.46,
      "end": 408.32,
      "text": "朗读的声音也可以自定义",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.11342611814800062,
      "no_speech_prob": 2.4398949874382403e-11,
      "compression_ratio": 1.2826086956521738
    },
    {
      "start": 408.32,
      "end": 411.32,
      "text": "比如我在这个OpenMagic AI应用里",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.11063346421787504,
      "no_speech_prob": 1.1084341777767293e-11,
      "compression_ratio": 1.2806451612903227
    },
    {
      "start": 411.32,
      "end": 413.94,
      "text": "直接调用AudioCPP的TTS服务",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.11063346421787504,
      "no_speech_prob": 1.1084341777767293e-11,
      "compression_ratio": 1.2806451612903227
    },
    {
      "start": 413.94,
      "end": 415.28,
      "text": "填写UIL",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.11063346421787504,
      "no_speech_prob": 1.1084341777767293e-11,
      "compression_ratio": 1.2806451612903227
    },
    {
      "start": 415.28,
      "end": 416.52,
      "text": "还有模型名称",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.11063346421787504,
      "no_speech_prob": 1.1084341777767293e-11,
      "compression_ratio": 1.2806451612903227
    },
    {
      "start": 416.52,
      "end": 418.8,
      "text": "由于这个应用没法填参考音频",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.11063346421787504,
      "no_speech_prob": 1.1084341777767293e-11,
      "compression_ratio": 1.2806451612903227
    },
    {
      "start": 418.8,
      "end": 422.12,
      "text": "所以我后台单独起一个AudioCPP服务",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.11063346421787504,
      "no_speech_prob": 1.1084341777767293e-11,
      "compression_ratio": 1.2806451612903227
    },
    {
      "start": 422.12,
      "end": 423.58,
      "text": "启动时带上参考音频",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.11063346421787504,
      "no_speech_prob": 1.1084341777767293e-11,
      "compression_ratio": 1.2806451612903227
    },
    {
      "start": 423.58,
      "end": 425.98,
      "text": "回到OpenMagic测试",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.11063346421787504,
      "no_speech_prob": 1.1084341777767293e-11,
      "compression_ratio": 1.2806451612903227
    },
    {
      "start": 425.98,
      "end": 427.84,
      "text": "你好 这是一段测试语音",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.11063346421787504,
      "no_speech_prob": 1.1084341777767293e-11,
      "compression_ratio": 1.2806451612903227
    },
    {
      "start": 427.84,
      "end": 429.48,
      "text": "能正确返回语音",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.11063346421787504,
      "no_speech_prob": 1.1084341777767293e-11,
      "compression_ratio": 1.2806451612903227
    },
    {
      "start": 429.48,
      "end": 430.6,
      "text": "这样就接好了",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.11063346421787504,
      "no_speech_prob": 1.1084341777767293e-11,
      "compression_ratio": 1.2806451612903227
    },
    {
      "start": 430.6,
      "end": 433.52,
      "text": "这就是一个完全免费的本地TTS服务",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.11063346421787504,
      "no_speech_prob": 1.1084341777767293e-11,
      "compression_ratio": 1.2806451612903227
    },
    {
      "start": 433.52,
      "end": 435.22,
      "text": "如果你在做本地Agent",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.11063346421787504,
      "no_speech_prob": 1.1084341777767293e-11,
      "compression_ratio": 1.2806451612903227
    },
    {
      "start": 435.22,
      "end": 436.82,
      "text": "或者想给自己的应用",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.11063346421787504,
      "no_speech_prob": 1.1084341777767293e-11,
      "compression_ratio": 1.2806451612903227
    },
    {
      "start": 436.82,
      "end": 438.32,
      "text": "加上语音输入输出",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10178256549423548,
      "no_speech_prob": 1.2452616195146593e-11,
      "compression_ratio": 1.179930795847751
    },
    {
      "start": 438.32,
      "end": 440.86,
      "text": "AudioCPP提供的接口就非常有用",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10178256549423548,
      "no_speech_prob": 1.2452616195146593e-11,
      "compression_ratio": 1.179930795847751
    },
    {
      "start": 440.86,
      "end": 442.84,
      "text": "接下来我们看音乐生成",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10178256549423548,
      "no_speech_prob": 1.2452616195146593e-11,
