optional MuFun extraction of lyrics from audio library #1986

Open

Labels

difficult-featureenhancement

opened

on Nov 19, 2025

while currently ACE-Step training requires manual tagging and writing lyrics into a separate file, it's possible to use the MuFun to speculatively decode the audio latents from MusicDCAE into text lyrics, sort of like BLIP for audio (BLAP?)

to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Assignees

No one assigned

Labels

difficult-featureenhancement

Projects

No projects

Milestone

No milestone

Relationships

None yet

Development

No branches or pull requests