HarmonyOS

In this section, we describe how to build an HarmonyOS app with sherpa-onnx.

Hint

For real-time speech recognition, it does not need to access the Internet. Everyting is processed locally on your phone.

• Pre-built HAPs
• On-device speaker identification (本地说话人识别)
  • Open the project with DevEco Studio
  • Select a model
    • Use 3dspeaker_speech_eres2net_base_200k_sv_zh-cn_16k-common.onnx
• On-device text-to-speech (TTS)
  • Open the project with DevEco Studio
  • Select a text-to-speech model
    • Use vits-melo-tts-zh_en
    • Use vits-piper-en_US-libritts_r-medium
• On-device VAD + ASR
  • Open the project with DevEco Studio
  • Download a VAD model
  • Select a non-streaming ASR model
    • Use sherpa-onnx-moonshine-tiny-en-int8
    • Use sherpa-onnx-sense-voice-zh-en-ja-ko-yue-2024-07-17
• How to build sherpa_onnx.har
  • Download commandline-tools
  • Build sherpa-onnx for HarmonyOS
  • Build sherpa_onnx.har
    • From the command-line
    • Use DevEco Studio
  • Use sherpa_onnx.har in your project
  • Colab demo