The Problem

Every voice app faces the same annoying problem: it can't tell when you're done talking.

You pause to think, and it cuts you off. You take a breath, and it responds too soon. You want to interrupt, and it keeps going.

The current solutions either:

• Detect silence (Silero VAD, ricky0123/vad): They know when sound stops, but they can't tell if you're thinking or finished.
• Use server-side AI (OpenAI Realtime, AssemblyAI): They are smart, but they add delay, costs, and privacy issues.

Utterance is different. It uses a lightweight ML model entirely on the client side. It recognizes the difference between a thinking pause and a completed turn. No cloud. No delay. No per-minute fees.

Quick Start

npm install @utterance/core

import { Utterance } from "@utterance/core";

const detector = new Utterance();

detector.on("turnEnd", (result) => {
  console.log("User is done speaking", result.confidence);
});

detector.on("pause", (result) => {
  console.log("User is thinking...", result.duration);
});

detector.on("interrupt", () => {
  console.log("User wants to speak — stop AI response");
});

await detector.start();

See the full documentation for detailed usage, API reference, and integration examples.

Comparison

| Feature | Silero VAD | ricky0123/vad | Picovoice Cobra | OpenAI Realtime | Utterance | | --- | --- | --- | --- | --- | --- | | Detects speech vs. silence | ✅ | ✅ | ✅ | ✅ | ✅ | | Semantic pause detection | ❌ | ❌ | ❌ | ✅ | ✅ | | Interrupt detection | ❌ | ❌ | ❌ | ✅ | ✅ | | Runs client-side | ✅ | ✅ | ✅ | ❌ | ✅ | | No API costs | ✅ | ✅ | ❌ | ❌ | ✅ | | Privacy (audio stays local) | ✅ | ✅ | ✅ | ❌ | ✅ |

Contributing

We're building Utterance in the open, and contributions are welcome.

git clone https://github.com/nizh0/Utterance.git
cd Utterance
npm install
npm start

See the contributing guide for development workflow, project structure, and areas where we need help.

Community

• Discord: Chat with contributors
• GitHub Issues: Bug reports & feature requests

License

MIT © Utterance