| --- |
| license: mit |
| library_name: onnx |
| tags: |
| - singing |
| - ultrastar |
| - note-transcription |
| - onnx |
| --- |
| |
| # Tune Studio note refiners |
|
|
| Two small gradient-boosted tree models (scikit-learn `HistGradientBoosting`, exported to ONNX) that correct notes |
| transcribed from a vocal stem, the way human UltraStar charters write them. Made for Tune Studio. |
|
|
| - `notefix.onnx`: pitch corrections. Charters often write a note a semitone or two away from the sung pitch |
| (slides, ornaments, intonation). Input `features` `[notes, 24]`; outputs `label` and `probabilities` |
| `[notes, 5]` for shifts of -2, -1, 0, +1, +2 semitones (`notefix.json`). Tune Studio shifts a note when the |
| most likely non-zero shift has probability ≥ 0.5. On held-out songs the corrected pitch matches the chart for 84.3% of |
| notes. |
| - `ends.onnx`: note ends. Charters end notes earlier than the voice stops and leave a gap before the next one. |
| Input `features` `[notes, 27]`; output `variable` `[notes, 1]`, seconds to add to each note's end. On held-out songs |
| the end is off by 62 ms on average, against 144 ms without it. |
|
|
| The features are computed by Tune Studio from the pitch track, the note transcription and the neighbouring notes. |
| Both were trained on Tune Studio's own transcriptions of about 2,600 songs, against those songs' hand-made |
| UltraStar charts (none of the songs used to test the pipeline). |
|
|
| The ONNX files give the same decisions as the scikit-learn models on all 319,616 notes of the training data. |
|
|
| ## License |
|
|
| MIT. |
|
|
