search

openai / jukebox

Code for the paper "Jukebox: A Generative Model for Music"
https://openai.com/blog/jukebox/
Other
7.72k stars 1.4k forks source link

Instrumental/Singing Voice only Generation #116

Open shoegazerstella opened 4 years ago

shoegazerstella commented 4 years ago

Hi all, this is a great project and I'm having lots of fun experimenting with it!

My question is: Would it be possible to force the model to only generate instrumental parts or singing voice only? I am trying the continuation from an instrumental song of mine, without any lyrics, and I notice sometimes there's some random singing. Do you think this could be done? Maybe by fine-tuning on voice-only tracks, do you think the generation, in this case, would only be random singing without any instrumental background? Thanks!

svntv commented 4 years ago

Great feature! I was wondering if there is a parameter that can be set or maybe putting an empty lyrics or something like that?

LucarioGamer commented 4 years ago

I would love to see this but as you know OpenAI makes a cool ai but abandons it and leaves it super unoptimized.

Cortexelus commented 3 years ago

You could use OpenMix or Spleeter or Wave-U-Net to remove the vocals (works on generative music too)