2026Kyutai and Mireloopen source
MuScriptor
v0.2.1
Music Transcription
Audio-to-MIDI
MuScriptor is an open-weight, multi-instrument music transcription model that converts complex, multi-instrument audio recordings into symbolic MIDI representations. Developed by Kyutai and Mirelo, the model leverages a decoder-only Transformer architecture trained on a large-scale dataset of 170,000 real-world songs across diverse genres.
Multi-instrument transcription from full audio mixes
Instrument conditioning to selectively transcribe specific instruments
Three-stage training pipeline (synthetic pre-training, real-world fine-tuning, and reinforcement learning)