      "compression_ratio": 1.179930795847751
    },
    {
      "start": 442.84,
      "end": 446.2,
      "text": "这部分也是我觉得AudioCPP很强的地方",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10178256549423548,
      "no_speech_prob": 1.2452616195146593e-11,
      "compression_ratio": 1.179930795847751
    },
    {
      "start": 446.2,
      "end": 455.08,
      "text": "它不只是把TTS和ASR放到一起",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10178256549423548,
      "no_speech_prob": 1.2452616195146593e-11,
      "compression_ratio": 1.179930795847751
    },
    {
      "start": 455.08,
      "end": 456.94,
      "text": "连音乐生成换词翻唱",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10178256549423548,
      "no_speech_prob": 1.2452616195146593e-11,
      "compression_ratio": 1.179930795847751
    },
    {
      "start": 456.94,
      "end": 458.58,
      "text": "这些原本比较重的模型",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10178256549423548,
      "no_speech_prob": 1.2452616195146593e-11,
      "compression_ratio": 1.179930795847751
    },
    {
      "start": 458.58,
      "end": 460.34,
      "text": "也开始接近同一个底座里",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10178256549423548,
      "no_speech_prob": 1.2452616195146593e-11,
      "compression_ratio": 1.179930795847751
    },
    {
      "start": 460.34,
      "end": 463.18,
      "text": "这个ACE STEP我之前介绍过",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10178256549423548,
      "no_speech_prob": 1.2452616195146593e-11,
      "compression_ratio": 1.179930795847751
    },
    {
      "start": 463.18,
      "end": 465.78,
      "text": "当时我在本地8G显存上跑不动翻唱",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.10178256549423548,
      "no_speech_prob": 1.2452616195146593e-11,
      "compression_ratio": 1.179930795847751
    },
    {
      "start": 465.78,
      "end": 467.54,
      "text": "最后是在Club上演示的",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.1670932326205941,
      "no_speech_prob": 8.015822380857962e-12,
      "compression_ratio": 0.9
    },
    {
      "start": 467.54,
      "end": 468.86,
      "text": "现在用AudioCPP",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.1670932326205941,
      "no_speech_prob": 8.015822380857962e-12,
      "compression_ratio": 0.9
    },
    {
      "start": 468.86,
      "end": 471.14,
      "text": "本地就能跑起来",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.1670932326205941,
      "no_speech_prob": 8.015822380857962e-12,
      "compression_ratio": 0.9
    },
    {
      "start": 471.14,
      "end": 482.2,
      "text": "翻唱可以按下面步骤操作",
      "chunk": 0,
      "language": "zh",
      "avg_logprob": -0.1670932326205941,
      "no_speech_prob": 8.015822380857962e-12,
      "compression_ratio": 0.9
    },
    {
      "start": 482.644,
      "end": 484.524,
      "text": "先上传要翻唱的歌曲",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.14869714187363448,
      "no_speech_prob": 1.1417101639099592e-11,
      "compression_ratio": 1.103896103896104
    },
    {
      "start": 484.524,
      "end": 486.064,
      "text": "然后点分析",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.14869714187363448,
      "no_speech_prob": 1.1417101639099592e-11,
      "compression_ratio": 1.103896103896104
    },
    {
      "start": 486.064,
      "end": 488.524,
      "text": "这个过程要等几十秒到几分钟",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.14869714187363448,
      "no_speech_prob": 1.1417101639099592e-11,
      "compression_ratio": 1.103896103896104
    },
    {
      "start": 488.524,
      "end": 489.984,
      "text": "具体看歌曲长度",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.14869714187363448,
      "no_speech_prob": 1.1417101639099592e-11,
      "compression_ratio": 1.103896103896104
    },
    {
      "start": 489.984,
      "end": 491.204,
      "text": "分析完成后",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.14869714187363448,
      "no_speech_prob": 1.1417101639099592e-11,
      "compression_ratio": 1.103896103896104
    },
    {
      "start": 491.204,
      "end": 493.664,
      "text": "合成翻唱的时长改成-1",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.14869714187363448,
      "no_speech_prob": 1.1417101639099592e-11,
      "compression_ratio": 1.103896103896104
    },
    {
      "start": 493.664,
      "end": 495.444,
      "text": "点开高级参数",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.14869714187363448,
      "no_speech_prob": 1.1417101639099592e-11,
