@voxd/sdk

Use Vox transcription and text-to-speech from Bun or Node. The SDK connects to Vox Companion, which runs the speech models on the Mac.

Install

bun add @voxd/sdk

Or with npm:

npm install @voxd/sdk

Requires Node 22+ or Bun, plus Vox Companion running on the same Mac.

Quick start

import { VoxClient } from "@voxd/sdk";

const vox = new VoxClient({ clientId: "my-tool" });
await vox.connect();

const transcript = await vox.transcribeFile("/tmp/sample.wav");
console.log(transcript.text);

const speech = await vox.synthesize("Hello from Vox");
console.log(speech.audioBytes);

vox.disconnect();

Warm up before a request

Warm-up is explicit, so your app decides when to spend the startup cost.

await vox.startWarmup("parakeet:v3");
const result = await vox.transcribeFile("/tmp/sample.wav", "parakeet:v3");

console.log(result.text);
console.log(result.metrics?.totalMs);

Main API

• connect() and disconnect()
• transcribeFile() and annotateFile()
• listVoices() and synthesize()
• startWarmup(), scheduleWarmup(), and getWarmupStatus()
• createLiveSession()
• listHistory() and deleteHistoryRecord()
• doctor() and health()

Use a stable clientId so metrics and sessions can be traced back to your tool.

Choose the right package

• Bun or Node: @voxd/sdk
• Browser: @voxd/client
• Terminal: @voxd/cli
• macOS or iOS app: embed VoxCore and VoxEngine

Learn more

• SDK guide
• API reference
• Download Vox Companion

License

MIT