---
license: cc-by-4.0
tags:
- speaker-diarization
- coreml
- mirror
- whispall
---

# Whispall mirror — Speaker diarization (CoreML)

**Cold-standby** availability mirror of the CoreML speaker-diarization models used by
**Whispall** (local, privacy-first macOS dictation & assistant app) for meeting-note
speaker labeling on Apple Silicon (via the FluidAudio runtime).

- **Mirrored from:**
  [FluidInference/speaker-diarization-coreml](https://huggingface.co/FluidInference/speaker-diarization-coreml)
  (upstream revision `1ed7a662fdc7109e36d822db793ee6eebdaf8594` at mirror time).
- **Note:** Whispall's FluidAudio runtime downloads from the upstream repo directly (the
  repo ID is baked into the library). This mirror exists so the bytes stay available if
  upstream ever changes; switching to it would require rebuilding the runtime.

## License & attribution

**CC-BY-4.0.** Original models: the pyannote project (segmentation) and upstream speaker
embedding model authors; CoreML conversion:
[FluidInference](https://huggingface.co/FluidInference) (FluidAudio project). Redistributed
unmodified with attribution.