Many dictation apps change what you said, deliberately and by default. They remove filler words, add punctuation you did not speak, correct grammar, drop mid-sentence false starts, and in several cases adjust tone to suit where the text is going. This is advertised as a feature rather than hidden, but it is rarely presented as a choice, and the result is that most people do not know which kind of app they are running.
The two jobs
Transcription turns audio into the words that were in it. The measure of a good one is word error rate — how often it heard something other than what was said.
Editing takes those words and produces different words that read better. Filler removal, grammar correction, restructuring and tone adjustment are all editing. The measure of a good one is subjective, which is exactly why it is hard to evaluate and easy to ship.
An app that does both well is doing two things you can only judge separately. An app that does both and stores only the result has thrown away the evidence you would need to tell them apart.
Why it matters more than it sounds
Transcription errors are visible: you said one word and a different word is on screen, and you fix it. Editing is invisible, because the output is well-formed and plausible. The sentence you get back reads better than the one you said, which is precisely why nobody re-reads it carefully.
The cost lands where phrasing carries meaning. Hedges removed from a clinical note or a legal memo turn an estimate into a claim. A prompt tidied before it reaches an AI assistant is a different instruction. A quote that has been grammar-corrected is no longer a quote. And for anyone whose writing has a voice, the polished version is reliably less like them than the rough one.
What to look for in a feature list
The vocabulary is fairly consistent across the category. Any of these describes editing rather than transcription:
- cleanup, auto cleanup, polish
- removes filler words — sometimes stated as a benefit with no switch
- formats as you speak, formatting that suits the destination
- tone, style, register, formal / casual
- grammar improvement, corrects grammar
- restructures, turns rambling notes into
- modes for email, messages, or documents
None of these is a bad thing to want. They are simply a different product from one that types what you said, and the distinction is not usually drawn on the pricing page.
The four-minute test
I think we should probably ship on Thursday — actually, Friday — assuming the tests pass.
Dictate exactly that sentence into whatever you use now. Then say one with deliberate fillers, and one sixty-word run-on with no spoken punctuation. If the hedges vanish, the false start disappears, the ums are gone and the run-on comes back as tidy sentences, you have an editor. If it all survives, you have a transcriber.
Then check the harder thing: does the app still have what it originally heard? If it stores only the edited version, an edit you disagree with is not recoverable.
Which one Coii VoiceInput is
A transcriber. Its post-processing is a fixed chain of text filters — repeat collapsing, your own dictionary, optional filler removal, punctuation and spacing — and no model is asked to improve the sentence. Filler removal ships off. Both the raw transcript and the processed one are kept on every history row, so a filter can be wrong without costing you the words.
That is a trade rather than a win: if you want a clean formatted draft without editing it yourself, an app that rewrites is doing work this one will not do for you. The longer argument, including who should buy which, is in dictation that doesn't rewrite your words, and the head-to-head against the most widely used editing tool is the Wispr Flow comparison.