Speech Recognition Options
Model Configuration Options
| Option | Type | Default | Description |
|---|---|---|---|
| device | string | ’webgpu’ | Device to run model on |
| onProgress | function | - | Callback for loading progress updates |
| quantization | string | ’q4’ | Model quantization level |
Available Models
| Model | Description | Size | Languages |
|---|---|---|---|
| whisper-tiny-en | Lightweight English-only model | ~150MB | English |
| whisper-base-all | Base multilingual model | ~290MB | 99 languages |
| whisper-small-all | Enhanced multilingual model | ~490MB | 99 languages |
Transcription Parameters
| Parameter | Type | Default | Description |
|---|---|---|---|
| language | string | - | Target language code (optional) |
| task | string | ’transcribe’ | Task type (‘transcribe’ or ‘translate’) |
| return_timestamps | boolean | false | Return word-level timestamps |
Example
const browserAI = new BrowserAI();
// Load speech recognition model
await browserAI.loadModel('whisper-tiny-en', {
device: 'webgpu',
onProgress: (progress) => {
console.log('Loading model:', progress.progress + '%');
}
});
// Transcribe audio from blob/array
const result = await browserAI.transcribe(audioInput, {
language: 'en',
return_timestamps: true,
chunk_length_s: 30
});
console.log('Transcription:', result.text);
// "Hello, how are you today?"Recording Audio
BrowserAI provides built-in methods for recording audio:
// Start recording
await browserAI.startRecording();
// Stop and get audio blob
const audioBlob = await browserAI.stopRecording();
// Transcribe the recorded audio
const transcription = await browserAI.transcribeAudio(audioBlob);
console.log(transcription.text);