Runs in the browser · Speaker diarization
Run pyannote Speaker Segmentation 3.0 in your browser
pyannote Speaker Segmentation 3.0 downloads 2.8 MB over WebGPU (q4f16), or 1.5 MB over WebAssembly (uint8), to work out who is speaking and when, from raw audio entirely in the tab with Transformers.js. No install, no server.
Speaker diarization · onnx-community/pyannote-segmentation-3.0
- Parameters
- -
- Pipeline task
- custom
Reading
Frame-wise speaker segmentation, not transcription. transformers.js has no dedicated diarization pipeline() task; use AutoModel and decode the frame-level speaker-activity output yourself.
The WASM build downloads smaller here: 1.5 MB (uint8) against 2.8 MB (q4f16) for WebGPU. The CPU-fallback quant compresses tighter than the GPU pick for this model.
The quant ladder spans 3.9x: 1.5 MB (uint8) to 5.7 MB (fp32).
Will it run in your browser?
Checking for WebGPU support…
All measured variants
| Variant | Size |
|---|---|
| uint8 WASM pick | 1.5 MB |
| q4f16 WebGPU pick | 2.8 MB |
| fp16 | 2.9 MB |
| q8 | 2.9 MB |
| bnb4 | 5.5 MB |
| q4 | 5.5 MB |
| fp32 | 5.7 MB |
Sizes measured from the HuggingFace API file tree for onnx-community/pyannote-segmentation-3.0, not estimated.
Use it with Transformers.js
There is no stable transformers.js pipeline() call for this model yet, and no verified low-level snippet either (see the Reading panel above). Check the model card on HuggingFace for current usage examples.
Sources
FAQ
WebGPU or WASM for pyannote Speaker Segmentation 3.0, in practice?
Chrome, Edge, and Safari 26+ run the WebGPU build (q4f16) on the GPU. Firefox has WebGPU on Windows and Apple Silicon Macs but not everywhere yet. Any browser without a WebGPU adapter falls back automatically to the WebAssembly build (uint8) on CPU: same model, slower to load, slower to run.
What does q4f16 mean for pyannote Speaker Segmentation 3.0?
The WebGPU build here uses q4f16: 4-bit weights with some layers kept at 16-bit for stability: WebGPU's usual smallest clean build. The WASM fallback uses uint8: 8-bit integers: about a quarter the size of fp32, with a smaller quality trade than 4-bit.
Sizes measured 2026-08-01 from the HuggingFace API. Last validated 2026-08-03. See all browser models or the methodology.