pyannote-onnx

Overview

pyannote-onnx is an open-source Python library hosted on GitHub by pengzhendong. It offers speaker diarization functionality using ONNX models, allowing developers to run pyannote-style audio speaker segmentation without requiring the full PyTorch-based pyannote.audio stack. By exporting models to the ONNX format, the project aims to make speaker diarization more portable and easier to deploy in environments where a lighter inference runtime is preferred. The project is licensed under Apache-2.0 and had its most recent release activity in December 2024. With 99 GitHub stars, it has attracted attention from developers working on audio processing and transcription pipelines. It is written in Python and suits teams or individuals who need speaker diarization capabilities integrated into custom workflows.

Highlights

• Speaker diarization using ONNX model inference
• Python-based library for audio processing pipelines
• Apache-2.0 open-source license allowing broad reuse
• Lightweight alternative to full PyTorch pyannote.audio deployment
• Active GitHub repository with recent 2024 release activity

Common use cases

• Running speaker diarization in Python applications
• Integrating speaker segmentation into transcription workflows
• Deploying audio analysis in environments with limited ML framework support
• Prototyping meeting or call analytics tools