Voxta: voxta-audio

Browser audio playback for the Voxta Server websocket API.

This library turns the audio referenced by replyChunk messages into ordered, gap-aware playback through the Web Audio API. It is decoupled from any UI: pair it with @voxta/voxta-client to hear a character speak.

This library is early beta and subject to change.

Example usage

import { AudioPlayback } from '@voxta/voxta-audio';
import { ServerReplyChunkMessage } from '@voxta/voxta-client';

const playback = new AudioPlayback<ServerReplyChunkMessage>();

// Unlock the AudioContext from a user gesture (e.g. a click), passing the output
// device id ('' for the default device).
await playback.resume('');

// When a replyChunk arrives with an audioUrl, queue it for playback.
voxtaClient.addEventListener('replyChunk', (e) => {
  if (e.detail.audioUrl) playback.play(e.detail.audioUrl, e.detail);
});

Announce the content types you can play when authenticating, via AudioPlayback.acceptedAudioContentTypes.

Building

Run nx build voxta-audio to build the library.

Running unit tests

Run nx test voxta-audio to execute the unit tests via Vitest.