      "compression_ratio": 1.103896103896104
    },
    {
      "start": 495.444,
      "end": 497.224,
      "text": "操作类型选Remix",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.14869714187363448,
      "no_speech_prob": 1.1417101639099592e-11,
      "compression_ratio": 1.103896103896104
    },
    {
      "start": 497.224,
      "end": 498.804,
      "text": "歌曲风格",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.0967322536632775,
      "no_speech_prob": 1.7989989706257425e-11,
      "compression_ratio": 1.289855072463768
    },
    {
      "start": 498.804,
      "end": 499.684,
      "text": "曲谱信息",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.0967322536632775,
      "no_speech_prob": 1.7989989706257425e-11,
      "compression_ratio": 1.289855072463768
    },
    {
      "start": 499.684,
      "end": 501.504,
      "text": "分析完以后就会自动填好",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.0967322536632775,
      "no_speech_prob": 1.7989989706257425e-11,
      "compression_ratio": 1.289855072463768
    },
    {
      "start": 501.504,
      "end": 502.044,
      "text": "不用动",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.0967322536632775,
      "no_speech_prob": 1.7989989706257425e-11,
      "compression_ratio": 1.289855072463768
    },
    {
      "start": 502.044,
      "end": 504.304,
      "text": "你只需要填上圆曲歌词",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.0967322536632775,
      "no_speech_prob": 1.7989989706257425e-11,
      "compression_ratio": 1.289855072463768
    },
    {
      "start": 504.304,
      "end": 507.044,
      "text": "提示词直接复制上面的圆曲描述",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.0967322536632775,
      "no_speech_prob": 1.7989989706257425e-11,
      "compression_ratio": 1.289855072463768
    },
    {
      "start": 507.044,
      "end": 509.084,
      "text": "最后填上你的翻唱歌词",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.0967322536632775,
      "no_speech_prob": 1.7989989706257425e-11,
      "compression_ratio": 1.289855072463768
    },
    {
      "start": 509.084,
      "end": 510.884,
      "text": "可以看一下显存占用",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.0967322536632775,
      "no_speech_prob": 1.7989989706257425e-11,
      "compression_ratio": 1.289855072463768
    },
    {
      "start": 510.884,
      "end": 512.484,
      "text": "虽然看上去超过8G",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.0967322536632775,
      "no_speech_prob": 1.7989989706257425e-11,
      "compression_ratio": 1.289855072463768
    },
    {
      "start": 512.484,
      "end": 513.584,
      "text": "但不会爆显存",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.0967322536632775,
      "no_speech_prob": 1.7989989706257425e-11,
      "compression_ratio": 1.289855072463768
    },
    {
      "start": 513.584,
      "end": 515.024,
      "text": "最后能顺利跑完",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.0967322536632775,
      "no_speech_prob": 1.7989989706257425e-11,
      "compression_ratio": 1.289855072463768
    },
    {
      "start": 515.024,
      "end": 516.924,
      "text": "换词翻唱的随机性比较大",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.0967322536632775,
      "no_speech_prob": 1.7989989706257425e-11,
      "compression_ratio": 1.289855072463768
    },
    {
      "start": 516.924,
      "end": 518.104,
      "text": "需要多试几次",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.0967322536632775,
      "no_speech_prob": 1.7989989706257425e-11,
      "compression_ratio": 1.289855072463768
    },
    {
      "start": 518.104,
      "end": 519.124,
      "text": "如果音质不好",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.0967322536632775,
      "no_speech_prob": 1.7989989706257425e-11,
      "compression_ratio": 1.289855072463768
    },
    {
      "start": 519.124,
      "end": 520.964,
      "text": "可以适当增加生存部署",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.0967322536632775,
      "no_speech_prob": 1.7989989706257425e-11,
      "compression_ratio": 1.289855072463768
    },
    {
      "start": 520.964,
      "end": 523.004,
      "text": "要是新词唱不准或者唱不出来",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.0967322536632775,
      "no_speech_prob": 1.7989989706257425e-11,
      "compression_ratio": 1.289855072463768
    },
    {
      "start": 523.004,
      "end": 525.044,
      "text": "可以调这个Flowedit参数",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.0967322536632775,
