{"text": "这是Handy,它是一款开源的语音识别和语音听写应用。完全离线运行,没有云端,没有订阅,也不用账号。你还可以自己选择喜欢的听写模型。老师说,我觉得它现在已经成了我最喜欢的语音转文字工具。所以这七视频里,我们会看看Handy。了解它是怎么工作的,再做几项测试,看看它表现如何。以及和商业版Whisper Flow这类重量级产品比起来怎么样。会很有意思,那我们开始吧。到目前为止,我一直对语音听写应用有点怀疑,而且一直找不到一款我真心想用的。我不想用任何商业产品来付费做语音转写。而且我也担心,用一个本地语音转写模型的开源工具会把CPU占满。不过,Handy其实把这两个顾虑都打消了。先说说Handy到底是怎么构建的,又是怎么工作的。Handy由开发者CJ Pace和Terry一起开发。后端用的是Rust。上层的设置界面则用React和TypeScript。Transcribe CPP负负责运行Whisper系列模型。格式是GGM和GGY。Transcribe RS运行的是ParaKit。CPAL负责跨平台音频输入。RDV负责全局键盘快捷键。它还用到一个叫Roboto的Core。提供实时音频流和重采样。这个组合让它用起来特别简单,也特别清亮。连在我的MacBook上都很轻松。从用户角度看,整个流程也很简单。你先设一个自定义快捷键,然后可以切缓开关。或者长按来实现Push to Talk。按住的时候,Colero VAD会在后台悄悄过滤掉静音。这样就不会白白消耗算力去转写整段空白。松开后,Handy会把音频发给你选定的模型。再把转写结果直接粘贴到当时当前聚焦的输入框里。Handy并不是这方面独有的。市面上还有不少语音转写应用也差不多是这种工作方式。不过我最喜欢的是你可以自己选喜欢的模型。模型列表里的每个模型都有速度和准确率的指标。这有点像在Mario Kart里选自己最喜欢的角色。你可以决定哪项指标对你来说最重要。对我来说,我发现用ParaKit Unify 0,6BP模型就很好用。因为ParaKit是一个CPU1模型,而且按他们自己的数据。它在终端爱武芯片上的运行速度大约是5倍10时。我的MR Max远远超过这个Page。所以我把它放在后台跑的时候甚至都感觉不到。而且可选项真的很多。它们甚至还有专门针对某一种的模型。比如我们有一些俄语版本,还有一些乌克兰语版本。我很喜欢嗨这句使命宣言里的坦诚。Handy不是想成为最好的语音转文字应用,而是想成为最容易二次开发的那个。而且你甚至可以把自己微调过的Whisper G Gmail模型放进App的Models文件夹里。重启之后,它就会直接作为一个自定义选项出现。我觉得,如果任何时候要我在Handy和Whisper Flow之间选,我会选Handy,最大的原因就是它可以离线运行。你不需要注册账号,也没有订阅费。而且我很确定,所有我的语音输入音频样本都会留在我的机器上,不会被偷偷拿去训练别的语音模型。所以我真的很高兴。现在有了这样一个MIT许可的语音输入工具。听起来都很棒,不过,我们就来实际测试一下Pandy,看看它表现如何。我会把Handy的表现和Whisper Flow以及Google自家的转写服务做个对比。后者也可以在Google Docs里用。哦,昨天我去买了点菜,回家路上突然想起来。我得回头把我的SQLite数据库接到一个运行在我自定义GPU集群上的Cobot服务上。我能想到的一个办法是用一个叫Grog的LLM。另外还从CheckGPT得到了一些帮助。最后,我们做出了一个能运行的应用,但过程真的很折腾。我们把这两部分分开来说。你看,这里的Google语音输入服务要差很多,它不会加标点。也不知道怎么分句,它把Cobot识别错了。SQLite数据库也识别错了。甚至连LLM这个词都没认出来。它以为Grog是Rock。这个我最喜欢,它还把ChatGPT认成了ChatGPT。ChatGPT不错。嗯,所以你能明显看出来Handy好多了。因为它有标点,还能识别一些特定术语。而且大多数技术术语都认对了。现在我们也来把这个结果和Whisper Flow比一下。我要尽量把同样的文本重复的尽可能接近。公平地测试一下。嗯,昨天我去买了些杂货。回来的路上。我突然想起我得回家。把我的SQLite数据库接到一个运行在我自定义GPU集群上的Cobot服务上。我其实不知道该怎么做,所以只好找人帮忙。我能想到的唯一能帮上忙的工具。是一个叫Grook的LLM。后来,我也从ChatGPT那里得到了一点帮助。最后,我们是把应用跑起来了,但过程真的很折腾。好吧,你可以清楚地看到Whisper Flow做得最好。它把句子分得很准确,而且所有术语都没出错。毕竟这是商业产品,你会以为它会比开源工具更好。但说实话,这个结果和Whisper Flow之间并没有那么大差别。所以,我还是更喜欢用Handy,因为它免费、私密。而且是开源的。不过,Whisper Flow在这里确实做得很不错。所以总的来说,它还是一个很靠谱的服务。好了,各位,概括来说就是这样。这个功能确实很实用。我刚才真不是开玩笑。说它已经成了我最喜欢的语音听写工具。我甚至有点担心,要是我用得太频繁。自己可能都会忘了怎么用键盘打字。不过,这只是我的个人看法和观察。那你觉得Handy怎么样?你是过了吗?你会用吗?欢迎在下方评论区告诉我们。各位,如果你喜欢这类技术解析,也请在视频下方点个赞。让我知道,别忘了订阅我们的频道。我是BetterStack的Andreas,我们下期视频再见。音乐", "segments": [{"id": 0, "seek": 0, "start": 0.0, "end": 28.36, "text": "这是Handy,它是一款开源的语音识别和语音听写应用。完全离线运行,没有云端,没有订阅,也不用账号。你还可以自己选择喜欢的听写模型。老师说,我觉得它现在已经成了我最喜欢的语音转文字工具。所以这七视频里,我们会看看Handy。了解它是怎么工作的,再做几项测试,看看它表现如何。", "tokens": [50365, 27455, 39, 11425, 11, 11284, 1541, 2257, 48798, 18937, 47402, 1546, 5233, 255, 18034, 5233, 228, 18453, 12565, 5233, 255, 18034, 31022, 5676, 247, 44297, 9254, 1543, 37100, 18614, 119, 16853, 123, 3316, 238, 8082, 11, 17944, 1369, 239, 11957, 107, 11, 17944, 7422, 95, 10034, 227, 11, 6404, 24384, 18464, 99, 26987, 1543, 2166, 14852, 6723, 17645, 2215, 231, 6852, 102, 41606, 1546, 31022, 5676, 247, 41908, 39823, 1543, 10439, 29186, 8090, 11, 45681, 11284, 25040, 49161, 11336, 2289, 1654, 8661, 41606, 1546, 5233, 255, 18034, 17819, 105, 17174, 22381, 23323, 39806, 1543, 7239, 5562, 29920, 40656, 39752, 15759, 11, 15003, 12949, 22884, 39, 11425, 1543, 2289, 17278, 11284, 1541, 15282, 41315, 1546, 11, 8623, 10907, 6336, 254, 10178, 117, 11038, 233, 5233, 243, 11, 22884, 11284, 17571, 20204, 43526, 1543, 51783], "temperature": 0, "avg_logprob": -0.06383182020748362, "compression_ratio": 1.2626262626262625, "no_speech_prob": 7.764875548099681e-12}, {"id": 1, "seek": 2836, "start": 28.36, "end": 56.06, "text": "以及和商业版Whisper Flow这类重量级产品比起来怎么样。会很有意思,那我们开始吧。到目前为止,我一直对语音听写应用有点怀疑,而且一直找不到一款我真心想用的。我不想用任何商业产品来付费做语音转写。而且我也担心,用一个本地语音转写模型的开源工具会把CPU占满。", "tokens": [50365, 40282, 12565, 45581, 940, 248, 42096, 2471, 271, 610, 32792, 5562, 22113, 119, 12624, 26748, 16853, 100, 1369, 100, 30246, 11706, 9147, 6912, 48200, 1543, 12949, 49295, 16697, 11, 4184, 15003, 45213, 6062, 1543, 4511, 39004, 13992, 30438, 11, 1654, 34448, 8713, 5233, 255, 18034, 31022, 5676, 247, 44297, 9254, 2412, 12579, 3757, 222, 18140, 239, 11, 22942, 34448, 25085, 35728, 2257, 48798, 1654, 6303, 7945, 7093, 9254, 1546, 1543, 25395, 7093, 9254, 26443, 11906, 45581, 940, 248, 1369, 100, 30246, 6912, 41469, 18464, 117, 10907, 5233, 255, 18034, 17819, 105, 5676, 247, 1543, 22942, 27761, 6852, 227, 7945, 11, 9254, 20182, 8802, 10928, 5233, 255, 18034, 17819, 105, 5676, 247, 41908, 39823, 1546, 18937, 47402, 23323, 39806, 12949, 16075, 34, 8115, 5322, 254, 15868, 94, 1543, 51750], "temperature": 0, "avg_logprob": -0.04374517921273035, "compression_ratio": 1.2218430034129693, "no_speech_prob": 9.745857766640942e-12}, {"id": 2, "seek": 5606, "start": 56.06, "end": 65.48, "text": "不过,Handy其实把这两个顾虑都打消了。先说说Handy到底是怎么构建的,又是怎么工作的。", "tokens": [50365, 1960, 16866, 11, 39, 11425, 41646, 16075, 5562, 36257, 7549, 10178, 122, 12026, 239, 7182, 12467, 28837, 2289, 1543, 10108, 8090, 8090, 39, 11425, 33883, 1541, 15282, 7360, 226, 34157, 1546, 11, 17047, 1541, 15282, 41315, 1546, 1543, 50836], "temperature": 0, "avg_logprob": -0.1285893808711659, "compression_ratio": 1.0845070422535212, "no_speech_prob": 1.6128581109553508e-11}, {"id": 3, "seek": 5606, "start": 66.44, "end": 75.68, "text": "Handy由开发者CJ Pace和Terry一起开发。