sherpa-onnx-wasm-main-asr.data

Speech ASR - WebAssembly Model Packages

Browser-based real-time speech recognition powered by sherpa-onnx WebAssembly.

🚀 Quick Start

1. Install the SDK

npm install speech-asr@latest

2. Download Model Package

Choose the package that fits your needs:

Option A: 1-Pass (Online Only)

File: sherpa-onnx-wasm-asr-1pass.zip

• Fast, real-time ASR
• Includes: Zipformer model + Silero VAD + Punctuation
• Best for: Basic real-time transcription

Option B: 1&2-Pass (Online + Offline Enhancement)

File: sherpa-onnx-wasm-asr-1&2pass.zip

• Two-pass recognition for higher accuracy
• Pass 1: Fast online Zipformer
• Pass 2: Offline SenseVoice enhancement
• Includes: All models + VAD + Punctuation + Offline worker
• Best for: Maximum accuracy requirements

3. Usage

Extract the zip file to your project directory, for example: ./models/

Simple Configuration (Recommended)

import { SpeechASR } from 'speech-asr';

const asr = new SpeechASR({
  vadMode: 'silero',
  twoPass: { enabled: true },  // Set false for 1-pass only
  modelPaths: {
    m_path: './models/sherpa-onnx-wasm-asr-20251211-120752'  // Your extracted directory
  },
  onReady: () => console.log('ASR Ready'),
  onResult: (text) => console.log('Result:', text),
  onTwoPassResult: ({ offlineText }) => console.log('Enhanced:', offlineText),
});

await asr.init();
await asr.start();

📦 Package Contents

1-Pass Package

sherpa-onnx-wasm-asr-1pass/
├── sherpa-onnx-wasm-main-asr.data    # Model weights
├── sherpa-onnx-wasm-main-asr.js      # WASM wrapper
├── sherpa-onnx-wasm-main-asr.wasm    # WASM runtime
├── sherpa-onnx-asr.js                # ASR helper
└── sherpa-onnx-vad.js                # VAD helper

1&2-Pass Package (includes all above plus)

sherpa-onnx-wasm-asr-1&2pass/
├── ... (all files from 1-pass)
└── offline-worker.js                 # Two-pass worker

🔗 Links

• npm Package: speech-asr
• GitHub: easy-asr
• Examples: See examples/ directory in the repository

📝 Notes

• The .data file and .js files must be from the same version/package
• For Vite/Webpack projects, ensure proper WASM handling configuration
• All models run entirely in the browser - no backend required

🆘 Support

For issues and questions, please visit the GitHub repository.