      "no_speech_prob": 1.7989989706257425e-11,
      "compression_ratio": 1.289855072463768
    },
    {
      "start": 525.044,
      "end": 527.184,
      "text": "从0.7到0.9去尝试",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.0967322536632775,
      "no_speech_prob": 1.7989989706257425e-11,
      "compression_ratio": 1.289855072463768
    },
    {
      "start": 527.184,
      "end": 529.304,
      "text": "如果你只想生成背景音乐",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.16193502289908274,
      "no_speech_prob": 1.1350067587179158e-11,
      "compression_ratio": 1.2846441947565543
    },
    {
      "start": 529.304,
      "end": 531.064,
      "text": "我推荐用Stable Audio",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.16193502289908274,
      "no_speech_prob": 1.1350067587179158e-11,
      "compression_ratio": 1.2846441947565543
    },
    {
      "start": 531.064,
      "end": 532.644,
      "text": "比ACE Stable更稳",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.16193502289908274,
      "no_speech_prob": 1.1350067587179158e-11,
      "compression_ratio": 1.2846441947565543
    },
    {
      "start": 532.644,
      "end": 535.624,
      "text": "声音转换相当于音色迁移",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.16193502289908274,
      "no_speech_prob": 1.1350067587179158e-11,
      "compression_ratio": 1.2846441947565543
    },
    {
      "start": 535.624,
      "end": 537.224,
      "text": "保持说话内容不变",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.16193502289908274,
      "no_speech_prob": 1.1350067587179158e-11,
      "compression_ratio": 1.2846441947565543
    },
    {
      "start": 537.224,
      "end": 538.284,
      "text": "语气不变",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.16193502289908274,
      "no_speech_prob": 1.1350067587179158e-11,
      "compression_ratio": 1.2846441947565543
    },
    {
      "start": 538.284,
      "end": 539.344,
      "text": "只把音色换掉",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.16193502289908274,
      "no_speech_prob": 1.1350067587179158e-11,
      "compression_ratio": 1.2846441947565543
    },
    {
      "start": 539.344,
      "end": 540.344,
      "text": "马匪",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.16193502289908274,
      "no_speech_prob": 1.1350067587179158e-11,
      "compression_ratio": 1.2846441947565543
    },
    {
      "start": 540.344,
      "end": 542.924,
      "text": "任何时候都要搅",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.16193502289908274,
      "no_speech_prob": 1.1350067587179158e-11,
      "compression_ratio": 1.2846441947565543
    },
    {
      "start": 542.924,
      "end": 544.164,
      "text": "不搅不行",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.16193502289908274,
      "no_speech_prob": 1.1350067587179158e-11,
      "compression_ratio": 1.2846441947565543
    },
    {
      "start": 544.164,
      "end": 545.864,
      "text": "马匪",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.16193502289908274,
      "no_speech_prob": 1.1350067587179158e-11,
      "compression_ratio": 1.2846441947565543
    },
    {
      "start": 545.864,
      "end": 548.324,
      "text": "任何时候都要搅",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.16193502289908274,
      "no_speech_prob": 1.1350067587179158e-11,
      "compression_ratio": 1.2846441947565543
    },
    {
      "start": 548.324,
      "end": 549.544,
      "text": "不搅不行",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.16193502289908274,
      "no_speech_prob": 1.1350067587179158e-11,
      "compression_ratio": 1.2846441947565543
    },
    {
      "start": 549.544,
      "end": 552.124,
      "text": "歌声转换也是类似逻辑",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.16193502289908274,
      "no_speech_prob": 1.1350067587179158e-11,
      "compression_ratio": 1.2846441947565543
    },
    {
      "start": 552.124,
      "end": 552.924,
      "text": "但我实测",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.16193502289908274,
      "no_speech_prob": 1.1350067587179158e-11,
      "compression_ratio": 1.2846441947565543
    },
    {
      "start": 552.924,
      "end": 554.504,
      "text": "如果你追求追踪质量",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.16193502289908274,
      "no_speech_prob": 1.1350067587179158e-11,
      "compression_ratio": 1.2846441947565543
    },
    {
      "start": 554.504,
      "end": 556.824,
