k2-fsa / sherpa-onnx

项目摘要

Sherpa-onnx is a comprehensive speech processing toolkit built on next-gen Kaldi with ONNX runtime, offering speech recognition, text-to-speech, speaker diarization, voice activity detection, and more. It supports offline operation across 12 programming languages and numerous platforms including mobile, embedded systems, and servers.

为什么值得关注

It provides exceptional cross-platform and cross-architecture support (including NPUs from RK, Axera, Ascend) with offline-capable speech AI, making it highly accessible for diverse deployment scenarios from Raspberry Pi to mobile devices.

优势

• Exceptional cross-platform and cross-architecture support (Android, iOS, Windows, macOS, Linux, HarmonyOS, RISC-V)
• Comprehensive speech processing capabilities including ASR, TTS, VAD, diarization, and enhancement
• Wide language bindings supporting 12 programming languages plus WebAssembly
• Offline operation without internet connection requirement
• Active development with 214 contributors and frequent releases

局限性

• No automated tests configured
• No Docker support for containerized deployment
• No examples provided in repository

使用场景

• Building offline speech recognition applications for mobile and embedded devices
• Cross-platform text-to-speech deployment
• Speaker diarization and identification systems
• Voice activity detection in IoT and edge computing devices