后端用的是Rust。上层的设置界面则用React和TypeScript。", "tokens": [50884, 39, 11425, 23786, 18937, 28926, 12444, 34, 41, 430, 617, 12565, 51, 5318, 29567, 18937, 28926, 1543, 13547, 11957, 107, 9254, 24620, 49, 381, 1543, 5708, 9636, 224, 1546, 7422, 122, 34719, 21185, 8833, 2437, 247, 9254, 8524, 578, 12565, 22464, 494, 14237, 1543, 51346], "temperature": 0, "avg_logprob": -0.1285893808711659, "compression_ratio": 1.0845070422535212, "no_speech_prob": 1.6128581109553508e-11}, {"id": 4, "seek": 7568, "start": 75.68, "end": 85.4, "text": "Transcribe CPP负负责运行Whisper系列模型。格式是GGM和GGY。Transcribe RS运行的是ParaKit。", "tokens": [50365, 33339, 8056, 383, 17755, 18464, 253, 18464, 253, 18464, 96, 3316, 238, 8082, 2471, 271, 610, 25368, 43338, 41908, 39823, 1543, 30921, 27584, 1541, 27561, 44, 12565, 27561, 56, 1543, 33339, 8056, 25855, 3316, 238, 8082, 24620, 47, 2419, 45626, 1543, 50851], "temperature": 0, "avg_logprob": -0.15612416700883344, "compression_ratio": 1.0833333333333333, "no_speech_prob": 2.065555539110253e-11}, {"id": 5, "seek": 7568, "start": 86.10000000000001, "end": 91.62, "text": "CPAL负责跨平台音频输入。RDV负责全局键盘快捷键。", "tokens": [50886, 20049, 3427, 18464, 253, 18464, 96, 6563, 101, 16716, 15433, 18034, 39752, 9830, 241, 14028, 1543, 49, 35, 53, 18464, 253, 18464, 96, 11319, 34703, 23049, 106, 5419, 246, 10251, 26075, 115, 23049, 106, 1543, 51162], "temperature": 0, "avg_logprob": -0.15612416700883344, "compression_ratio": 1.0833333333333333, "no_speech_prob": 2.065555539110253e-11}, {"id": 6, "seek": 7568, "start": 92.5, "end": 98.0, "text": "它还用到一个叫Roboto的Core。提供实时音频流和重采样。", "tokens": [51206, 11284, 14852, 9254, 4511, 20182, 19855, 26332, 6738, 1546, 34, 418, 1543, 20949, 3254, 249, 24726, 15729, 18034, 39752, 27854, 12565, 12624, 5873, 229, 14496, 1543, 51481], "temperature": 0, "avg_logprob": -0.15612416700883344, "compression_ratio": 1.0833333333333333, "no_speech_prob": 2.065555539110253e-11}, {"id": 7, "seek": 9800, "start": 98.0, "end": 105.42, "text": "这个组合让它用起来特别简单,也特别清亮。连在我的MacBook上都很轻松。", "tokens": [50365, 15368, 10115, 226, 14245, 33650, 11284, 9254, 9147, 6912, 17682, 18453, 11249, 222, 47446, 11, 6404, 17682, 18453, 21784, 22991, 1543, 3316, 252, 3581, 14200, 34355, 19203, 5708, 7182, 4563, 17819, 119, 36897, 1543, 50736], "temperature": 0, "avg_logprob": -0.13048562770936548, "compression_ratio": 1.094890510948905, "no_speech_prob": 1.8870223092437577e-11}, {"id": 8, "seek": 9800, "start": 107.1, "end": 109.58, "text": "从用户角度看,整个流程也很简单。", "tokens": [50820, 35630, 9254, 1486, 115, 29389, 13127, 4200, 11, 27662, 7549, 27854, 29649, 6404, 4563, 11249, 222, 47446, 1543, 50944], "temperature": 0, "avg_logprob": -0.13048562770936548, "compression_ratio": 1.094890510948905, "no_speech_prob": 1.8870223092437577e-11}, {"id": 9, "seek": 9800, "start": 110.34, "end": 113.46000000000001, "text": "你先设一个自定义快捷键,然后可以切缓开关。", "tokens": [50982, 2166, 10108, 7422, 122, 20182, 9722, 12088, 2930, 231, 10251, 26075, 115, 23049, 106, 11, 26636, 6723, 23632, 38109, 241, 18937, 28053, 1543, 51138], "temperature": 0, "avg_logprob": -0.13048562770936548, "compression_ratio": 1.094890510948905, "no_speech_prob": 1.8870223092437577e-11}, {"id": 10, "seek": 9800, "start": 114.34, "end": 116.36, "text": "或者长按来实现Push to Talk。", "tokens": [51182, 31148, 32271, 26613, 6912, 24726, 20204, 47, 1498, 281, 8780, 1543, 51283], "temperature": 0, "avg_logprob": -0.13048562770936548, "compression_ratio": 1.094890510948905, "no_speech_prob": 1.8870223092437577e-11}, {"id": 11, "seek": 9800, "start": 117.36, "end": 120.32, "text": "按住的时候,Colero VAD会在后台悄悄过滤掉静音。", "tokens": [51333, 26613, 21632, 49873, 11, 34, 401, 2032, 691, 6112, 12949, 3581, 13547, 15433, 14696, 226, 14696, 226, 16866, 15868, 97, 29327, 5363, 247, 18034, 1543, 51481], "temperature": 0, "avg_logprob": -0.13048562770936548, "compression_ratio": 1.094890510948905, "no_speech_prob": 1.8870223092437577e-11}, {"id": 12, "seek": 12032, "start": 120.32, "end": 123.72, "text": "这样就不会白白消耗算力去转写整段空白。", "tokens": [50365, 21209, 3111, 47928, 13558, 13558, 28837, 4450, 245, 19497, 13486, 6734, 17819, 105, 5676, 247, 27662, 28427, 23322, 13558, 1543, 50535], "temperature": 0, "avg_logprob": -0.08909798860549926, "compression_ratio": 1.3040752351097178, "no_speech_prob": 1.4412372337235801e-11}, {"id": 13, "seek": 12032, "start": 124.55999999999999, "end": 128.16, "text": "松开后,Handy会把音频发给你选定的模型。", "tokens": [50577, 36897, 18937, 13547, 11, 39, 11425, 12949, 16075, 18034, 39752, 28926, 23197, 2166, 2215, 231, 12088, 1546, 41908, 39823, 1543, 50757], "temperature": 0, "avg_logprob": -0.08909798860549926, "compression_ratio": 1.3040752351097178, "no_speech_prob": 1.4412372337235801e-11}, {"id": 14, "seek": 12032, "start": 129.2, "end": 133.64, "text": "再把转写结果直接粘贴到当时当前聚焦的输入框里。", "tokens": [50809, 8623, 16075, 17819, 105, 5676, 247, 