      "text": "我更推荐之前那套RVC流程",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.16193502289908274,
      "no_speech_prob": 1.1350067587179158e-11,
      "compression_ratio": 1.2846441947565543
    },
    {
      "start": 556.824,
      "end": 558.184,
      "text": "做歌声转换之前",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.10593068976151315,
      "no_speech_prob": 1.6132428726223225e-11,
      "compression_ratio": 1.419463087248322
    },
    {
      "start": 558.184,
      "end": 559.824,
      "text": "最好先做人声分离",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.10593068976151315,
      "no_speech_prob": 1.6132428726223225e-11,
      "compression_ratio": 1.419463087248322
    },
    {
      "start": 559.824,
      "end": 561.304,
      "text": "音频分离标签下",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.10593068976151315,
      "no_speech_prob": 1.6132428726223225e-11,
      "compression_ratio": 1.419463087248322
    },
    {
      "start": 561.304,
      "end": 562.544,
      "text": "就有相应的模型",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.10593068976151315,
      "no_speech_prob": 1.6132428726223225e-11,
      "compression_ratio": 1.419463087248322
    },
    {
      "start": 562.544,
      "end": 565.064,
      "text": "再往下是音频分析这一组",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.10593068976151315,
      "no_speech_prob": 1.6132428726223225e-11,
      "compression_ratio": 1.419463087248322
    },
    {
      "start": 565.064,
      "end": 566.644,
      "text": "它们都是工具类模型",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.10593068976151315,
      "no_speech_prob": 1.6132428726223225e-11,
      "compression_ratio": 1.419463087248322
    },
    {
      "start": 566.644,
      "end": 568.744,
      "text": "比如这个是实时语音检测",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.10593068976151315,
      "no_speech_prob": 1.6132428726223225e-11,
      "compression_ratio": 1.419463087248322
    },
    {
      "start": 568.744,
      "end": 570.724,
      "text": "下面这个是多人对话识别",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.10593068976151315,
      "no_speech_prob": 1.6132428726223225e-11,
      "compression_ratio": 1.419463087248322
    },
    {
      "start": 570.724,
      "end": 572.424,
      "text": "最后一个是声音设计",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.10593068976151315,
      "no_speech_prob": 1.6132428726223225e-11,
      "compression_ratio": 1.419463087248322
    },
    {
      "start": 572.424,
      "end": 573.504,
      "text": "这个挺有意思",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.10593068976151315,
      "no_speech_prob": 1.6132428726223225e-11,
      "compression_ratio": 1.419463087248322
    },
    {
      "start": 573.504,
      "end": 575.084,
      "text": "它其实类似声音克隆",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.10593068976151315,
      "no_speech_prob": 1.6132428726223225e-11,
      "compression_ratio": 1.419463087248322
    },
    {
      "start": 575.084,
      "end": 576.664,
      "text": "只是不需要参考音频",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.10593068976151315,
      "no_speech_prob": 1.6132428726223225e-11,
      "compression_ratio": 1.419463087248322
    },
    {
      "start": 576.664,
      "end": 578.804,
      "text": "你直接用文字描述一个声音",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.10593068976151315,
      "no_speech_prob": 1.6132428726223225e-11,
      "compression_ratio": 1.419463087248322
    },
    {
      "start": 578.804,
      "end": 580.564,
      "text": "比如磁性的中年男生",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.10593068976151315,
      "no_speech_prob": 1.6132428726223225e-11,
      "compression_ratio": 1.419463087248322
    },
    {
      "start": 580.564,
      "end": 581.404,
      "text": "语速偏慢",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.10593068976151315,
      "no_speech_prob": 1.6132428726223225e-11,
      "compression_ratio": 1.419463087248322
    },
    {
      "start": 581.404,
      "end": 582.044,
      "text": "波音枪",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.10593068976151315,
      "no_speech_prob": 1.6132428726223225e-11,
      "compression_ratio": 1.419463087248322
    },
    {
      "start": 582.044,
      "end": 583.804,
      "text": "模型就会按描述生成",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.10593068976151315,
      "no_speech_prob": 1.6132428726223225e-11,
      "compression_ratio": 1.419463087248322
    },
    {
      "start": 583.804,
