search

nomadkaraoke / python-audio-separator

Easy to use stem (e.g. instrumental/vocals) separation from CLI or as a python package, using a variety of amazing pre-trained models (primarily from UVR)
MIT License
425 stars 72 forks source link

Killed for the long recording #120

Open Ylw2014 opened 1 week ago

Ylw2014 commented 1 week ago

GPU memory will be OOM when the recording time is too long ,for example 1 ~2 hours, any update plan for this scenerio?

by the way, great job! great work!

beveradb commented 2 days ago

I have no plan to solve that as it's well outside of my own use case, sorry! You can try other models, or try changing the parameters to use a different batch size etc. - but depending on which model architecture you're using it may also have some kinda memory leak which might be causing the issue.

You can also always chunk your input audio into smaller clips (e.g. using ffmpeg), separate each smaller segment, then join the outputs!

If you decide you want to improve audio-separator to make it handle larger input files more gracefully like this yourself, PRs are welcome :)