45641, 9319, 43297, 18393, 246, 18464, 112, 4511, 16233, 15729, 16233, 8945, 8171, 248, 4971, 99, 1546, 9830, 241, 14028, 21416, 228, 15759, 1543, 51031], "temperature": 0, "avg_logprob": -0.08909798860549926, "compression_ratio": 1.3040752351097178, "no_speech_prob": 1.4412372337235801e-11}, {"id": 15, "seek": 12032, "start": 134.66, "end": 136.45999999999998, "text": "Handy并不是这方面独有的。", "tokens": [51082, 39, 11425, 3509, 114, 7296, 5562, 39604, 18637, 105, 2412, 1546, 1543, 51172], "temperature": 0, "avg_logprob": -0.08909798860549926, "compression_ratio": 1.3040752351097178, "no_speech_prob": 1.4412372337235801e-11}, {"id": 16, "seek": 12032, "start": 137.4, "end": 141.45999999999998, "text": "市面上还有不少语音转写应用也差不多是这种工作方式。", "tokens": [51219, 27261, 8833, 5708, 35091, 1960, 15686, 5233, 255, 18034, 17819, 105, 5676, 247, 44297, 9254, 6404, 37876, 1541, 5562, 39810, 41315, 9249, 27584, 1543, 51422], "temperature": 0, "avg_logprob": -0.08909798860549926, "compression_ratio": 1.3040752351097178, "no_speech_prob": 1.4412372337235801e-11}, {"id": 17, "seek": 12032, "start": 142.16, "end": 145.68, "text": "不过我最喜欢的是你可以自己选喜欢的模型。", "tokens": [51457, 1960, 16866, 1654, 8661, 41606, 24620, 42766, 17645, 2215, 231, 41606, 1546, 41908, 39823, 1543, 51633], "temperature": 0, "avg_logprob": -0.08909798860549926, "compression_ratio": 1.3040752351097178, "no_speech_prob": 1.4412372337235801e-11}, {"id": 18, "seek": 12032, "start": 146.34, "end": 149.9, "text": "模型列表里的每个模型都有速度和准确率的指标。", "tokens": [51666, 41908, 39823, 43338, 17571, 15759, 1546, 23664, 7549, 41908, 39823, 48121, 31217, 13127, 12565, 6336, 228, 38114, 106, 44866, 1546, 25922, 162, 3921, 1543, 51844], "temperature": 0, "avg_logprob": -0.08909798860549926, "compression_ratio": 1.3040752351097178, "no_speech_prob": 1.4412372337235801e-11}, {"id": 19, "seek": 15032, "start": 150.32, "end": 153.44, "text": "这有点像在Mario Kart里选自己最喜欢的角色。", "tokens": [50365, 5562, 2412, 12579, 12760, 3581, 44, 4912, 27365, 15759, 2215, 231, 17645, 8661, 41606, 1546, 29389, 17673, 1543, 50521], "temperature": 0, "avg_logprob": -0.19431482873311856, "compression_ratio": 1.1961325966850829, "no_speech_prob": 1.1679083915250299e-11}, {"id": 20, "seek": 15032, "start": 154.16, "end": 156.44, "text": "你可以决定哪项指标对你来说最重要。", "tokens": [50557, 42766, 5676, 111, 12088, 17028, 10178, 117, 25922, 162, 3921, 8713, 2166, 6912, 8090, 8661, 24928, 1543, 50671], "temperature": 0, "avg_logprob": -0.19431482873311856, "compression_ratio": 1.1961325966850829, "no_speech_prob": 1.1679083915250299e-11}, {"id": 21, "seek": 15032, "start": 156.94, "end": 162.18, "text": "对我来说,我发现用ParaKit Unify 0,6BP模型就很好用。", "tokens": [50696, 8713, 1654, 6912, 8090, 11, 1654, 28926, 20204, 9254, 47, 2419, 45626, 1156, 2505, 1958, 11, 21, 33, 47, 41908, 39823, 3111, 23801, 9254, 1543, 50958], "temperature": 0, "avg_logprob": -0.19431482873311856, "compression_ratio": 1.1961325966850829, "no_speech_prob": 1.1679083915250299e-11}, {"id": 22, "seek": 15032, "start": 162.84, "end": 166.98, "text": "因为ParaKit是一个CPU1模型,而且按他们自己的数据。", "tokens": [50991, 34627, 47, 2419, 45626, 1541, 20182, 34, 8115, 16, 41908, 39823, 11, 22942, 26613, 47911, 17645, 1546, 33188, 26075, 106, 1543, 51198], "temperature": 0, "avg_logprob": -0.19431482873311856, "compression_ratio": 1.1961325966850829, "no_speech_prob": 1.1679083915250299e-11}, {"id": 23, "seek": 15032, "start": 167.7, "end": 170.64, "text": "它在终端爱武芯片上的运行速度大约是5倍10时。", "tokens": [51234, 11284, 3581, 10115, 230, 11957, 107, 27324, 38194, 13778, 107, 16668, 5708, 1546, 3316, 238, 8082, 31217, 13127, 3582, 16853, 99, 1541, 20, 35477, 3279, 15729, 1543, 51381], "temperature": 0, "avg_logprob": -0.19431482873311856, "compression_ratio": 1.1961325966850829, "no_speech_prob": 1.1679083915250299e-11}, {"id": 24, "seek": 15032, "start": 171.2, "end": 173.29999999999998, "text": "我的MR Max远远超过这个Page。", "tokens": [51409, 14200, 21173, 7402, 3316, 250, 3316, 250, 19869, 16866, 15368, 47, 609, 1543, 51514], "temperature": 0, "avg_logprob": -0.19431482873311856, "compression_ratio": 1.1961325966850829, "no_speech_prob": 1.1679083915250299e-11}, {"id": 25, "seek": 15032, "start": 173.95999999999998, "end": 176.82, "text": "所以我把它放在后台跑的时候甚至都感觉不到。", "tokens": [51547, 7239, 1654, 42061, 12744, 3581, 13547, 15433, 32585, 49873, 17556, 20844, 7182, 9709, 24447, 35728, 1543, 51690], "temperature": 0, "avg_logprob": -0.19431482873311856, "compression_ratio": 1.1961325966850829, "no_speech_prob": 1.1679083915250299e-11}, {"id": 26, "seek": 15032, "start": 177.62, "end": 179.57999999999998, "text": "而且可选项真的很多。", "tokens": [51730, 22942, 4429, 2215, 231, 10178, 117, 8034, 20778, 1543, 51828], "temperature": 0, "avg_logprob": -0.19431482873311856, "compression_ratio": 1.1961325966850829, "no_speech_prob": 1.1679083915250299e-11}, {"id": 27, "seek": 17958, "start": 179.58, "end": 182.92000000000002, "text": "它们甚至还有专门针对某一种的模型。", "tokens": [50365, 11284, 9497, 17556, 20844, 35091, 940, 241, 8259, 101, 25153, 230, 8713, 17238, 238, 2257, 39810, 1546, 41908, 39823, 