SheetSage-Pro open-sources bilingual audio-to-lead-sheet transcription ahead of ISMIR 2026
affige_yang · x · 2026-11-10
SheetSage-Pro, a Late-Breaking Demo at ISMIR 2026, is now open source under MIT. Feed it a song (local file or YouTube link) and it produces an editable lead sheet with melody, chord symbols, bar lines from downbeat detection, section labels, and note-aligned lyrics in Mandarin or English, output as JSON, MusicXML, or MIDI.
The pipeline chains pretrained models on one metrical grid: Mel-Band RoFormer for source separation, Beat This! for beat/downbeat tracking, All-In-One (+SongPrep labels) for structure, GAME for melody transcription, BTC variants for chords, SongPrep for lyric transcription, and Phonsa/SOFA for lyric alignment.
More from Research
- MSK and Illumina AI Lab publish bioRxiv study on how hosts regulate anellovirus persistence — SeyoneC · 2026-11-10
- Reviewer: paper I rejected at NeurIPS 2026 was resubmitted barely changed to ICLR 2027 — 3scorciav · 2026-11-10
- HARP-Bench: learned robot policies score 6-19% on 37 tasks involving humans — gan_chuang · 2026-11-10
- One Euler step suffices: VLAs get 1.6-4.8x faster with no retraining — kaixhin · 2026-11-10
- Bengio-led team releases 182-page mechanistic interpretability survey, tops alphaXiv — burny_tech · 2026-11-10
- Do LLM agents spread reward-hacking like a virus? Experiment says no — burny_tech · 2026-11-10