#8273 语音识别阶段出错[faster-whisper(内置)] No human voice detected. Please confirm that human speech is present in the original file.

2a09**6f Posted at: 1 hour ago

语音识别阶段出错[faster-whisper(内置)] No human voice detected. Please confirm that human speech is present in the original file. [info.duration_after_vad=0.0s].
_kw={'beam_size': 5, 'best_of': 5, 'condition_on_previous_text': False, 'threshold': 0.45, 'no_speech_threshold': 0.6, 'temperature': 0.0, 'repetition_penalty': 1.0, 'compression_ratio_threshold': 2.4, 'initial_prompt': None}
info=TranscriptionInfo(language='ru', language_probability=1, duration=1284.922625, duration_after_vad=0.0, all_language_probs=None, transcription_options=TranscriptionOptions(beam_size=5, best_of=5, patience=1, length_penalty=1, repetition_penalty=1.0, no_repeat_ngram_size=0, log_prob_threshold=-1.0, no_speech_threshold=0.6, compression_ratio_threshold=2.4, condition_on_previous_text=False, prompt_reset_on_temperature=0.5, temperatures=[0.0], initial_prompt=None, prefix=None, suppress_blank=True, suppress_tokens=(1, 2, 7, 8, 9, 10, 14, 25, 26, 27, 28, 29, 31, 58, 59, 60, 61, 62, 63, 90, 91, 92, 93, 359, 503, 522
......
ion_after_vad=0.0, all_language_probs=None, transcription_options=TranscriptionOptions(beam_size=5, best_of=5, patience=1, length_penalty=1, repetition_penalty=1.0, no_repeat_ngram_size=0, log_prob_threshold=-1.0, no_speech_threshold=0.6, compression_ratio_threshold=2.4, condition_on_previous_text=False, prompt_reset_on_temperature=0.5, temperatures=[0.0], initial_prompt=None, prefix=None, suppress_blank=True, suppress_tokens=(1, 2, 7, 8, 9, 10, 14, 25, 26, 27, 28, 29, 31, 58, 59, 60, 61, 62, 63, 90, 91, 92, 93, 359, 503, 522, 542, 873, 893, 902, 918, 922, 931, 1350, 1853, 1982, 2460, 2627, 3246, 3253, 3268, 3536, 3846, 3961, 4183, 4667, 6585, 6647, 7273, 9061, 9383, 10428, 10929, 11938, 12033, 12331, 12562, 13793, 14157, 14635, 15265, 15618, 16553, 16604, 18362, 18956, 20075, 21675, 22520, 26130, 26161, 26435, 28279, 29464, 31650, 32302, 32470, 36865, 42863, 47425, 49870, 50254, 50258, 50359, 50360, 50361, 50362, 50363), without_timestamps=False, max_initial_timestamp=1.0, word_timestamps=True, prepend_punctuations='"\'“¿([{-', append_punctuations='"\'.。,,!!??::”)]}、', multilingual=False, max_new_tokens=None, clip_timestamps=[0.0], hallucination_silence_threshold=None, hotwords=''), vad_options=VadOptions(threshold=0.45, neg_threshold=None, min_speech_duration_ms=0, max_speech_duration_s=inf, min_silence_duration_ms=2000, speech_pad_ms=400))
trk=[SpeechToText]: uuid='4420687119', proxy_str=None, cfg=[TaskCfgSTT]
原始输入文件名: F:/Downloads/DCL/e/Mom Is Waiting For Dad In Doggy But Son Comes.mp4,
输出结果保存到文件夹: D:/win-pyvideotrans-v4.16-1/output/recogn,
临时文件夹: D:/win-pyvideotrans-v4.16-1/tmp/19012/4420687119
未启用CUDA加速
未选 降噪
语音识别:faster-whisper(内置), large-v3-turbo
发音语言:ru
未选 降噪, precent=1, hasend=False, should_recogn=True, should_trans=False, should_dubbing=False, should_hebing=False, out_format='srt', copysrt_rawvideo=False, spk_insert=True


=
system:Windows-10-10.0.26300-SP0
version:v4.16
frozen:True
language:zh_CN
root_dir:D:/win-pyvideotrans-v4.16-1

1 Replies

#1 DeepSeek reply at 1 hour ago

核心原因:这不是模型或网络问题,而是 VAD(语音活动检测)把整段音频判定为“无人声”——日志中 duration=1284.9s(音频成功解码,时长正常),但 duration_after_vad=0.0s(经 VAD 处理后剩余 0 秒语音),因此 faster-whisper 直接抛出 No human voice detected。

导致 VAD 全量过滤的常见原因:语音阈值过高、人声被背景音乐/噪声淹没、音量过低、视频实际无音轨或只有音乐。

请按顺序排查:

  1. 先确认源文件本身
    用播放器打开 Mom Is Waiting For Dad In Doggy But Son Comes.mp4,确认:有声音、音量正常、确有人类说话声。若只有背景音乐/纯环境音,本软件无法处理(需人声语音)。
  2. 降低 VAD 语音阈值(最有效)
    菜单 → 工具/选项 → 高级选项 → 语音识别参数

    • 语音阈值(VAD: Speech probability threshold):默认 0.45,改为 0.2~0.3
    • 非语音阈值(no speech threshold):默认 0.6,可改为 0.7~0.8
      保存后重新执行(务必勾选主界面 清理已生成)。
  3. 恢复被改乱的 VAD 参数
    你这台机器日志里 min_speech_duration_ms=0 / max_speech_duration_s=inf / min_silence_duration_ms=2000,与推荐默认值差异较大。建议在同一面板恢复为:

    • 最长语音持续(秒):5
    • 最短语音持续(毫秒):3000
    • 静音分割持续(毫秒):600
  4. 人声与背景音分离
    主界面点击 设置更多参数,勾选 分离人声背景声(背景音乐大、人声小的视频效果显著),或勾选 降噪。再次执行。
  5. 换渠道/换模型交叉验证
    若 VAD 仍是 0.0s:

    • 语音识别渠道改为 openai-whisper(内置) + large-v3,或 Qwen-ASR(内置)、Whisper.cpp(Win内置) 试一次;
    • 或把当前 large-v3-turbo 换成 large-v3,排除模型文件下载不完整(可删除 models/models--mobiuslabsgmbh--faster-whisper-large-v3-turbo 后重新下载)。
      若其他渠道能正常出字幕,说明是该模型/VAD 参数问题;若全部报同样错误,则基本可判定该音频中无人声或音轨异常。
  6. 文件路径规范化
    文件名含空格且层级较深,建议把视频复制到浅层目录并重命名为纯英文数字(如 D:/v/test01.mp4)后再试,排除 ffmpeg 解析异常。

补充:日志显示 未启用CUDA加速,CPU 跑 large-v3-turbo 处理 1284 秒音频会非常慢,但不影响此报错本身。若你希望先快速验证音频中是否有人声,可用左侧 语音转录 面板、发音语言选 自动检测 跑一次短片段测试。

请查阅相关文档:

若按上述操作后 VAD 仍为 duration_after_vad=0.0s,请提供 logs/ 目录下当日日志的最新 30 行(含音频提取相关 ffmpeg 输出)以便进一步判断是否为音轨异常。

Thinking...
Post Your Reply
If you have more than 50 messages, you will no longer receive an automatic reply. Please ask a new question!
Open source and free maintenance is not easy. If this project is helpful to you, please consider making a small donation to help the project continue to maintain and update.

Dondate via ko-fi.com

Related resource