1543, 50532], "temperature": 0, "avg_logprob": -0.14345688138689314, "compression_ratio": 1.2114093959731544, "no_speech_prob": 1.4638429357560767e-11}, {"id": 28, "seek": 17958, "start": 183.88000000000002, "end": 187.20000000000002, "text": "比如我们有一些俄语版本,还有一些乌克兰语版本。", "tokens": [50580, 36757, 15003, 32241, 13824, 7792, 226, 5233, 255, 42096, 8802, 11, 14852, 32241, 13824, 2930, 234, 24881, 2347, 108, 5233, 255, 42096, 8802, 1543, 50746], "temperature": 0, "avg_logprob": -0.14345688138689314, "compression_ratio": 1.2114093959731544, "no_speech_prob": 1.4638429357560767e-11}, {"id": 29, "seek": 17958, "start": 188.0, "end": 190.24, "text": "我很喜欢嗨这句使命宣言里的坦诚。", "tokens": [50786, 46257, 41606, 6083, 101, 5562, 34592, 22982, 25236, 2415, 96, 12009, 15759, 1546, 14872, 99, 5233, 248, 1543, 50898], "temperature": 0, "avg_logprob": -0.14345688138689314, "compression_ratio": 1.2114093959731544, "no_speech_prob": 1.4638429357560767e-11}, {"id": 30, "seek": 17958, "start": 191.02, "end": 193.44, "text": "Handy不是想成为最好的语音转文字应用,", "tokens": [50937, 39, 11425, 7296, 7093, 11336, 13992, 8661, 20715, 5233, 255, 18034, 17819, 105, 17174, 22381, 44297, 9254, 11, 51058], "temperature": 0, "avg_logprob": -0.14345688138689314, "compression_ratio": 1.2114093959731544, "no_speech_prob": 1.4638429357560767e-11}, {"id": 31, "seek": 17958, "start": 194.12, "end": 196.28, "text": "而是想成为最容易二次开发的那个。", "tokens": [51092, 11070, 1541, 7093, 11336, 13992, 8661, 49212, 11217, 9487, 18937, 28926, 1546, 43083, 1543, 51200], "temperature": 0, "avg_logprob": -0.14345688138689314, "compression_ratio": 1.2114093959731544, "no_speech_prob": 1.4638429357560767e-11}, {"id": 32, "seek": 17958, "start": 197.06, "end": 203.34, "text": "而且你甚至可以把自己微调过的Whisper G Gmail模型放进App的Models文件夹里。", "tokens": [51239, 22942, 2166, 17556, 20844, 6723, 16075, 17645, 39152, 8897, 225, 16866, 1546, 2471, 271, 610, 460, 36732, 41908, 39823, 12744, 36700, 9132, 1546, 44, 378, 1625, 17174, 20485, 1787, 117, 15759, 1543, 51553], "temperature": 0, "avg_logprob": -0.14345688138689314, "compression_ratio": 1.2114093959731544, "no_speech_prob": 1.4638429357560767e-11}, {"id": 33, "seek": 20334, "start": 203.34, "end": 208.20000000000002, "text": "重启之后,它就会直接作为一个自定义选项出现。", "tokens": [50365, 12624, 2392, 107, 9574, 13547, 11, 11284, 3111, 12949, 43297, 11914, 13992, 20182, 9722, 12088, 2930, 231, 2215, 231, 10178, 117, 7781, 20204, 1543, 50608], "temperature": 0, "avg_logprob": -0.12548497382630694, "compression_ratio": 1.0304347826086957, "no_speech_prob": 1.7507646374315122e-11}, {"id": 34, "seek": 20334, "start": 209.18, "end": 214.34, "text": "我觉得,如果任何时候要我在Handy和Whisper Flow之间选,", "tokens": [50657, 45681, 11, 13119, 26443, 11906, 29111, 4275, 40528, 39, 11425, 12565, 2471, 271, 610, 32792, 9574, 31685, 2215, 231, 11, 50915], "temperature": 0, "avg_logprob": -0.12548497382630694, "compression_ratio": 1.0304347826086957, "no_speech_prob": 1.7507646374315122e-11}, {"id": 35, "seek": 20334, "start": 215.36, "end": 219.20000000000002, "text": "我会选Handy,最大的原因就是它可以离线运行。", "tokens": [50966, 1654, 12949, 2215, 231, 39, 11425, 11, 8661, 39156, 42577, 5620, 11284, 6723, 18614, 119, 16853, 123, 3316, 238, 8082, 1543, 51158], "temperature": 0, "avg_logprob": -0.12548497382630694, "compression_ratio": 1.0304347826086957, "no_speech_prob": 1.7507646374315122e-11}, {"id": 36, "seek": 20334, "start": 220.16, "end": 222.98000000000002, "text": "你不需要注册账号,也没有订阅费。", "tokens": [51206, 2166, 1960, 35748, 26432, 5676, 234, 18464, 99, 26987, 11, 6404, 17944, 7422, 95, 10034, 227, 18464, 117, 1543, 51347], "temperature": 0, "avg_logprob": -0.12548497382630694, "compression_ratio": 1.0304347826086957, "no_speech_prob": 1.7507646374315122e-11}, {"id": 37, "seek": 22298, "start": 222.98, "end": 228.54, "text": "而且我很确定,所有我的语音输入音频样本都会留在我的机器上,", "tokens": [50365, 22942, 46257, 38114, 106, 12088, 11, 39300, 14200, 5233, 255, 18034, 9830, 241, 14028, 18034, 39752, 14496, 8802, 7182, 12949, 24456, 3581, 14200, 37960, 34386, 5708, 11, 50643], "temperature": 0, "avg_logprob": -0.11425744374593098, "compression_ratio": 1.213166144200627, "no_speech_prob": 1.4954067498185175e-11}, {"id": 38, "seek": 22298, "start": 229.2, "end": 231.22, "text": "不会被偷偷拿去训练别的语音模型。", "tokens": [50676, 47928, 23238, 7437, 115, 7437, 115, 24351, 6734, 7422, 255, 10115, 225, 18453, 1546, 5233, 255, 18034, 41908, 39823, 1543, 50777], "temperature": 0, "avg_logprob": -0.11425744374593098, "compression_ratio": 1.213166144200627, "no_speech_prob": 1.4954067498185175e-11}, {"id": 39, "seek": 22298, "start": 231.67999999999998, "end": 232.89999999999998, "text": "所以我真的很高兴。", "tokens": [50800, 7239, 1654, 38317, 12979, 2347, 112, 1543, 50861], "temperature": 0, "avg_logprob": -0.11425744374593098, "compression_ratio": 1.213166144200627, "no_speech_prob": 1.4954067498185175e-11}, {"id": 40, "seek": 22298, "start": 233.64, "end": 236.45999999999998, "text": "现在有了这样一个MIT许可的语音输入工具。", "tokens": [50898, 25040, 2412, 2289, 21209, 20182, 44, 3927, 7422, 116, 4429, 1546, 5233, 255, 18034, 