simen-onnx-node

Node.js addon for speech-to-text, text-to-speech, and speaker diarization using ONNX Runtime.

Features

• Streaming ASR: Real-time speech recognition
• Non-streaming ASR: Batch speech recognition
• TTS: Text-to-speech synthesis
• VAD: Voice activity detection
• Speaker Identification: Speaker recognition and verification
• Keyword Spotting: Wake word detection

Installation

Prerequisites

1. Node.js >= 16.0.0
2. CMake >= 3.15
3. C++ compiler (GCC, Clang, or MSVC)

Download Dependencies

# macOS/Linux
chmod +x scripts/download-deps.sh
./scripts/download-deps.sh

# Windows (PowerShell)
.\scripts\download-deps.ps1

Build sherpa-onnx Library

git clone https://github.com/k2-fsa/sherpa-onnx
cd sherpa-onnx
mkdir build && cd build
cmake -DBUILD_SHARED_LIBS=ON -DCMAKE_INSTALL_PREFIX=./install ..
make -j4 install

# Copy to prebuilt directory
cp -r ./install /path/to/simen-onnx-node/prebuilt/sherpa-onnx

Build the Addon

npm install
npm run build

Usage

const sherpaOnnx = require('simen-onnx-node');

// Create streaming recognizer
const recognizer = sherpaOnnx.createOnlineRecognizer({
  modelConfig: {
    transducer: {
      encoder: './encoder.onnx',
      decoder: './decoder.onnx',
      joiner: './joiner.onnx',
    },
    tokens: './tokens.txt',
    numThreads: 2,
  },
  enableEndpoint: true,
});

// Create stream and process audio
const stream = sherpaOnnx.createOnlineStream(recognizer);
sherpaOnnx.acceptWaveformOnline(stream, { samples: audioData, sampleRate: 16000 });

// Get result
const result = sherpaOnnx.getOnlineStreamResultAsJson(recognizer, stream);
console.log(JSON.parse(result));

Project Structure

simen-onnx-node/
├── src/                 # Node addon C++ bindings
├── lib/                 # JavaScript wrappers
├── native/              # Core C++ source (reference)
│   ├── csrc/            # Core implementation
│   ├── c-api/           # C API layer
│   └── cmake/           # CMake modules
├── prebuilt/            # Prebuilt dependencies (gitignored)
│   ├── onnxruntime/     # ONNX Runtime library
│   └── sherpa-onnx/     # sherpa-onnx library
└── scripts/             # Build scripts

License

Apache-2.0