@ishant89/vosk-speech-kit (React Native)

Offline Vosk speech-to-text for React Native (iOS). Thin bridge over the native VoskSpeechKit Swift core.

Install

npm install @ishant89/vosk-speech-kit
cd ios && pod install

The native core (VoskSpeechKit, which vendors libvosk) resolves automatically from CocoaPods — it's a dependency of this package's podspec, so no extra Podfile entry is needed.

Add the model (required)

The model is not shipped. Bundle it with your iOS app:

1. Get a model (https://alphacephei.com/vosk/models), unzip so it has am/ conf/ graph/.
2. Drag the atc-lgraph-v10/ folder into your Xcode project as a folder reference (blue folder), e.g. under a Models/ group.
3. At runtime, resolve its on-device path and load it.

Usage

import VoskSpeech from '@ishant89/vosk-speech-kit';
import { MainBundlePath } from 'react-native-fs';

const info = await VoskSpeech.loadLatestModel(`${MainBundlePath}/Models`); // or loadModel(exactPath)
VoskSpeech.onPartialResult(t => console.log('partial', t));
const sub = VoskSpeech.onResult(t => console.log('result', t));
await VoskSpeech.start();           // optional grammar: string[]
// …
VoskSpeech.stop();
sub.remove();

API: loadModel, loadLatestModel, availableModels, engineVersion, start, stop, unload, onPartialResult, onResult, onError. iOS only for now.