What is AI dictation, and how is it different from speech-to-text?

AI dictation is speech-to-text followed by an AI writing step. Classic speech-to-text turns your voice into the exact words you said; AI dictation then hands that raw transcript to a language model that removes fillers and false starts, adds punctuation and paragraphs, keeps names and numbers, and — if you ask — rewrites, summarizes or translates. The output is a text you would have typed, not a transcript you have to fix.

A concrete example

You say: "uhm hi Marco just wanted to tell you that the meeting tomorrow uhm we're moving it to three".

Speech-to-text writes exactly that. AI dictation writes: "Hi Marco, tomorrow's meeting has been moved to 3:00 PM. See you soon." — and with a "Write in: Italian" setting it writes "Ciao Marco, la riunione di domani è spostata alle 15:00. A presto."

What the AI step does — and does not do

• Removes "uhm", "like", repetitions, self-corrections ("at two, no, at three" → "at three").
• Adds punctuation, capital letters, line breaks; keeps lists as lists.
• Keeps your meaning and your register: it does not add content, and a good tool never cuts what you said.
• On request: rewrite (more formal, shorter), summarize, translate.

Where it runs

Most AI dictation apps (Loquil included) run transcription and cleanup in the cloud, so they need a connection and take a few seconds per dictation; some tools run local models on the computer, trading setup and hardware for offline use. In both cases the app runs alongside your programs and pastes the result where the cursor is.

Who uses it

Anyone who writes a lot and talks faster than they type: professionals answering email, lawyers and doctors drafting notes, support teams, people who write in a second language, people with RSI or dyslexia. See the FAQ or try Loquil free.

Questions

Updated: 2026-08-16

Related guides and comparisons