Difference between this checkpoint and the repo’s

#27
by nateraw - opened

Hi there! Thanks so much for this massive release :)

I noticed the config here is different than stable audio 1.0 config in the stable audio tools repo. So this is a new model that was previously unreported in the paper?

This is a new model based on the architecture from the Stable Audio 2 paper, but trained on CC data. There’ll be a technical report covering training and metrics for this model in the near future.

thanks!

nateraw changed discussion status to closed

Sign up or log in to comment