Sherpa-onnx

Offline, cross-platform speech processing suite (ASR, TTS, VAD) leveraging ONNX Runtime and Next-gen Kaldi technology.

Sherpa-onnx is a high-performance, open-source speech processing framework from the Next-gen Kaldi team. It delivers real-time, offline Speech-to-Text (ASR), Text-to-Speech (TTS), and Voice Activity Detection (VAD) by utilizing ONNX Runtime for efficient neural network computation (replacing PyTorch). The core strength is its vast platform support: it runs on embedded systems, Android, iOS, Linux, macOS, and Windows, supporting multiple NPUs and over 12 programming languages. This architecture ensures low-resource consumption and local processing, making it ideal for privacy-focused and constrained environments.

https://github.com/k2-fsa/sherpa-onnx