#2566 TaskCfg(cache_folder='D:/win-videotrans-v3.81/tmp/14792/d1dae2bb1b', target_dir='20180617-進化班縮錄2-mp4', remove_noise=Fals

The error "No subs recognized. Check Audio/Lang." indicates that the faster-whisper model failed to detect any speech in your audio file. This is a common issue and can be resolved by checking several key points.

Here are the most likely causes and solutions, in order of probability:

1. Incorrect Source Language Selection: You have set source_language='Simplified Chinese' (zh-cn), but the audio in your video might be in a different language or dialect (e.g., Traditional Chinese, Japanese, or another). The model large-v3-turbo is good at multilingual detection, but specifying the wrong language can hinder it.

   • Solution: In the main interface, under "Translation Settings," change the "Source Language" to auto. Let the model detect the language automatically.

2. Audio File Issues: The video might have no audible speech, extremely low volume, or heavy background noise that masks the voice.

   • Solution:

     • Play the original video to confirm there is clear human speech.
     • In the main interface, check the box for "Remove Noise" before starting the task. This can help clean the audio for better recognition.
     • If the speech is very quiet, you may need to pre-process the audio volume externally.

3. Model or CUDA Issue: You are not using CUDA (cuda=False). While the CPU works, it's much slower. More importantly, the large-v3-turbo model might be too large for your system's memory if you have limited RAM.

   • Solution:

     • If you have an NVIDIA GPU, enable the "CUDA" checkbox on the main interface for significant speed and potential accuracy improvements.
     • If you have limited system RAM (e.g., less than 16GB), try switching to a smaller, faster model. In the "Speech Recognition" dropdown, select a model like medium or small instead of large-v3-turbo.

4. File Path/Name Issues: Your file path contains Chinese characters (.../2016-2017-2018自化高级晋级第一部（未流通）/...). While the software handles Unicode, extremely long paths or certain characters can sometimes cause issues with underlying libraries like ffmpeg.

   • Solution (Simplest Test): Copy the video file to a simple, short path on your D: drive, such as D:\test_video.mp4, and try processing it from there.

Recommended Action Plan:

1. First, change the Source Language to auto.
2. Enable the "Remove Noise" option.
3. If you have an NVIDIA GPU, enable CUDA.
4. If the problem persists, try using the medium model.
5. As a final test, move the video to a simple directory like D:\ and rename it to a short English name (e.g., test.mp4).

If none of these steps work, please check the logs folder inside your D:/win-videotrans-v3.81/ directory for more detailed error messages and provide them.

请查看更多参考文档:

• 语音识别渠道说明与介绍
• faster-whisper本地