search

RVC-Boss / GPT-SoVITS

1 min voice data can also be used to train a good TTS model! (few shot voice cloning)
MIT License
35.72k stars 4.07k forks source link

I can't use Faster Whisper/ASR because the "dependency files" downloading doesn't complete. #560

Open nekogecko2 opened 9 months ago

nekogecko2 commented 9 months ago

I'm using the windows prezip. When I select "faster whisper" and click the "batch asr" button to start the process, it never completes downloading one of the required files. It gets stuck on a random download percentage and never finishes. For example it's stuck on 81% after minutes passed and despite my fast internet speed.

image

I've decided to randomly try it with the "Chinese lang ASR" option too and the download gets stuck too.

RVC-Boss commented 8 months ago

I can't reproduce it. Maybe it's because the network. You can 1-wait or 2-retry: by double clicking the button to stop and start the asr process

MissingTwins commented 8 months ago

For the first deployment, you may refer to the jupyter notebook file in this project root folder, and test run with google colab before trying on your local environment.

Iamgoofball commented 8 months ago

I'm having the exact same issue myself with the exact same file and the exact same maximum download amount, and I've got 1G/1G symmetrical; it's definitely not an issue with my local network and likely . Can you add debugging to the file downloading so we can identify which file this is that's having issues? The webserver is likely killing the download for being too large or something. cmd_w5fnksmhaK

Iamgoofball commented 8 months ago

Tracked it down; something's up with the import on funASR, removing it from the fasterwhisper_asr.py entirely bypasses this issue. Ditch this line on line 12: from tools.asr.funasr_asr import only_asr then remove the section using that import in execute_asr() if you need just english transcription work done via Faster Whisper