pyVideoTrans

Appearance

v3.95 Changelog ​

  1. Added secondary recognition in the video translation feature. It can be seen by clicking Set More Parameters on the main interface.
  2. Added VAD Type Switch under Menu -> Tools -> Advanced Options. Default is ten-vad, which is faster and more accurate. Switch back to silero if issues are encountered.
  3. Enabled batch inference by default for faster-whisper to increase speed. Modify faster-whisper Batch Size to 1 in Advanced Options to switch to non-batch mode, which is slower but slightly improves accuracy.
  4. Optimized model auto-download: first test huggingface.co, then fall back to the domestic mirror hf-mirror.com if it fails.
  5. Added several custom options in Advanced Options to control hallucination issues.
  6. Added Transcript Alignment feature, simply implemented based on the faster-whisper local model and diff algorithm.
  7. Added supertonic dubbing channel, supporting English, Korean, Spanish, Portuguese, and French. It's compact and fast.
  8. Switched audio denoising to use modelscope.cn by default. Processing time will increase significantly, but denoising effect is better.
  9. Speaker diarization now supports: built-in Chinese/English model, pyannote3.0 model, Alibaba cam++ Chinese/English model, and reverb-diarization-v1 model. Can be switched in Advanced Options. Default uses the built-in Chinese/English model.
  10. During single video translation, when the subtitle editing interface pops up, you can open the subtitle folder, then manually or via third-party tools modify the generated subtitle file. Clicking Continue Without Saving in that interface will then use the modified subtitle file, making it convenient for manual adjustments.
  11. Added support for multi-GPU concurrency to speed up processing. Can be enabled by selecting Multi-GPU Mode in Advanced Options.
  12. Heavy tasks now use a multi-process approach to better utilize multi-core CPUs. Can be sped up by setting a higher number in Max Process Count in Advanced Options.
  13. Advanced Options -> Subtitle Translation: Added Include Full Subtitle checkbox. When selected and using an AI translation channel, the original subtitle will also be sent to the AI, potentially achieving better translation quality.
  14. Added command-line mode. For specific usage, see CLI Command Line.
  15. Optimized the default translation prompt.
  16. Added support for the open-source qwen-tts model.