Dictation cleanup
Talking is faster than typing, but raw voice input comes out messy: "um", "uh", half-started sentences, repeated words, no punctuation. Paste it here and an AI model gives you back clean text — fillers and false starts gone, punctuation and paragraphs fixed, your words and meaning left alone. Handy if you draft briefs, outreach, or posts by voice.
What the cleanup looks like
A sample so you can see the kind of cleanup you get: fillers and false starts removed, punctuation and paragraphs fixed, your words kept. Illustrative example, not a live run.
um so basically the the idea is like we want to, we want to ship the feature you know by friday and uh then maybe do the announcement next week i think
So basically, the idea is we want to ship the feature by Friday, and then maybe do the announcement next week.
Clean your own text free. You get 1000 tokens (about 125 cleanups), no card.
Cleaned text
What it does, and what it won't touch
Speech-to-text is accurate on the words but keeps everything you said out loud, including the noise: the "ums", the sentence you started and abandoned, the word you said twice while thinking. This tool sends your text to an AI model with one job — remove that noise and fix the formatting, while keeping your actual words.
- Removes fillers: um, uh, er, like, you know, I mean, sort of, and the same in other languages.
- Drops false starts and repeats — abandoned fragments and "the the" stutters.
- Fixes formatting: punctuation, capitalization, sentence breaks, and paragraphs.
- Keeps your words. No paraphrasing, no summarizing, no reordering your points.
- Keeps your language. It won't translate — paste Ukrainian, get Ukrainian back.
It is an AI model, not a deterministic script, so it can occasionally drop a word it shouldn't or miss one it should. Read the cleaned text before you use it — it's a strong first pass, not a signed-off final.
How to use it
- Sign in and add tokens once. Each run makes a real AI call, so it runs on tokens (you start with 1000 free).
- Paste your raw text — straight from your phone's voice typing, a dictation app, or a transcript.
- Run the cleanup and read the result. It shows a one-line note of what it changed.
- Copy it out and drop it into your doc, email, or post.
Frequently asked questions
What does it remove?
Filler words and sounds (um, uh, like, you know, I mean, and the Russian/Ukrainian equivalents), false starts and abandoned fragments, and stutters or repeated words. It also fixes punctuation, capitalization, and paragraph breaks. It does not touch your actual wording beyond removing that dictation noise.
Does it change my words or meaning?
No. It does faithful cleanup, not rewriting. It will not paraphrase, summarize, translate, reorder your points, or invent anything. If a sentence is a little rough but clear, it leaves your wording and only fixes the punctuation. It's an AI model, so it can occasionally slip — read the result before you use it.
Does it translate the text?
No. It keeps your original language. Paste English, Russian, Ukrainian, or another language and it cleans the text in that same language.
Is my text private?
Your text is sent to our server and our AI provider only to run the cleanup, and we don't store it after the request. Don't paste anything you wouldn't send to a third-party AI service.
Why does this need sign-in and cost tokens?
Each cleanup makes a real AI model call, which costs us money, so unlike a static tool it runs on tokens after a one-time top-up. You get 1000 free tokens on signup, which covers a lot of cleanups, and the top-up requirement keeps the tool from being drained by bots.
How long can the text be?
Up to 6000 characters per run, which is roughly 1000 words. For a longer transcript, clean it in chunks.
Last updated: 2026-05-25