Skip to content

A speech recognition library running in the browser thanks to a WebAssembly build of Vosk

License

Notifications You must be signed in to change notification settings

ennuicastr/vosk-browser

 
 

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

54 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Vosk-Browser

A somewhat opinionated speech recognition library for the browser using a WebAssembly build of Vosk

This library picks up the work done by Denis Treskunov and packages an updated Vosk WebAssembly build as an easy-to-use browser library.

Note: WebAssembly builds can target NodeJS, the browser's main thread or web workers. This library explicitly compiles Vosk to be used in a WebWorker context. If you want to use Vosk in a NodeJS application it is recommended to use the official node bindings.

Live Demo

Checkout the demo running in-browser speech recognition of microphone input or audio files in 13 languages.

Installation

You can install vosk-browser as a module:

$ npm i vosk-browser

You can also use a CDN like jsdelivr to add the library to your page, which will be accessible via the global variable Vosk:

<script type="application/javascript" src="https://cdn.jsdelivr.net/npm/vosk-browser@0.0.5/dist/vosk.js"></script>

Usage

See the README in ./lib for API reference documentation or check out the examples folder for some ways of using the library

Basic example

One of the simplest examples that assumes vosk-browser is loaded via a script tag. It loads the model named model.tar.gzlocated in the same path as the script and starts listening to the microphone. Recognition results are logged to the console.

async function init() {
    const model = await Vosk.createModel('model.tar.gz');

    const recognizer = new model.KaldiRecognizer();
    recognizer.on("result", (message) => {
        console.log(`Result: ${message.result.text}`);
    });
    recognizer.on("partialresult", (message) => {
        console.log(`Partial result: ${message.result.partial}`);
    });
    
    const mediaStream = await navigator.mediaDevices.getUserMedia({
        video: false,
        audio: {
            echoCancellation: true,
            noiseSuppression: true,
            channelCount: 1,
            sampleRate: 16000
        },
    });
    
    const audioContext = new AudioContext();
    const recognizerNode = audioContext.createScriptProcessor(4096, 1, 1)
    recognizerNode.onaudioprocess = (event) => {
        try {
            recognizer.acceptWaveform(event.inputBuffer)
        } catch (error) {
            console.error('acceptWaveform failed', error)
        }
    }
    const source = audioContext.createMediaStreamSource(mediaStream);
    source.connect(recognizerNode);
}

window.onload = init;

Todos

  • Write tests
  • Automate npm publish

About

A speech recognition library running in the browser thanks to a WebAssembly build of Vosk

Resources

License

Stars

Watchers

Forks

Packages

No packages published

Languages

  • JavaScript 46.1%
  • TypeScript 29.4%
  • C++ 14.8%
  • Makefile 5.7%
  • Dockerfile 3.3%
  • HTML 0.7%