Skip to content

Free tool · No sign-up

Free online dictation that stays on your device.

Speak, and your words become text you can edit, copy or download. Speech is recognized right here in your browser, so your voice is never uploaded. No account, and as many recordings as you like.

Press Space or Start speaking. A one-time model download comes first.

English model. Up to 2 minutes per recording, as many recordings as you like. Audio and text stay in this tab, so copy or download your text before you close it.

DeutschFrançaisEspañolItalianoPortuguês

How it works

Three steps, all in this tab.

  1. 1

    Set up once.

    Your browser downloads a Whisper speech model (about 80 MB, or about 165 MB where your browser uses the graphics chip) and keeps it cached. Your microphone stays off during setup.

  2. 2

    Speak.

    Press Space or Start speaking, allow the microphone, and talk for up to 2 minutes. Press Space again to finish. Esc cancels a recording.

  3. 3

    Edit, copy or download.

    Each recording is added to your text. Fix anything by typing, then copy it or save it as a .txt file. Nothing is kept after you close the tab.

Why it is different

Dictation without sending your voice anywhere.

Many free dictation websites pass your audio to a cloud speech service. This one runs Whisper, an open speech recognition model from OpenAI, on your own computer’s processor or graphics chip.

Private by design.

Audio and text stay in this tab. Nothing you say is stored on a server, and there is no account to tie it to.

No caps on use.

There is no daily quota or word limit. Record in chunks of up to 2 minutes, as often as you like.

English, or almost any language.

English mode is the most accurate for English. Any language mode recognizes 99 languages and detects each recording’s language, with lower accuracy.

Honest about its limits.

A model small enough for a browser makes more mistakes than large ones, especially with names, jargon and background noise. In English, “um” and “uh” are removed, sentences are capitalized and a closing period is added.

Questions.

Is this online dictation really free?
Yes. There is no account, no trial and no limit on how many recordings you make. Each recording can last up to 2 minutes. Yap’s paid product is its desktop app, which types into any app on your computer and is free for its first 3 days.
Is my voice uploaded?
No. Speech is recognized inside this browser tab, on your device. Your audio and your text are not sent to a server or saved by us. The site records usage events, such as that a recording finished, never your audio or what you said.
Why does it download a model first?
Recognizing speech on your device needs a speech model on your device. The first time, your browser downloads a Whisper model (about 80 MB, or about 165 MB where your browser uses the graphics chip) and the library that runs it, and keeps them in its cache for next time.
Does it work offline?
Once the model is in your browser’s cache, recognition itself needs no connection. You still need to open this page, and private browsing or clearing site data removes the cached model.
Which languages can I dictate in?
English mode uses Whisper’s English model, the most accurate choice for English. Any language mode uses the multilingual model, which recognizes 99 languages and detects the language of each recording. It is less accurate than the English model.
Can I say punctuation like “comma”?
Whisper adds commas, periods and question marks from the way you speak. Spoken words such as “comma” or “new line” are written as words here. You can edit the text directly, before or after you copy it.
Which browsers work?
A current desktop browser with microphone access. Browsers that let the model use your graphics chip, such as Chrome and Edge, recognize speech faster. Phones and older computers may be slow.

Dictating on your computer instead?

From the library

Dictation and Voice Typing: How to Talk Instead of Type

Dictation is speaking your words so a computer types them as you talk, and every current phone and computer does it for free with no voice training. Doing it well is a skill rather than a setting: compose each sentence before you say it, keep the microphone close, draft by voice and edit afterward, and proofread, because raw speech recognition still leaves errors. In a classic IBM study, novices learned to dictate letters as fast and as well as they wrote them within a few hours.