Dictation that types while you are still talking
Why watching the words arrive changes dictation more than accuracy does — and how to set up live typing, including what happens when the connection drops mid-sentence.
Step by step
- Turn on live typingSettings → Dictation → “Type as I speak”. Each phrase lands at your cursor the moment you finish saying it, instead of everything arriving at the end.
- Watch the preview while you talkThe recorder shows two lines: what has already been typed, and the model’s current guess at what you are saying now. The guess is never typed — it settles first.
- Move the cursor mid-sentenceDelete a word, click somewhere else, keep talking: the next phrase goes to the new place, because it is typed at the cursor rather than pasted as a block.
- Stop by stoppingThree seconds of silence ends the run. No button to reach for, which matters when your hands are already somewhere else.
- Teach it your wordsSettings → Vocabulary holds the names and jargon a general model has never heard. They are sent to the transcriber itself, so they are recognised rather than repaired afterwards.
Turn on live typing and watch the sentence build
Dictation is free on every plan, streaming included.
The preview is an error-correction tool, not a decoration
The value of seeing the words is not that it looks impressive. It is that dictation errors compound: a name transcribed wrongly in sentence two is usually transcribed the same way in sentences five and nine, and finding that at the end means re-reading everything.
The other half is that people talk differently when they can see what landed. Sentences get finished. A phrase that came out badly gets said again immediately rather than fixed in an editing pass that never happens.
See the words as you say them
Download for free — nothing to configure before the first run.
Frequently asked
How is this different from Apple or Windows dictation?
Those transcribe; this one also punctuates, capitalises to fit the sentence it is landing in, and can offer a tightened version before it types. It also switches language mid-sentence, which built-in dictation does not.
Does the preview cost extra?
No. Streaming and batch are billed the same way — by the length of the recording, not by how it was delivered.
What happens if the connection drops?
The recording keeps going locally and is transcribed in one piece when you stop, so a dropped socket costs you the live preview and nothing else.
Which operating systems does aiCat support?
macOS (Apple Silicon and Intel), Windows 10/11, and Linux as AppImage, .deb or .rpm.
How much does aiCat cost?
Free to start, with 10 aiCatCoins a month and everything aiCat does included. Premium starts from $6 a month (40% off the $10 list price) and adds Think More, read-aloud, encrypted memory, 5000-character selections and 3000 coins a month. Team + Offline starts from $18 per member and adds local models, a memory you can edit by hand, and MCP access from your other tools. Coins can also be bought outright instead of subscribing, and they do not expire.