SenseVoice

This section describes how to use models from https://github.com/FunAudioLLM/SenseVoice.

A single model from SenseVoice supports the following languages

• Chinese (Mandarin, 普通话)

• Cantonese (粤语, 广东话)

• English

• Japanese

• Korean

which is similar to what multilingual Whisper is doing.

We have converted SenseVoice to onnx and provided APIs for the following programming languages

• 1. C++

• 2. C

• 3. Python

• 4. C#

• 5. Go

• 6. Kotlin

• 7. Java

• 8. JavaScript (Support WebAssembly and Node)

• 9. Swift

• 10. Dart (Support Flutter)

Note that you can use SenseVoice with sherpa-onnx on the following platforms:

• Linux (x64, aarch64, arm, riscv64)

• macOS (x64, arm64)

• Windows (x64, x86, arm64)

• Android (arm64-v8a, armv7-eabi, x86, x86_64)

• iOS (arm64)

In the following, we describe how to download pre-trained SenseVoice models and use them in sherpa-onnx.

• Huggingface space
• Export SenseVoice to sherpa-onnx
  • The code
  • Test the exported model
  • Where to find exported models
• Pre-trained Models
  • sherpa-onnx-sense-voice-zh-en-ja-ko-yue-2024-07-17
    • Download
    • Decode a file with model.onnx
      • Without inverse text normalization
      • With inverse text normalization
      • Specify a language
    • Speech recognition from a microphone
    • Speech recognition from a microphone with VAD
• C API for SenseVoice
  • Explanations
    • 1. Download sherpa-onnx
    • 2. Download the model
    • 3. Build sherpa-onnx
    • 4. View the build result
    • 5. Build the C API example
    • 6. Run it
    • 7. Where to find sense-voice-c-api.c
• Dart API for SenseVoice
  • Explanations
    • 1. Download the code
    • 2. Download the sherpa-onnx package
    • 3. Run it
• Python API for SenseVoice
  • Decode a file
  • Speech recognition from a microphone
  • Generate subtitles
    • Chinese
    • English
  • WebSocket server and client example
    • 1. Start the server
    • 2. Start the client (decode files sequentially)
    • 3. Start the client (decode files in parallel)
    • 4. Start the Web browser client
      • Upload a file for recognition
      • Record your speech with a microphone for recognition