KdaiP / StableTTS

Next-generation TTS model using flow-matching and DiT, inspired by Stable Diffusion 3
MIT License
290 stars 31 forks source link

Cqt-Diff + #17

Open emmanuelinfante opened 3 weeks ago

emmanuelinfante commented 3 weeks ago

Hello good afternoon. I found this project very interesting, its combination and inspiration with different projects. I have a proposal for you, I hope you are interested.

I am working on the modification of a project called "Cqt-Diff +" is a unet-based diffusion model that analyzes the audio with the help of the Constant Q Transform analysis algorithm for audio, thanks to which it seeks to increase the resolution of an mp3 (128k for example) to the standard quality of a cd (44.1Khz/16Bits).

If you are interested in listening to the results, you can see them here:

http://research.spa.aalto.fi/publications/papers/icassp23-cqt-diff/

I am interested in having you work with me, in the modification of a project in order to improve the architecture. I hope you will consider it, If you have a way to contact you, it would be great to discuss it further, thank you.