rknn

This section describes how to use sherpa-onnx with RKNPU from Rockchip.

The following boards are known to work:

• RK3588

• RK3576

• RK3568

• RK3566

• RK3562

• Tutorials
  • Zipformer on LubanCat
• Install
  • From pre-built wheels using pip install
  • Download pre-compiled binaries
  • Build sherpa-onnx directly on your board
  • Cross-compiling
    • Dynamic link
    • Static link
• Pre-trained models
  • sherpa-onnx-rk3588-streaming-zipformer-small-bilingual-zh-en-2023-02-16
    • Decode files
    • Real-time speech recognition from a microphone
  • sherpa-onnx-rk3588-streaming-zipformer-bilingual-zh-en-2023-02-20
    • Decode files
    • Real-time speech recognition from a microphone
  • sherpa-onnx-rk3588-20-seconds-sense-voice-zh-en-ja-ko-yue-2024-07-17 (Chinese, English, Japanese, Korean, Cantonese, 中英日韩粤语)
    • Decode long files with a VAD
    • Decode a short file
    • Speed test
      • Sense-voice int8 ONNX model on 1 Cortex A55 CPU
      • Sense-voice int8 ONNX model on 1 Cortex A76 CPU
      • Sense-voice RKNN model on 1 RK3588 NPU
  • sherpa-onnx-rk3588-20-seconds-sense-voice-zh-en-ja-ko-yue-2025-09-09 (Chinese, English, Japanese, Korean, Cantonese, 中英日韩粤语)