Speech Segmentation / Speaker Diarization

Offline speaker diarization using ONNX models (pyannote-segmentation-3.0).

Models

Model	Description
model.onnx	Base segmentation model from onnx-community/pyannote-segmentation-3.0
model_with_embedding.onnx	Extended version with speaker embeddings as an additional output (generated via speech_embedding_export.py)

Usage

Basic diarization

Outputs detected speakers with timestamps and confidence scores:

uv run python speech_diarizer.py

Automatically downloads the model and sample audio (mlk.wav), then prints segments like:

  SPEAKER_01      0.37s -    2.84s  (conf=0.951)
  SPEAKER_02      2.84s -    5.21s  (conf=0.876)

Diarization with embeddings

Extract per-segment speaker embeddings alongside timestamps:

uv run python speech_embedding.py

Output includes embedding dimensions for each segment, useful for downstream clustering or verification.

Export model with embeddings

Re-exports the base ONNX model to include the LeakyRelu activation (speaker embeddings) as a graph output:

uv run python speech_embedding_export.py
# Produces: model_with_embedding.onnx

Setup

Dependencies are managed via uv:

pip install uv          # if not already installed
uv sync                 # installs dependencies from pyproject.toml