Slower then UVR GUI

[deletedcookie]

This takes 3 times longer to process the same video then on the GUI version.

I used the UVR-MDX-NET-Inst_Main model and the same test video (2:42 long) for my testing.

Using my GPU with PyTorch CUDA 11.8, I tried changing the chunk size and found the fastest results with chunk size 8 at 55 seconds. Using my CPU with PyTorch CPU, I got the fastest results without changing the chunk size at 52 seconds.

Using GPU conversion on UVR GUI and default settings, it takes 19 seconds. Using CPU conversion on UVR GUI and default settings, it takes 35 seconds.

Is there anything I can do to make it faster?

I also get this warning when I run it UserWarning: PySoundFile failed. Trying audioread instead. Would that have anything to do with the slower performance?

[seanghay]

What's is your audio format wav or mp3?

[deletedcookie]

The input is a webm file. I just tried it with wav and mp3 and it's not any faster but the warning goes away.

[xinliu9451]

This takes 3 times longer to process the same video then on the GUI version.

I used the UVR-MDX-NET-Inst_Main model and the same test video (2:42 long) for my testing.

Using my GPU with PyTorch CUDA 11.8, I tried changing the chunk size and found the fastest results with chunk size 8 at 55 seconds. Using my CPU with PyTorch CPU, I got the fastest results without changing the chunk size at 52 seconds.

Using GPU conversion on UVR GUI and default settings, it takes 19 seconds. Using CPU conversion on UVR GUI and default settings, it takes 35 seconds.

Is there anything I can do to make it faster?

I also get this warning when I run it UserWarning: PySoundFile failed. Trying audioread instead. Would that have anything to do with the slower performance?

I encountered the same issue. Are there any good solutions?