does anyone knows a software like Ultimate Vocal Remover but with 24 bits 48 khz support?

thank you in advance

 

If you own Intel hardware, Audacity + Inte's OpenVINO™ AI Plugins for Audacity could be an alternative.

 

uvr support 32 bit but no 48khz... and that is enough.
@tommyzai was also in search for alternative to uvr with 48khz support, maybe you can ask him what he find...
but consider, other platforms use other models and architectures... its better to use 44 instead of 48 if the quality o separation is better.

 

LALAL.AI ---> 24 bits 48 khz ---> www.lalal.ai

While LALAL.AI can work with various audio qualities, we recommend a sample rate of 44.1kHz or 48kHz and a bit depth of 24 bits to capture more audio detail and enhance voice modification accuracy. www.lalal.ai/guides/voice-sample-guidelines

 

I mean - the process grinds down your audio into millions of tiny pieces, then an AI hallucinates a very educated guess of the other missing millions of pieces, then this atomised AI-goop is recombobulated into audio. what is there to improve on by going from 44.1/16 to 48/24?

 

it's only going to be barely or marginally better, because the nyquist frequency is 24k instead of 22k. If there is no real information in that tiny band, isn't the AI going to basically ignore it anyway?

 

Exactly, they've all been trained on 44.1/16 sources anyway. What useful information with this little training data could possibly be constructed for 48/24 material?