Open-source voice typing · Powered by your Groq account

Speak naturally.
Keep the clean version.

Watch your words appear while you speak—or use the original single-upload mode—then remove filler words and false starts with a fast cleanup model.

No account, database, analytics, or server proxy. Your key stays in this tab.

03 · Dictate

Ready when you are

Enter your key, then record a short message.

Result

Clean transcript

Your cleaned dictation will appear here.
View raw Whisper output
Nothing transcribed yet.

Performance

Where the time went

— ms total
Recording
Whisper + upload
AI cleanup
Pipeline total

Run a dictation to see models, audio size, and exact timings.

Use it in your project

One library, any interface.

The demo is powered by the published npm package. The original API remains unchanged, with near-live sessions available as an additive preview.

npm install groq-dictation-kit

import { DictationPipeline } from "groq-dictation-kit";

const pipeline = new DictationPipeline({ apiKey });
const result = await pipeline.dictate(audio);

// Near-live partials
const session = pipeline.startLiveConversation({ onEvent });