news
115

A new Show HN audio-to-MIDI model claims to be 60x more efficient while outperforming MuScriptor 1.3B and slightly beating Mirelo’s closed-source model. It uses only 20 million parameters and was trained without copyrighted recordings.

A new Show HN audio-to-MIDI model claims to be 60x more efficient while outperforming MuScriptor 1.3B and slightly beating Mirelo’s closed-source model. It uses only 20 million parameters and was trained without copyrighted recordings.
Comments
0Discuss on Hacker NewsOnly full members can comment
Log in to express your opinion.
No comments yet.