#4384 TaskCfgVTT(is_cuda=True, uuid='1e2b9bb1a7', cache_folder='D:/win-pyvideotrans-v3.98-314/tmp/18908/1e2b9bb1a7', target_di

218.48* Posted at: 2 hours ago 👁9

语音识别阶段出错 [faster-whisper(本地)] 出错了,可能内存或显存不足
A process in the process pool was terminated abruptly while the future was running or pending.
Traceback (most recent call last):
File "videotrans\configure\_base.py", line 284, in _new_process
File "concurrent\futures\_base.py", line 458, in result
File "concurrent\futures\_base.py", line 403, in __get_result
concurrent.futures.process.BrokenProcessPool: A process in the process pool was terminated abruptly while the future was running or pending.

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
File "videotrans\task\job.py", line 105, in run
File "videotrans\task\trans_create.py", line 353, in recogn
File "videotrans\recognition\__init__.py", line 265, in run
File "videotrans\recognition\_base.py", line 143, in run
File "videotrans\recognition\_overall.py", line 33, in _exec
File "videotrans\recognition\_overall.py", line 105, in _faster
File "videotrans\configure\_base.py", line 294, in _new_process
RuntimeError: 出错了,可能内存或显存不足
A process in the process pool was terminated abruptly while the future was running or pending.
TaskCfgVTT(is_cuda=True, uuid='1e2b9bb1a7', cache_folder='D:/win-pyvideotrans-v3.98-314/tmp/18908/1e2b9bb1a7', target_dir='C:/课程/《STAAD.Pro澳洲规范钢结构设计教程》/_video_out/1、Creating Physical Members in STAAD.Pro-mp4', source_language='英语', source_language_code='en', source_sub='C:/课程/《STAAD.Pro澳洲规范钢结构设计教程》/_video_out/1、Creating Physical Members in STAAD.Pro-mp4/en.srt', source_wav='D:/win-pyvideotrans-v3.98-314/tmp/18908/1e2b9bb1a7/en.wav', source_wav_output='C:/课程/《STAAD.Pro澳洲规范钢结构设计教程》/_video_out/1、Creating Physical Members in STAAD.Pro-mp4/en.m4a', target_language='简体中文', target_language_code='zh-cn', target_sub='C:/课程/《STAAD.Pro澳洲规范钢结构设计教程》/_video_out/1、Creating Physical Members in STAAD.Pro-mp4/zh-cn.srt', target_wav='D:/win-pyvideotrans-v3.98-314/tmp/18908/1e2b9bb1a7/target.wav', target_wav_output='C:/课程/《STAAD.Pro澳洲规范钢结构设计教程》/_video_out/1、Creating Physical Members in STAAD.Pro-mp4/zh-cn.m4a', name='C:/课程/《STAAD.Pro澳洲规范钢结构设计教程》/1、Creating Physical Members in STAAD.Pro.mp4', noextname='1、Creating Physical Members in STAAD.Pro', basename='1、Creating Physical Members in STAAD.Pro.mp4', ext='mp4', dirname='C:/课程/《STAAD.Pro澳洲规范钢结构设计教程》', shound_del_name=None, translate_type=0, tts_type=0, volume='+0%', pitch='+0Hz', voice_rate='+0%', voice_role='No', voice_autorate=False, video_autorate=False, remove_silent_mid=False, align_sub_audio=True, detect_language='en', recogn_type=0, model_name='large-v3-turbo', shibie_audio=None, remove_noise=False, enable_diariz=False, nums_diariz=0, rephrase=0, fix_punc=False, subtitle_language=None, app_mode='tiqu', subtitles='', targetdir_mp4='C:/课程/《STAAD.Pro澳洲规范钢结构设计教程》/_video_out/1、Creating Physical Members in STAAD.Pro-mp4/1、Creating Physical Members in STAAD.Pro.mp4', novoice_mp4='D:/win-pyvideotrans-v3.98-314/tmp/18908/1e2b9bb1a7/novoice.mp4', is_separate=False, instrument=None, vocal=None, back_audio='', clear_cache=False, background_music=None, subtitle_type=0, only_out_mp4=False, recogn2pass=True, output_srt=2, copysrt_rawvideo=True)


=
system:Windows-10-10.0.26200-SP0
version:v3.98
frozen:True
language:zh
root_dir:D:/win-pyvideotrans-v3.98-314

1 Replies

这是显存或内存不足导致的进程崩溃。错误发生在整个视频同时发送给faster-whisper进行识别时(_overall.py_faster 函数),因为您开启了 二次识别 (recogn2pass=True) 并选择了 large-v3-turbo 模型,在CUDA加速下双任务并发加剧了显存消耗。

请按顺序尝试以下方案:

  1. 立即取消“二次识别”
    关闭主界面“配音渠道”行右侧的 二次识别 开关。该功能会在配音后对音频再次进行完整转录,同时处理一个视频时可能并行加载两个识别任务,导致需要2-3倍显存。
  2. 换用更小的语音识别模型
    将右侧模型从 large-v3-turbo 换成 smallmediumlarge-v3-turbo 模型仍需约6G以上显存,如果你的显卡显存有限(尤其是当显存小于8G并且系统已占用部分显存时)容易触发不足。
  3. 降低CUDA推理精度和搜索宽度
    进入 菜单 → 工具/选项 → 高级选项,找到 语音识别参数 组,进行如下修改:

    • CUDA数据类型:改为 int8(牺牲少量精度大幅降低显存)
    • 识别准确度 beam_size:改为 1
    • 识别准确度 best_of:改为 1
    • 启用上下文感知:取消勾选
  4. 升级到最新版本
    您目前使用的 v3.98 可能仍存在某些显存泄露风险,该问题在 v3.99 中已修复。请下载最新版本并覆盖:
    下载页面

如果上述步骤均无效,请提供 logs 目录下最新的日志文件最后30行内容,以便进一步分析。

请查阅相关文档:

Post Your Reply
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.

Related resource