9830, 241, 14028, 23323, 39806, 1543, 51039], "temperature": 0, "avg_logprob": -0.11425744374593098, "compression_ratio": 1.213166144200627, "no_speech_prob": 1.4954067498185175e-11}, {"id": 41, "seek": 22298, "start": 237.17999999999998, "end": 238.44, "text": "听起来都很棒,不过,", "tokens": [51075, 31022, 9147, 6912, 7182, 4563, 23514, 11, 1960, 16866, 11, 51138], "temperature": 0, "avg_logprob": -0.11425744374593098, "compression_ratio": 1.213166144200627, "no_speech_prob": 1.4954067498185175e-11}, {"id": 42, "seek": 22298, "start": 239.23999999999998, "end": 242.73999999999998, "text": "我们就来实际测试一下Pandy,看看它表现如何。", "tokens": [51178, 15003, 3111, 6912, 24726, 8842, 227, 11038, 233, 5233, 243, 8861, 47, 11425, 11, 22884, 11284, 17571, 20204, 43526, 1543, 51353], "temperature": 0, "avg_logprob": -0.11425744374593098, "compression_ratio": 1.213166144200627, "no_speech_prob": 1.4954067498185175e-11}, {"id": 43, "seek": 22298, "start": 243.62, "end": 248.85999999999999, "text": "我会把Handy的表现和Whisper Flow以及Google自家的转写服务做个对比。", "tokens": [51397, 1654, 12949, 16075, 39, 11425, 1546, 17571, 20204, 12565, 2471, 271, 610, 32792, 40282, 12104, 3127, 9722, 5155, 1546, 17819, 105, 5676, 247, 27408, 5087, 94, 10907, 7549, 8713, 11706, 1543, 51659], "temperature": 0, "avg_logprob": -0.11425744374593098, "compression_ratio": 1.213166144200627, "no_speech_prob": 1.4954067498185175e-11}, {"id": 44, "seek": 25298, "start": 252.98, "end": 260.94, "text": "后者也可以在Google Docs里用。", "tokens": [50365, 13547, 12444, 47935, 3581, 12104, 3127, 16024, 82, 15759, 9254, 1543, 50763], "temperature": 0, "avg_logprob": -0.11494940519332886, "compression_ratio": 0.9850746268656716, "no_speech_prob": 1.9903250922381766e-11}, {"id": 45, "seek": 25298, "start": 266.12, "end": 270.12, "text": "哦,昨天我去买了点菜,回家路上突然想起来。", "tokens": [51022, 14659, 11, 47404, 6135, 1654, 6734, 2930, 108, 2289, 12579, 33025, 11, 8350, 5155, 24658, 5708, 40859, 5823, 7093, 9147, 6912, 1543, 51222], "temperature": 0, "avg_logprob": -0.11494940519332886, "compression_ratio": 0.9850746268656716, "no_speech_prob": 1.9903250922381766e-11}, {"id": 46, "seek": 25298, "start": 275.12, "end": 282.06, "text": "我得回头把我的SQLite数据库接到一个运行在我自定义GPU集群上的Cobot服务上。", "tokens": [51472, 1654, 5916, 8350, 39862, 16075, 14200, 39934, 642, 33188, 26075, 106, 6346, 241, 14468, 4511, 20182, 3316, 238, 8082, 3581, 1654, 9722, 12088, 2930, 231, 38, 8115, 26020, 7248, 97, 5708, 1546, 34, 996, 310, 27408, 5087, 94, 5708, 1543, 51819], "temperature": 0, "avg_logprob": -0.11494940519332886, "compression_ratio": 0.9850746268656716, "no_speech_prob": 1.9903250922381766e-11}, {"id": 47, "seek": 28298, "start": 282.98, "end": 296.64000000000004, "text": "我能想到的一个办法是用一个叫Grog的LLM。", "tokens": [50365, 1654, 8225, 7093, 4511, 1546, 20182, 39453, 11148, 1541, 9254, 20182, 19855, 38, 6675, 1546, 24010, 44, 1543, 51048], "temperature": 0, "avg_logprob": -0.1713928956251878, "compression_ratio": 1.0236686390532543, "no_speech_prob": 2.1539211386700785e-11}, {"id": 48, "seek": 28298, "start": 298.54, "end": 301.22, "text": "另外还从CheckGPT得到了一些帮助。", "tokens": [51143, 26202, 14852, 35630, 47769, 38, 47, 51, 5916, 21381, 38515, 4845, 106, 37618, 1543, 51277], "temperature": 0, "avg_logprob": -0.1713928956251878, "compression_ratio": 1.0236686390532543, "no_speech_prob": 2.1539211386700785e-11}, {"id": 49, "seek": 28298, "start": 301.86, "end": 308.06, "text": "最后,我们做出了一个能运行的应用,但过程真的很折腾。", "tokens": [51309, 8661, 13547, 11, 15003, 10907, 7781, 2289, 20182, 8225, 3316, 238, 8082, 1546, 44297, 9254, 11, 8395, 16866, 29649, 38317, 7235, 246, 21184, 122, 1543, 51619], "temperature": 0, "avg_logprob": -0.1713928956251878, "compression_ratio": 1.0236686390532543, "no_speech_prob": 2.1539211386700785e-11}, {"id": 50, "seek": 30806, "start": 308.06, "end": 312.14, "text": "我们把这两部分分开来说。", "tokens": [50365, 15003, 16075, 5562, 36257, 32174, 6627, 18937, 6912, 8090, 1543, 50569], "temperature": 0, "avg_logprob": -0.15856294508104202, "compression_ratio": 1.0601092896174864, "no_speech_prob": 1.809806818298121e-11}, {"id": 51, "seek": 30806, "start": 312.82, "end": 318.64, "text": "你看,这里的Google语音输入服务要差很多,它不会加标点。", "tokens": [50603, 16529, 11, 35102, 1546, 12104, 3127, 5233, 255, 18034, 9830, 241, 14028, 27408, 5087, 94, 4275, 21679, 20778, 11, 11284, 47928, 9990, 162, 3921, 12579, 1543, 50894], "temperature": 0, "avg_logprob": -0.15856294508104202, "compression_ratio": 1.0601092896174864, "no_speech_prob": 1.809806818298121e-11}, {"id": 52, "seek": 30806, "start": 320.32, "end": 324.28, "text": "也不知道怎么分句,它把Cobot识别错了。", "tokens": [50978, 6404, 17572, 15282, 6627, 34592, 11, 11284, 16075, 34, 996, 310, 5233, 228, 18453, 29900, 2289, 1543, 51176], "temperature": 0, "avg_logprob": -0.15856294508104202, "compression_ratio": 1.0601092896174864, "no_speech_prob": 1.809806818298121e-11}, {"id": 53, "seek": 30806, "start": 325.66, "end": 327.78, "text": "SQLite数据库也识别错了。", "tokens": [51245, 39934, 642, 33188, 26075, 106, 6346, 241, 6404, 5233, 228, 18453, 29900, 2289, 1543, 51351], "temperature": 