      "end": 594.664,
      "text": "这个功能做临时配音比较好用",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.10120298679058369,
      "no_speech_prob": 1.4337619633209009e-11,
      "compression_ratio": 1.125
    },
    {
      "start": 594.664,
      "end": 596.864,
      "text": "最后说一下整合包",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.10120298679058369,
      "no_speech_prob": 1.4337619633209009e-11,
      "compression_ratio": 1.125
    },
    {
      "start": 596.864,
      "end": 599.264,
      "text": "Audio CPP本身是个C++项目",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.10120298679058369,
      "no_speech_prob": 1.4337619633209009e-11,
      "compression_ratio": 1.125
    },
    {
      "start": 599.264,
      "end": 601.484,
      "text": "官方目前主要靠命令行来使用",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.10120298679058369,
      "no_speech_prob": 1.4337619633209009e-11,
      "compression_ratio": 1.125
    },
    {
      "start": 601.484,
      "end": 603.484,
      "text": "对有经验的朋友来说问题不大",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.10120298679058369,
      "no_speech_prob": 1.4337619633209009e-11,
      "compression_ratio": 1.125
    },
    {
      "start": 603.484,
      "end": 605.024,
      "text": "但对没技术基础的",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.10120298679058369,
      "no_speech_prob": 1.4337619633209009e-11,
      "compression_ratio": 1.125
    },
    {
      "start": 605.024,
      "end": 606.584,
      "text": "这一步确实比较劝退",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.10120298679058369,
      "no_speech_prob": 1.4337619633209009e-11,
      "compression_ratio": 1.125
    },
    {
      "start": 606.584,
      "end": 608.644,
      "text": "所以我做了带界面的整合包",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.10120298679058369,
      "no_speech_prob": 1.4337619633209009e-11,
      "compression_ratio": 1.125
    },
    {
      "start": 608.644,
      "end": 609.744,
      "text": "解压就能用",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.10120298679058369,
      "no_speech_prob": 1.4337619633209009e-11,
      "compression_ratio": 1.125
    },
    {
      "start": 609.744,
      "end": 611.804,
      "text": "支持16-50系列的N卡",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.10120298679058369,
      "no_speech_prob": 1.4337619633209009e-11,
      "compression_ratio": 1.125
    },
    {
      "start": 611.804,
      "end": 614.064,
      "text": "目前A卡以及没有独显的机器",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.1433333953221639,
      "no_speech_prob": 1.5787504983877376e-11,
      "compression_ratio": 1.2108108108108109
    },
    {
      "start": 614.064,
      "end": 615.644,
      "text": "会自动走CPU模式",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.1433333953221639,
      "no_speech_prob": 1.5787504983877376e-11,
      "compression_ratio": 1.2108108108108109
    },
    {
      "start": 615.644,
      "end": 616.764,
      "text": "速度会慢一些",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.1433333953221639,
      "no_speech_prob": 1.5787504983877376e-11,
      "compression_ratio": 1.2108108108108109
    },
    {
      "start": 616.764,
      "end": 618.444,
      "text": "但跑几个轻量模型没问题",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.1433333953221639,
      "no_speech_prob": 1.5787504983877376e-11,
      "compression_ratio": 1.2108108108108109
    },
    {
      "start": 618.444,
      "end": 619.264,
      "text": "整合包里",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.1433333953221639,
      "no_speech_prob": 1.5787504983877376e-11,
      "compression_ratio": 1.2108108108108109
    },
    {
      "start": 619.264,
      "end": 621.304,
      "text": "Run Web UI是启动网页界面",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.1433333953221639,
      "no_speech_prob": 1.5787504983877376e-11,
      "compression_ratio": 1.2108108108108109
    },
    {
      "start": 621.304,
      "end": 622.484,
      "text": "想用哪个模型",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.1433333953221639,
      "no_speech_prob": 1.5787504983877376e-11,
      "compression_ratio": 1.2108108108108109
    },
    {
      "start": 622.484,
      "end": 623.884,
      "text": "界面里就能直接下载",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.1433333953221639,
      "no_speech_prob": 1.5787504983877376e-11,
      "compression_ratio": 1.2108108108108109
    },
    {
      "start": 623.884,
      "end": 625.844,
