Open sersalih opened 1 year ago
Last month I tried the Ultimate Vocal Remover GUI which contains Demucs model. It has better separation result, but the processing speed is extremely slow.
I'm using a desktop PC with AMD 5900X CPU. Just now I tested the official python program of Demucs. Separating a song of 07:32 into 4 tracks costed 02:10 * 4 = 08:40, about 0.87x in average.
But using SpleeterMsvcExe with 4stems-22khz model, it only took 25 seconds, about 18.08x speed.
Due to the huge processing speed problem, I'm not much interested in it. So writing a program that support Demucs model may take a while.
Hi, thanks for great effort about SpleeterMsvcExe :-) It's easy and fast :-) Can you add the Demucs 3 algorithm to your project? The Demucs 3 has high quality separation model B. Please check here
An example project is written in python : here and main website here
I hope, you'll do it, Best regards,