0, "avg_logprob": -0.15856294508104202, "compression_ratio": 1.0601092896174864, "no_speech_prob": 1.809806818298121e-11}, {"id": 54, "seek": 32778, "start": 327.78, "end": 331.44, "text": "甚至连LLM这个词都没认出来。", "tokens": [50365, 17556, 20844, 3316, 252, 24010, 44, 15368, 5233, 235, 7182, 10062, 7422, 97, 44561, 1543, 50548], "temperature": 0, "avg_logprob": -0.18794196165060695, "compression_ratio": 1.0650887573964498, "no_speech_prob": 1.6151044043843932e-11}, {"id": 55, "seek": 32778, "start": 332.08, "end": 333.44, "text": "它以为Grog是Rock。", "tokens": [50580, 11284, 3588, 13992, 38, 6675, 1541, 39061, 1543, 50648], "temperature": 0, "avg_logprob": -0.18794196165060695, "compression_ratio": 1.0650887573964498, "no_speech_prob": 1.6151044043843932e-11}, {"id": 56, "seek": 32778, "start": 335.2, "end": 339.11999999999995, "text": "这个我最喜欢,它还把ChatGPT认成了ChatGPT。", "tokens": [50736, 15368, 1654, 8661, 41606, 11, 11284, 14852, 16075, 41683, 38, 47, 51, 7422, 97, 11336, 2289, 41683, 38, 47, 51, 1543, 50932], "temperature": 0, "avg_logprob": -0.18794196165060695, "compression_ratio": 1.0650887573964498, "no_speech_prob": 1.6151044043843932e-11}, {"id": 57, "seek": 32778, "start": 340.79999999999995, "end": 341.94, "text": "ChatGPT不错。", "tokens": [51016, 41683, 38, 47, 51, 1960, 29900, 1543, 51073], "temperature": 0, "avg_logprob": -0.18794196165060695, "compression_ratio": 1.0650887573964498, "no_speech_prob": 1.6151044043843932e-11}, {"id": 58, "seek": 32778, "start": 342.65999999999997, "end": 345.7, "text": "嗯,所以你能明显看出来Handy好多了。", "tokens": [51109, 21206, 11, 7239, 2166, 8225, 11100, 1431, 122, 4200, 44561, 39, 11425, 2131, 6392, 2289, 1543, 51261], "temperature": 0, "avg_logprob": -0.18794196165060695, "compression_ratio": 1.0650887573964498, "no_speech_prob": 1.6151044043843932e-11}, {"id": 59, "seek": 34570, "start": 345.7, "end": 350.14, "text": "因为它有标点,还能识别一些特定术语。", "tokens": [50365, 34627, 11284, 2412, 162, 3921, 12579, 11, 14852, 8225, 5233, 228, 18453, 38515, 17682, 12088, 1474, 107, 5233, 255, 1543, 50587], "temperature": 0, "avg_logprob": -0.13285771335463925, "compression_ratio": 1.055, "no_speech_prob": 1.8650126581976068e-11}, {"id": 60, "seek": 34570, "start": 351.65999999999997, "end": 354.3, "text": "而且大多数技术术语都认对了。", "tokens": [50663, 22942, 3582, 6392, 33188, 32502, 1474, 107, 1474, 107, 5233, 255, 7182, 7422, 97, 8713, 2289, 1543, 50795], "temperature": 0, "avg_logprob": -0.13285771335463925, "compression_ratio": 1.055, "no_speech_prob": 1.8650126581976068e-11}, {"id": 61, "seek": 34570, "start": 355.44, "end": 358.64, "text": "现在我们也来把这个结果和Whisper Flow比一下。", "tokens": [50852, 25040, 15003, 6404, 6912, 16075, 15368, 45641, 9319, 12565, 2471, 271, 610, 32792, 11706, 8861, 1543, 51012], "temperature": 0, "avg_logprob": -0.13285771335463925, "compression_ratio": 1.055, "no_speech_prob": 1.8650126581976068e-11}, {"id": 62, "seek": 34570, "start": 362.38, "end": 366.2, "text": "我要尽量把同样的文本重复的尽可能接近。", "tokens": [51199, 32186, 1530, 121, 26748, 16075, 13089, 14496, 1546, 17174, 8802, 12624, 1787, 235, 1546, 1530, 121, 16657, 14468, 17463, 1543, 51390], "temperature": 0, "avg_logprob": -0.13285771335463925, "compression_ratio": 1.055, "no_speech_prob": 1.8650126581976068e-11}, {"id": 63, "seek": 36620, "start": 366.2, "end": 369.97999999999996, "text": "公平地测试一下。", "tokens": [50365, 13545, 16716, 10928, 11038, 233, 5233, 243, 8861, 1543, 50554], "temperature": 0, "avg_logprob": -0.13561457508015182, "compression_ratio": 1.0952380952380953, "no_speech_prob": 1.670154639588084e-11}, {"id": 64, "seek": 36620, "start": 371.08, "end": 373.14, "text": "嗯,昨天我去买了些杂货。", "tokens": [50609, 21206, 11, 47404, 6135, 1654, 6734, 2930, 108, 2289, 13824, 4422, 224, 18464, 100, 1543, 50712], "temperature": 0, "avg_logprob": -0.13561457508015182, "compression_ratio": 1.0952380952380953, "no_speech_prob": 1.670154639588084e-11}, {"id": 65, "seek": 36620, "start": 373.84, "end": 374.76, "text": "回来的路上。", "tokens": [50747, 8350, 6912, 1546, 24658, 5708, 1543, 50793], "temperature": 0, "avg_logprob": -0.13561457508015182, "compression_ratio": 1.0952380952380953, "no_speech_prob": 1.670154639588084e-11}, {"id": 66, "seek": 36620, "start": 376.18, "end": 378.24, "text": "我突然想起我得回家。", "tokens": [50864, 1654, 40859, 5823, 7093, 9147, 1654, 5916, 8350, 5155, 1543, 50967], "temperature": 0, "avg_logprob": -0.13561457508015182, "compression_ratio": 1.0952380952380953, "no_speech_prob": 1.670154639588084e-11}, {"id": 67, "seek": 36620, "start": 379.58, "end": 385.52, "text": "把我的SQLite数据库接到一个运行在我自定义GPU集群上的Cobot服务上。", "tokens": [51034, 16075, 14200, 39934, 642, 33188, 26075, 106, 6346, 241, 14468, 4511, 20182, 3316, 238, 8082, 3581, 1654, 9722, 12088, 2930, 231, 38, 8115, 26020, 7248, 97, 5708, 1546, 34, 996, 310, 27408, 5087, 94, 5708, 1543, 51331], "temperature": 0, "avg_logprob": -0.13561457508015182, "compression_ratio": 1.0952380952380953, "no_speech_prob": 1.670154639588084e-11}, {"id": 68, "seek": 36620, "start": 390.38, "end": 393.82, "text": "我其实不知道该怎么做,所以只好找人帮忙。", "tokens": [51574, 