      "text": "想体验实时语音",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.1433333953221639,
      "no_speech_prob": 1.5787504983877376e-11,
      "compression_ratio": 1.2108108108108109
    },
    {
      "start": 625.844,
      "end": 627.644,
      "text": "先在Web UI里加载模型",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.1433333953221639,
      "no_speech_prob": 1.5787504983877376e-11,
      "compression_ratio": 1.2108108108108109
    },
    {
      "start": 627.644,
      "end": 629.744,
      "text": "再启动这个ASR服务",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.1433333953221639,
      "no_speech_prob": 1.5787504983877376e-11,
      "compression_ratio": 1.2108108108108109
    },
    {
      "start": 629.744,
      "end": 631.784,
      "text": "最后运行Run Real Time",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.1433333953221639,
      "no_speech_prob": 1.5787504983877376e-11,
      "compression_ratio": 1.2108108108108109
    },
    {
      "start": 631.784,
      "end": 634.264,
      "text": "配置里把接入的大模型改一下",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.1433333953221639,
      "no_speech_prob": 1.5787504983877376e-11,
      "compression_ratio": 1.2108108108108109
    },
    {
      "start": 634.264,
      "end": 635.824,
      "text": "如果你也用Deep Seek",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.1433333953221639,
      "no_speech_prob": 1.5787504983877376e-11,
      "compression_ratio": 1.2108108108108109
    },
    {
      "start": 635.824,
      "end": 638.124,
      "text": "API Key换成自己的就可以了",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.1433333953221639,
      "no_speech_prob": 1.5787504983877376e-11,
      "compression_ratio": 1.2108108108108109
    },
    {
      "start": 638.124,
      "end": 640.144,
      "text": "这一整套原代码我已经开完了",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.1433333953221639,
      "no_speech_prob": 1.5787504983877376e-11,
      "compression_ratio": 1.2108108108108109
    },
    {
      "start": 640.144,
      "end": 641.444,
      "text": "这是GitHub地址",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.1433333953221639,
      "no_speech_prob": 1.5787504983877376e-11,
      "compression_ratio": 1.2108108108108109
    },
    {
      "start": 641.444,
      "end": 643.344,
      "text": "使用中如果遇到问题",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.11580949900101642,
      "no_speech_prob": 1.4289483658835866e-11,
      "compression_ratio": 1.2737430167597765
    },
    {
      "start": 643.344,
      "end": 644.784,
      "text": "或者希望增加功能",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.11580949900101642,
      "no_speech_prob": 1.4289483658835866e-11,
      "compression_ratio": 1.2737430167597765
    },
    {
      "start": 644.784,
      "end": 646.444,
      "text": "可以在评论区告诉我",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.11580949900101642,
      "no_speech_prob": 1.4289483658835866e-11,
      "compression_ratio": 1.2737430167597765
    },
    {
      "start": 646.444,
      "end": 648.764,
      "text": "我会收集整理定期更新版本",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.11580949900101642,
      "no_speech_prob": 1.4289483658835866e-11,
      "compression_ratio": 1.2737430167597765
    },
    {
      "start": 648.764,
      "end": 649.704,
      "text": "总结一下",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.11580949900101642,
      "no_speech_prob": 1.4289483658835866e-11,
      "compression_ratio": 1.2737430167597765
    },
    {
      "start": 649.704,
      "end": 651.384,
      "text": "Audio CPP最大的意义",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.11580949900101642,
      "no_speech_prob": 1.4289483658835866e-11,
      "compression_ratio": 1.2737430167597765
    },
    {
      "start": 651.384,
      "end": 653.404,
      "text": "不是又多了一个声音客户工具",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.11580949900101642,
      "no_speech_prob": 1.4289483658835866e-11,
      "compression_ratio": 1.2737430167597765
    },
    {
      "start": 653.404,
      "end": 655.004,
      "text": "而是本地语音模型",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.11580949900101642,
      "no_speech_prob": 1.4289483658835866e-11,
      "compression_ratio": 1.2737430167597765
    },
    {
      "start": 655.004,
      "end": 657.264,
      "text": "开始变成一套统一的系统",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.11580949900101642,
      "no_speech_prob": 1.4289483658835866e-11,
