docs: PR #654 merged — pip install mlx-audio>=0.4.3

README.md

@@ -65,8 +65,8 @@ The relicense was independently confirmed with the original author the week of A
 
 ## Conversion
 
-- **Tool:** `mlx_audio.sts.models.mel_roformer.convert` (custom port — pending upstream merge to [Blaizzy/mlx-audio](https://github.com/Blaizzy/mlx-audio))
-- **Tool version:** [`8380ab8`](https://github.com/xocialize/mlx-audio/commit/8380ab8) on the `feat/mel-band-roformer` branch (xocialize fork)
+- **Tool:** `mlx_audio.sts.models.mel_roformer.convert` — merged upstream into [Blaizzy/mlx-audio](https://github.com/Blaizzy/mlx-audio) via [PR #654](https://github.com/Blaizzy/mlx-audio/pull/654) (2026-04-27) and shipped in [`mlx-audio==0.4.3`](https://pypi.org/project/mlx-audio/0.4.3/) and later.
+- **Tool version at conversion time:** [`8380ab8`](https://github.com/xocialize/mlx-audio/commit/8380ab8) on the `feat/mel-band-roformer` branch (xocialize fork) — this is the exact commit that produced `model.safetensors`. The merged upstream code is functionally equivalent.
 - **mlx (Python) version:** 0.31.0
 - **Architecture preset:** `MelRoFormerConfig.kim_vocal_2()`
 - **Output precision:** `bfloat16`
@@ -96,7 +96,7 @@ Vocal isolation for music source separation. Input is a stereo music mixture; ou
 
 ### Python (mlx-audio)
 
-> Requires the Mel-Band-RoFormer architecture from the [pending mlx-audio PR](https://github.com/Blaizzy/mlx-audio); not yet on the released `pip install mlx-audio` line. Once merged, the snippet below works as-is. In the meantime, install the fork branch directly.
+> The Mel-Band-RoFormer architecture is included in [`mlx-audio>=0.4.3`](https://pypi.org/project/mlx-audio/0.4.3/) (merged via [PR #654](https://github.com/Blaizzy/mlx-audio/pull/654) on 2026-04-27). Install with `pip install "mlx-audio>=0.4.3"`.
 
 ```python
 import soundfile as sf