🎤 AI Voice to Text

Speak into your device and AI writes down what you say.

The spoken language is detected automatically. Speak in short bursts (up to 2 minutes). Recordings are processed on our server and not stored.

Voice to Text — Speak and It Writes Down What You Say

Press record, say what you want written, and the transcript appears. This is the quickest way to turn speech into text when typing is inconvenient — when your hands are busy, when you think faster than you type, or when you simply want to get an idea down before it disappears. The spoken language is detected automatically, so there is nothing to set up first. Everything is processed privately on our server by an offline AI engine, so your voice is never handed to an outside service. No sign-up and no watermarks.

How to Convert Speech to Text

  1. Press Start Recording and allow microphone access when your browser asks.
  2. Speak normally — there is no need to slow down or over-pronounce.
  3. Press Stop & Transcribe. Your words appear in the box a few seconds later.
  4. Press Copy and paste the text wherever you need it.

When You Might Need This

  • Getting an idea down before you forget it, without opening anything
  • Dictating a message, an email or a note while your hands are busy
  • Drafting out loud first, then tidying the words up afterwards
  • Writing down a phone number, address or instruction someone has just told you
  • Saying something in a language you speak more easily than you type
  • Getting past a blank page by talking your way into it
  • Taking notes when holding a phone is easier than holding a keyboard

Key Features

  • Automatic language detection – no need to set the spoken language
  • Many languages supported – transcribes a wide range of languages
  • Punctuation included – full stops and capitals are added for you
  • Private – recordings are processed on our server and not stored
  • Works on any device with a microphone and a modern browser

How Accurate Is It?

Reliable on clear speech, and it punctuates as it goes. Accuracy depends far more on the recording than on the engine: speak at a normal pace, keep the microphone reasonably close, and avoid a room with a television on. Background noise, several people talking at once and a heavy echo are what cause most mistakes. Names, places and technical terms are the words worth glancing over afterwards, since there is no context from which to guess their spelling.

Recording Length

Each recording can run up to 2 minutes, which is roughly 250 to 300 words — more than most people dictate in one breath. Short bursts also transcribe faster and more accurately than long ones. If you already have a longer recording saved as a file, use Audio to Text instead, which takes an uploaded file and returns a Word document.

Frequently Asked Questions

Press Start Recording, say what you want written, then press Stop & Transcribe. Your words appear in the box, and the Copy button puts them on your clipboard ready to paste. There is nothing to install and no account to create — the only thing your browser asks for is permission to use the microphone, which you can withdraw again at any time.

You can, in stages. Each recording is capped at 2 minutes, and a new transcript replaces the previous one rather than being added to it — so copy each piece into your document before you record the next. For long-form dictation that tends to be the better way to work anyway: pausing to paste gives you a natural moment to collect the next thought, and short recordings come back more accurate than long ones.

This page listens to your microphone as you speak; Audio to Text works from a recording you already have. Use this one for dictation and quick notes, where the words do not exist yet. Use the other one for a file — a voice memo, an interview, a recorded call — where you want the whole thing transcribed and handed back as a Word document.

Yes. Full stops, commas, question marks and capitals are worked out from how you speak — your pauses and intonation — so you do not have to say “comma” or “full stop” out loud the way older dictation software required. It is least reliable on one very long unbroken sentence, so speaking in ordinary sentences with small pauses between them gives noticeably tidier text.

Nearly always the recording rather than the engine. Background noise, a microphone that is too far away, two people talking over one another, or a hard-surfaced room with an echo are the usual causes. The other common case is proper nouns — names of people, places and products have nothing in the surrounding sentence to hint at their spelling, so they are worth a glance afterwards. Re-recording just that sentence on its own usually fixes it.

No. It runs in your browser — allow microphone access and record. The transcription happens on our server, so nothing is downloaded to your device and it behaves the same on a phone, a tablet or a computer.

Yes. The audio is processed on our own server by an offline engine, is not stored, and is never sent to a third-party service or used to train anything. Microphone access also requires a secure (HTTPS) connection, which your browser enforces.

Check that you allowed microphone access when you were prompted. If you dismissed or blocked that request, reloading the page will not bring it back — the permission has to be reset in your browser's site settings for this page. The page also has to be loaded over HTTPS. Recording works in current versions of Chrome, Edge, Firefox and Safari, on both desktop and mobile.

Related Tools