Insight for: Show HN: Speakrs Full PyAnnotate pipeline in Rust/ONNX 20-37x times faster macOS

Speakrs, a Rust/ONNX implementation of the PyAnnotate diarization pipeline.

Analyzed: May 27, 2026

This project directly addresses critical performance bottlenecks in speech processing pipelines, a key concern for B2B applications in call center analytics, meeting transcription, and voice AI. The 20-37x speed improvement on macOS, achieved by leveraging native CoreML and optimized hardware utilization, provides a substantial competitive advantage for developers targeting Apple's ecosystem. While Linux/CUDA gains are more modest, the elimination of the Python runtime simplifies deployment and reduces dependency overhead. This performance uplift translates directly into lower operational costs, faster processing times, and enhanced real-time capabilities for enterprises. The 'fast mode' offering a speed-accuracy trade-off provides crucial flexibility for diverse use cases, positioning Speakrs as a high-performance foundation for advanced audio analytics solutions.

diarization pipeline Rust ONNX Runtime CoreML Python runtime segmentation powerset decode overlap-add aggregation binarization embedding PLDA VBx clustering PyTorch MPS (Metal) CPU Neural Engine GPU batch mode fast mode accuracy

Hacker News Post

Parent Entity

Show HN: Speakrs Full PyAnnotate pipeline in Rust/ONNX 20-37x times faster macOS

Score: 2