1654, 41646, 17572, 44646, 15282, 10907, 11, 7239, 14003, 2131, 25085, 4035, 4845, 106, 48839, 1543, 51746], "temperature": 0, "avg_logprob": -0.13561457508015182, "compression_ratio": 1.0952380952380953, "no_speech_prob": 1.670154639588084e-11}, {"id": 69, "seek": 39382, "start": 393.82, "end": 397.09999999999997, "text": "我能想到的唯一能帮上忙的工具。", "tokens": [50365, 1654, 8225, 7093, 4511, 1546, 17198, 107, 2257, 8225, 4845, 106, 5708, 48839, 1546, 23323, 39806, 1543, 50529], "temperature": 0, "avg_logprob": -0.12290783586173222, "compression_ratio": 1.1829652996845426, "no_speech_prob": 1.376922794532609e-11}, {"id": 70, "seek": 39382, "start": 397.78, "end": 399.44, "text": "是一个叫Grook的LLM。", "tokens": [50563, 1541, 20182, 19855, 38, 340, 453, 1546, 24010, 44, 1543, 50646], "temperature": 0, "avg_logprob": -0.12290783586173222, "compression_ratio": 1.1829652996845426, "no_speech_prob": 1.376922794532609e-11}, {"id": 71, "seek": 39382, "start": 400.34, "end": 403.62, "text": "后来,我也从ChatGPT那里得到了一点帮助。", "tokens": [50691, 13547, 6912, 11, 27761, 35630, 41683, 38, 47, 51, 4184, 15759, 5916, 21381, 49042, 4845, 106, 37618, 1543, 50855], "temperature": 0, "avg_logprob": -0.12290783586173222, "compression_ratio": 1.1829652996845426, "no_speech_prob": 1.376922794532609e-11}, {"id": 72, "seek": 39382, "start": 404.65999999999997, "end": 408.62, "text": "最后,我们是把应用跑起来了,但过程真的很折腾。", "tokens": [50907, 8661, 13547, 11, 15003, 1541, 16075, 44297, 9254, 32585, 9147, 34912, 11, 8395, 16866, 29649, 38317, 7235, 246, 21184, 122, 1543, 51105], "temperature": 0, "avg_logprob": -0.12290783586173222, "compression_ratio": 1.1829652996845426, "no_speech_prob": 1.376922794532609e-11}, {"id": 73, "seek": 39382, "start": 409.65999999999997, "end": 413.42, "text": "好吧,你可以清楚地看到Whisper Flow做得最好。", "tokens": [51157, 40221, 11, 42766, 38433, 10928, 18032, 2471, 271, 610, 32792, 10907, 5916, 8661, 2131, 1543, 51345], "temperature": 0, "avg_logprob": -0.12290783586173222, "compression_ratio": 1.1829652996845426, "no_speech_prob": 1.376922794532609e-11}, {"id": 74, "seek": 39382, "start": 414.42, "end": 418.12, "text": "它把句子分得很准确,而且所有术语都没出错。", "tokens": [51395, 11284, 16075, 34592, 7626, 6627, 5916, 4563, 6336, 228, 38114, 106, 11, 22942, 39300, 1474, 107, 5233, 255, 7182, 10062, 7781, 29900, 1543, 51580], "temperature": 0, "avg_logprob": -0.12290783586173222, "compression_ratio": 1.1829652996845426, "no_speech_prob": 1.376922794532609e-11}, {"id": 75, "seek": 39382, "start": 419.18, "end": 423.42, "text": "毕竟这是商业产品,你会以为它会比开源工具更好。", "tokens": [51633, 7256, 243, 11957, 253, 27455, 45581, 940, 248, 1369, 100, 30246, 11, 2166, 12949, 3588, 13992, 11284, 12949, 11706, 18937, 47402, 23323, 39806, 19002, 2131, 1543, 51845], "temperature": 0, "avg_logprob": -0.12290783586173222, "compression_ratio": 1.1829652996845426, "no_speech_prob": 1.376922794532609e-11}, {"id": 76, "seek": 42382, "start": 423.82, "end": 429.4, "text": "但说实话,这个结果和Whisper Flow之间并没有那么大差别。", "tokens": [50365, 8395, 8090, 24726, 21596, 11, 15368, 45641, 9319, 12565, 2471, 271, 610, 32792, 9574, 31685, 3509, 114, 17944, 35693, 3582, 21679, 18453, 1543, 50644], "temperature": 0, "avg_logprob": -0.18442789713541666, "compression_ratio": 0.9337349397590361, "no_speech_prob": 1.5088708060773115e-11}, {"id": 77, "seek": 42382, "start": 430.34, "end": 434.52, "text": "所以,我还是更喜欢用Handy,因为它免费、私密。", "tokens": [50691, 7239, 11, 1654, 45726, 19002, 41606, 9254, 39, 11425, 11, 34627, 11284, 2347, 235, 18464, 117, 1231, 20083, 47838, 1543, 50900], "temperature": 0, "avg_logprob": -0.18442789713541666, "compression_ratio": 0.9337349397590361, "no_speech_prob": 1.5088708060773115e-11}, {"id": 78, "seek": 42382, "start": 435.48, "end": 436.44, "text": "而且是开源的。", "tokens": [50948, 22942, 1541, 18937, 47402, 1546, 1543, 50996], "temperature": 0, "avg_logprob": -0.18442789713541666, "compression_ratio": 0.9337349397590361, "no_speech_prob": 1.5088708060773115e-11}, {"id": 79, "seek": 43644, "start": 436.44, "end": 439.94, "text": "不过,Whisper Flow在这里确实做得很不错。", "tokens": [50365, 1960, 16866, 11, 2471, 271, 610, 32792, 3581, 35102, 38114, 106, 24726, 10907, 5916, 4563, 1960, 29900, 1543, 50540], "temperature": 0, "avg_logprob": -0.12326780954996745, "compression_ratio": 1.1148936170212767, "no_speech_prob": 1.4102624450174073e-11}, {"id": 80, "seek": 43644, "start": 440.98, "end": 444.32, "text": "所以总的来说,它还是一个很靠谱的服务。", "tokens": [50592, 7239, 33440, 1546, 6912, 8090, 11, 11284, 45726, 20182, 4563, 5363, 254, 8897, 109, 1546, 27408, 5087, 94, 1543, 50759], "temperature": 0, "avg_logprob": -0.12326780954996745, "compression_ratio": 1.1148936170212767, "no_speech_prob": 1.4102624450174073e-11}, {"id": 81, "seek": 43644, "start": 445.46, "end": 447.5, "text": "好了,各位,概括来说就是这样。", "tokens": [50816, 12621, 11, 22073, 11, 26181, 40322, 6912, 8090, 5620, 21209, 1543, 50918], "temperature": 0, "avg_logprob": -0.12326780954996745, "compression_ratio": 1.1148936170212767, "no_speech_prob": 1.4102624450174073e-11}, {"id": 82, "seek": 