Audio-to-MIDI model 60x more efficient
kobi.music

Audio-to-MIDI model 60x more efficient

The gist

Developer kobimusic released Dettato, an open-source audio-to-MIDI model at 20M parameters, roughly 60x smaller than MuScriptor 1.3B.
It claims to beat the next-best open model without training on copyrighted recordings.

How it unfolded

  1. Nov 16, 2026kobimusic publishes "introducing dettato," describing it as a state-of-the-art audio-to-MIDI model at 20M parameters, roughly 65x smaller than MuScriptor, trained on no copyrighted recordings and open source with commercially usable outputs.
  2. Nov 17, 2026WO! Discover lists the project as "I made an audio-to-MIDI model 60x more efficient."
  3. Nov 20, 2026Hacker News user hidude562 posts a Show HN: "I released dettato, my open weight audio to midi transcription model. It beats muscriptor 1.3b on every set and is roughly 60x smaller, without training on copyrighted recordings."

Key terms

audio-to-MIDI
Conversion of audio signals into MIDI data, which encodes musical instructions such as pitch, timing, and velocity.
open-source
Software whose source code is freely available for anyone to use, study, modify, and distribute.

Source: Wikipedia

Sources

Read the full story in the app Gistline — free on the App Store