      "compression_ratio": 1.2737430167597765
    },
    {
      "start": 657.264,
      "end": 658.804,
      "text": "大语言模型有Lama CPP",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.11580949900101642,
      "no_speech_prob": 1.4289483658835866e-11,
      "compression_ratio": 1.2737430167597765
    },
    {
      "start": 658.804,
      "end": 660.864,
      "text": "图片视频生成有Conf UI",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.11580949900101642,
      "no_speech_prob": 1.4289483658835866e-11,
      "compression_ratio": 1.2737430167597765
    },
    {
      "start": 660.864,
      "end": 662.964,
      "text": "语音模型现在也开始有了",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.11580949900101642,
      "no_speech_prob": 1.4289483658835866e-11,
      "compression_ratio": 1.2737430167597765
    },
    {
      "start": 662.964,
      "end": 664.324,
      "text": "自己的本地运行中心",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.11580949900101642,
      "no_speech_prob": 1.4289483658835866e-11,
      "compression_ratio": 1.2737430167597765
    },
    {
      "start": 664.324,
      "end": 665.904,
      "text": "对于普通用户来说",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.11580949900101642,
      "no_speech_prob": 1.4289483658835866e-11,
      "compression_ratio": 1.2737430167597765
    },
    {
      "start": 665.904,
      "end": 667.164,
      "text": "使用门槛更低了",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.11580949900101642,
      "no_speech_prob": 1.4289483658835866e-11,
      "compression_ratio": 1.2737430167597765
    },
    {
      "start": 667.164,
      "end": 668.424,
      "text": "对开发者来说",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.11580949900101642,
      "no_speech_prob": 1.4289483658835866e-11,
      "compression_ratio": 1.2737430167597765
    },
    {
      "start": 668.424,
      "end": 670.224,
      "text": "本地语音服务接入更容易",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.11580949900101642,
      "no_speech_prob": 1.4289483658835866e-11,
      "compression_ratio": 1.2737430167597765
    },
    {
      "start": 670.224,
      "end": 672.604,
      "text": "这也是我这期最想讲清楚的地方",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.1080185283314098,
      "no_speech_prob": 9.188244783076005e-12,
      "compression_ratio": 1.0419580419580419
    },
    {
      "start": 672.604,
      "end": 674.644,
      "text": "本期视频用到的所有链接",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.1080185283314098,
      "no_speech_prob": 9.188244783076005e-12,
      "compression_ratio": 1.0419580419580419
    },
    {
      "start": 674.644,
      "end": 676.124,
      "text": "我都放在评论区置顶",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.1080185283314098,
      "no_speech_prob": 9.188244783076005e-12,
      "compression_ratio": 1.0419580419580419
    },
    {
      "start": 676.124,
      "end": 677.784,
      "text": "这里是AI探索已发现",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.1080185283314098,
      "no_speech_prob": 9.188244783076005e-12,
      "compression_ratio": 1.0419580419580419
    },
    {
      "start": 677.784,
      "end": 678.624,
      "text": "感谢观看",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.1080185283314098,
      "no_speech_prob": 9.188244783076005e-12,
      "compression_ratio": 1.0419580419580419
    },
    {
      "start": 678.624,
      "end": 679.204,
      "text": "下期见",
      "chunk": 1,
      "language": "zh",
      "avg_logprob": -0.1080185283314098,
      "no_speech_prob": 9.188244783076005e-12,
      "compression_ratio": 1.0419580419580419
    }
  ],
  "segmented_transcription": {
    "status": "SEGMENTED_TRANSCRIPT_OK",
    "created_at": "2026-08-01T16:59:18",
    "manifest": "/Users/sagawa/AI_WORK/video_notes/url/20260801_165831__本地语音_AI_终于统一了_实时对话_声音克隆_AI_翻唱8G_显存全跑通_audio.cpp_整合包/segmented_work/chunk_manifest.json",
    "audio": "/Users/sagawa/AI_WORK/video_notes/url/20260801_165831__本地语音_AI_终于统一了_实时对话_声音克隆_AI_翻唱8G_显存全跑通_audio.cpp_整合包/audio_16k_mono.wav",
    "audio_sha256": "4aadc873e2b4e11e8c27eeb18e7fcd9a5db7cdce731c7206eee2434dec975ecf",
    "audio_duration": 686.016,
    "model": "mlx-community/whisper-large-v3-turbo",
    "language_hint": "zh",
    "condition_on_previous_text": false,
    "chunk_count": 2,
    "resumed_chunks": 0,
    "merged_segment_count": 352,
    "final_segment_end": 679.204,
    "tail_repeat_run": 1,
    "tail_repeat_text": "再启动这个ASR服务",
    "qc_ok": true,
    "qc_errors": []
  }
}