43644, "start": 447.96, "end": 449.24, "text": "这个功能确实很实用。", "tokens": [50941, 15368, 32311, 8225, 38114, 106, 24726, 4563, 24726, 9254, 1543, 51005], "temperature": 0, "avg_logprob": -0.12326780954996745, "compression_ratio": 1.1148936170212767, "no_speech_prob": 1.4102624450174073e-11}, {"id": 83, "seek": 43644, "start": 449.94, "end": 451.24, "text": "我刚才真不是开玩笑。", "tokens": [51040, 1654, 49160, 18888, 6303, 7296, 18937, 19912, 11600, 1543, 51105], "temperature": 0, "avg_logprob": -0.12326780954996745, "compression_ratio": 1.1148936170212767, "no_speech_prob": 1.4102624450174073e-11}, {"id": 84, "seek": 43644, "start": 452.02, "end": 454.68, "text": "说它已经成了我最喜欢的语音听写工具。", "tokens": [51144, 8090, 11284, 49161, 11336, 2289, 1654, 8661, 41606, 1546, 5233, 255, 18034, 31022, 5676, 247, 23323, 39806, 1543, 51277], "temperature": 0, "avg_logprob": -0.12326780954996745, "compression_ratio": 1.1148936170212767, "no_speech_prob": 1.4102624450174073e-11}, {"id": 85, "seek": 45468, "start": 454.68, "end": 457.84000000000003, "text": "我甚至有点担心,要是我用得太频繁。", "tokens": [50365, 1654, 17556, 20844, 2412, 12579, 6852, 227, 7945, 11, 4275, 42614, 9254, 5916, 9455, 39752, 23141, 223, 1543, 50523], "temperature": 0, "avg_logprob": -0.1628261843035298, "compression_ratio": 1.1954022988505748, "no_speech_prob": 1.4379071718029213e-11}, {"id": 86, "seek": 45468, "start": 458.6, "end": 460.96, "text": "自己可能都会忘了怎么用键盘打字。", "tokens": [50561, 17645, 16657, 7182, 12949, 26677, 2289, 15282, 9254, 23049, 106, 5419, 246, 12467, 22381, 1543, 50679], "temperature": 0, "avg_logprob": -0.1628261843035298, "compression_ratio": 1.1954022988505748, "no_speech_prob": 1.4379071718029213e-11}, {"id": 87, "seek": 45468, "start": 461.7, "end": 463.72, "text": "不过,这只是我的个人看法和观察。", "tokens": [50716, 1960, 16866, 11, 5562, 36859, 14200, 7549, 4035, 4200, 11148, 12565, 42350, 47550, 1543, 50817], "temperature": 0, "avg_logprob": -0.1628261843035298, "compression_ratio": 1.1954022988505748, "no_speech_prob": 1.4379071718029213e-11}, {"id": 88, "seek": 45468, "start": 464.5, "end": 465.90000000000003, "text": "那你觉得Handy怎么样?", "tokens": [50856, 4184, 2166, 29992, 39, 11425, 48200, 30, 50926], "temperature": 0, "avg_logprob": -0.1628261843035298, "compression_ratio": 1.1954022988505748, "no_speech_prob": 1.4379071718029213e-11}, {"id": 89, "seek": 45468, "start": 466.46, "end": 467.3, "text": "你是过了吗?", "tokens": [50954, 32526, 16866, 2289, 14769, 30, 50996], "temperature": 0, "avg_logprob": -0.1628261843035298, "compression_ratio": 1.1954022988505748, "no_speech_prob": 1.4379071718029213e-11}, {"id": 90, "seek": 45468, "start": 467.8, "end": 468.5, "text": "你会用吗?", "tokens": [51021, 2166, 12949, 9254, 14769, 30, 51056], "temperature": 0, "avg_logprob": -0.1628261843035298, "compression_ratio": 1.1954022988505748, "no_speech_prob": 1.4379071718029213e-11}, {"id": 91, "seek": 45468, "start": 469.36, "end": 471.2, "text": "欢迎在下方评论区告诉我们。", "tokens": [51099, 28566, 17699, 3581, 4438, 9249, 5233, 226, 7422, 118, 9937, 118, 44816, 15003, 1543, 51191], "temperature": 0, "avg_logprob": -0.1628261843035298, "compression_ratio": 1.1954022988505748, "no_speech_prob": 1.4379071718029213e-11}, {"id": 92, "seek": 45468, "start": 471.76, "end": 474.88, "text": "各位,如果你喜欢这类技术解析,", "tokens": [51219, 22073, 11, 45669, 41606, 5562, 22113, 119, 32502, 1474, 107, 17278, 7360, 238, 11, 51375], "temperature": 0, "avg_logprob": -0.1628261843035298, "compression_ratio": 1.1954022988505748, "no_speech_prob": 1.4379071718029213e-11}, {"id": 93, "seek": 45468, "start": 475.2, "end": 477.2, "text": "也请在视频下方点个赞。", "tokens": [51391, 6404, 27908, 3581, 40656, 39752, 4438, 9249, 12579, 7549, 5266, 252, 1543, 51491], "temperature": 0, "avg_logprob": -0.1628261843035298, "compression_ratio": 1.1954022988505748, "no_speech_prob": 1.4379071718029213e-11}, {"id": 94, "seek": 47720, "start": 477.2, "end": 480.94, "text": "让我知道,别忘了订阅我们的频道。", "tokens": [50365, 33650, 33838, 11, 18453, 26677, 2289, 7422, 95, 10034, 227, 15003, 1546, 39752, 6025, 1543, 50552], "temperature": 0, "avg_logprob": -0.2620470637366885, "compression_ratio": 0.9727272727272728, "no_speech_prob": 1.495946769236589e-11}, {"id": 95, "seek": 47720, "start": 481.82, "end": 483.46, "text": "我是BetterStack的Andreas,", "tokens": [50596, 15914, 33, 27296, 4520, 501, 1546, 5289, 37868, 11, 50678], "temperature": 0, "avg_logprob": -0.2620470637366885, "compression_ratio": 0.9727272727272728, "no_speech_prob": 1.495946769236589e-11}, {"id": 96, "seek": 47720, "start": 483.78, "end": 485.02, "text": "我们下期视频再见。", "tokens": [50694, 15003, 4438, 16786, 40656, 39752, 44176, 1543, 50756], "temperature": 0, "avg_logprob": -0.2620470637366885, "compression_ratio": 0.9727272727272728, "no_speech_prob": 1.495946769236589e-11}, {"id": 97, "seek": 47720, "start": 486.12, "end": 486.56, "text": "音乐", "tokens": [50811, 18034, 44365, 50833], "temperature": 0, "avg_logprob": -0.2620470637366885, "compression_ratio": 0.9727272727272728, "no_speech_prob": 1.495946769236589e-11}], "language": "zh"}