---
license: mit
library_name: coreml
tags:
- audio
- source-separation
- vocals
- coreml
---

# spoti.pw Sing voice model

The vocal separator [spoti.pw](https://spoti.pw)'s Sing runs on the iPhone to turn a song's vocals down
while it plays. It is Mel-Band RoFormer with KimberleyJensen's vocal checkpoint, its spectral core
exported for Core ML with two-second windows: a float32 `spectrum` of shape `[1, 2050, 201, 2]` in, the
`vocals_spectrum` of the same shape out, the STFT around it done by the app. Normalization, attention,
softmax and matrix products are float32, the rest float16. The files are a compiled `separator.mlmodelc`
as they are: the app downloads each one and keeps it only if its size and SHA-256 are the ones it pins.
It needs iOS 27, and no audio leaves the phone.

| File | Bytes | SHA-256 |
| --- | ---: | --- |
| `weights/weight.bin` | 488986336 | `970a99fb4b15724bf76d2918ceb177df592c69265d3e2fabaab6e5ba72738e62` |
| `model.mil` | 669060 | `c612ac3798e1ce0f89603453ff72ca78bf4f1b07701925afa726ffce4c6e3ecb` |
| `metadata.json` | 2416 | `af098e7f6b0af360cb2d5df7c3059fe83e3e2f38c27e3f67c875417c8d9d4102` |
| `coremldata.bin` | 507 | `5ff2c235fcf4e153f1ea47ee9939132ebad276dc7509fc1069b45f7ef8c0207d` |
| `analytics/coremldata.bin` | 243 | `9a949d4e0b28ab778d750f30ec3bb120b22a8c01ef8dce25f11b0832d6d5fe39` |

## Provenance

* Checkpoint: [KimberleyJSN/melbandroformer](https://huggingface.co/KimberleyJSN/melbandroformer/tree/ac9b0614ab3cd7f77219e18ba494dfd93956c348), MIT.
* Conversion: [john-rocky/coreai-model-zoo](https://github.com/john-rocky/coreai-model-zoo/tree/5029e6df8100650fe175d3e276fffb0177754ca1/conversion/melband_roformer).
* Reference implementation: Mel-Band-Roformer-Vocal-Model at revision `25f44ffb55ee3c301281bba21b2d6d311cb69ae2`.
* Exported by `harness/sing/export_coreml.py` in the spoti.pw repository.

Mel-Band RoFormer by Ju-Chiang Wang, Wei-Tsung Lu and Minz Won; the vocal checkpoint by KimberleyJensen;
lucidrains' BS-RoFormer implementation; ZFTurbo's training code. The MIT notices are in `NOTICE`.