Any thoughts on model distillation for faster realtime on CPU? | RVC Developers | Page 1

supple crescent May 31, 2024, 6:58 PM

#

recently some papers claims to have achieved a great infer performance improvement by distillation technique (on so-vitis), I was thinking that this could be applied to RVC's pretrained as well... any thoughts?

link to Distillation paper -> https://arxiv.org/abs/2401.01792

arXiv.org

CoMoSVC: Consistency Model-based Singing Voice Conversion

The diffusion-based Singing Voice Conversion (SVC) methods have achieved remarkable performances, producing natural audios with high similarity to the target timbre. However, the iterative sampling process results in slow inference speed, and acceleration thus becomes crucial. In this paper, we propose CoMoSVC, a consistency model-based SVC meth...

supple crescent May 31, 2024, 7:31 PM

#

Any thoughts on model distillation for faster realtime on CPU?

inland isle May 31, 2024, 7:46 PM

#

This is really interesting. I'd suggest posting this here as well: https://github.com/RVC-Project/Retrieval-based-Voice-Conversion-WebUI/issues since the developers are more likely to see it there

GitHub

Issues · RVC-Project/Retrieval-based-Voice-Conversion-WebUI

Easily train a good VC model with voice data <= 10 mins! - Issues · RVC-Project/Retrieval-based-Voice-Conversion-WebUI

supple crescent May 31, 2024, 11:28 PM

#

ty gonna do that 😉

#Any thoughts on model distillation for faster realtime on CPU?