SheetSage2
An open-source audio-to-sheet-music transcription model that automatically converts audio files into ABC notation format. The underlying system relies on pitch detection and beat-tracking algorithms to perform the transcription.
Timeline (last 90 days)
SheetSage2 performs full-song lead-sheet transcription, converting audio recordings into ABC notation, providing the symbolic supervision target s for YuE2 training.
YuE2's model weights, code, and the two auxiliary models MERT2 and SheetSage2 have all been released on GitHub and Hugging Face, and the project page is map-yue2.github.io.
The built-in SheetSage2 first extracts actual musical score information from the audio, including melody, chords, phrases, and segment structure.
The cover pipeline introduced in version 3.0 is: source audio → SheetSage2 → ABC notation → LLM-rewritten cover prompt → YuE2 generates cover.
SheetSage2 is an open-source model specifically designed to automatically transcribe audio into ABC format, relying on pitch detection and beat tracking algorithms at its core.
SheetSage2 transcribes the music itself rather than the original song's lyrics, so when making a cover, users need to either provide their own lyrics, ask AI to generate new lyrics, or take the purely instrumental route.
All Facts (6)
SheetSage2 is an open-source model specifically designed to automatically transcribe audio into ABC format, relying on pitch detection and beat tracking algorithms at its core.
60%UnverifiedSheetSage2 transcribes the music itself rather than the original song's lyrics, so when making a cover, users need to either provide their own lyrics, ask AI to generate new lyrics, or take the purely instrumental route.
60%UnverifiedSheetSage2 performs full-song lead-sheet transcription, converting audio recordings into ABC notation, providing the symbolic supervision target s for YuE2 training.
50%UnverifiedThe built-in SheetSage2 first extracts actual musical score information from the audio, including melody, chords, phrases, and segment structure.
50%UnverifiedThe cover pipeline introduced in version 3.0 is: source audio → SheetSage2 → ABC notation → LLM-rewritten cover prompt → YuE2 generates cover.
50%UnverifiedYuE2's model weights, code, and the two auxiliary models MERT2 and SheetSage2 have all been released on GitHub and Hugging Face, and the project page is map-yue2.github.io.
50%