News09/07/2026
Mirelo and Kyutai release audio-to-MIDI research and a Studio transcription tool
Mirelo AI GmbH →Mirelo announced an audio-to-MIDI model and research release with Kyutai on July 9, 2026. The model transcribes a musical recording into separate instrument tracks, identifying parts such as vocals, drums, bass and keyboards. Mirelo also introduced an improved version through its Studio product, allowing users to upload WAV or MP3 recordings and work with the resulting notation. The workflow includes instrument detection, editable notes on a piano roll and MIDI export for use in digital audio workstations. The announcement linked model weights, source code and a research paper, while distinguishing the openly released model from the improved version offered in Studio.