One of these is a transcription workbench. The other is a key.

These two apps share exactly one feature and are otherwise not competitors at all. MacWhisper is a transcription application: you give it audio — a file, a meeting, a YouTube URL, a podcast, a folder full of interviews — and it gives you text, subtitles and exports. Somewhere in that feature list is system-wide dictation. Coii VoiceInput is only the dictation, and it is $19.

So the useful question is not which app is better. It is which of those two jobs you are actually trying to do, and whether you are about to buy a workbench because you wanted a hotkey — or a hotkey when what you needed was the workbench.

What MacWhisper is

Checked on 4 September 2026, their site prices it at €0 free forever, with Pro at €64 once, including lifetime updates. The release their page talks about is MacWhisper 11.

The free tier alone is substantial. Their listing includes recording from the Mac microphone, mp3, wav, m4a, ogg, opus, mov and mp4, support for 100 different languages, export to .srt, .vtt and .txt, automatic speaker recognition, YouTube and media URL transcription, podcast transcription with a file per speaker, batch transcription, system-wide dictation, and meeting recording covering Zoom, Teams, Webex, Skype, Chime and Discord.

Pro adds .md, .pdf, .html and .docx export, grammar improvement for dictation, automatic transcript cleanup, translation, subtitle generation and translation, integrations with a long list of outside AI services, watched folders, workflow automation, and a command-line interface.

There is also a Mac App Store product from the same developer. Apple's listing shows Whisper Transcription by Good Snooze, free with in-app purchases, including a lifetime Pro at $99.99, requiring macOS 14 or later. If you are comparing prices, check which of the two you are looking at.

What Coii VoiceInput is

A hotkey. Hold ⌥ Space, say the sentence, let go, and the words are typed into the field you were already in, about half a second later. $19 once, three devices, macOS 13 Ventura or later on Apple Silicon and Intel, thirty days' trial with every feature and no card.

It does not open files. It has no export formats because there is nothing to export from — the text went into somebody else's application as you spoke. It has no batch anything, no subtitles, no speakers, no watched folders and no command line.

The table

Most of that table is a list of things this app does not do. That is not a weakness being conceded reluctantly; it is what a €64 transcription suite is for and what a $19 hotkey is not.

Where MacWhisper wins

Everything to do with recorded audio. Files, meetings, podcasts, URLs, batches, speakers, subtitles, watched folders. If any part of your week involves audio that already exists, this comparison is finished on their side and nothing below changes it.

The free tier is genuinely free and genuinely large. System-wide dictation is in it. Somebody who dictates occasionally can install MacWhisper, never pay, and be perfectly well served — and we have no answer to that at all beyond thirty days.

A hundred languages. Ours is several. If you write in Portuguese on Tuesdays this is not a close decision.

Exports and automation. Watched folders, workflow integrations, a CLI. This app has no seam you could automate against, because its output does not go to a file.

One app instead of two. If you want both dictation and transcription, that is one €64 purchase from them. From us it is two apps — this one and Coii AudioNotes for recording and summarising meetings — which is two menu bar items, two trials and two licences. That is a real cost and it belongs on this page rather than in a footnote.

Where Coii VoiceInput wins

The wins here are narrow and specific, because the overlap is narrow and specific. All of them are about what happens in the second and a half between letting go of the key and seeing text.

The sentence is saved before the engine is asked anything

When you release the key, the transcript is written to your history first, and the engine is asked afterwards. Not "and also saved", not "recovered on next launch" — first, as a matter of ordering.

This is the difference between two failure modes that sound similar and are not. The bad one is the sentence is gone: you said something, it did not arrive, and the version you rewrite is flatter than the version you had. The tolerable one is the sentence is in a list and did not land in the window: irritating, thirty seconds to fix, and no loss of the thing you actually wanted, which was the phrasing.

MacWhisper does not publish an ordering guarantee for its dictation, so this page will not tell you what it does under a failure. It tells you what ours does, and how to check it: dictate a long sentence, force-quit before the text appears, and open the history.

Dictation is the whole product, not a feature in it

A transcription application's dictation is a mode you enter. A dictation application's dictation is the entire surface, and the difference shows up in the small decisions.

The target window is fixed when the key goes down, not when it comes up — so clicking into another app mid-sentence, which everybody does while thinking, does not redirect your text. Where the accessibility API cannot place the words, the fallback is a synthetic paste and your clipboard is put back afterwards, so the app does not silently discard whatever you had copied. The pill shows a level meter, so you can see it is hearing you before you have said forty words into a muted microphone.

None of these are big features. They are the kind of thing that gets decided well when a hotkey is the only thing an app has to get right.

$19 once, and the trial is the whole app

€64 and $19 are both one-time prices in a category full of subscriptions, and the honest thing to say is that MacWhisper's price is reasonable for what it contains. It is more money because it is more software.

The difference worth noticing is the trial rather than the price. MacWhisper's free tier is a subset of the app; ours is thirty days of the entire app, including everything. Neither approach is better in the abstract — a permanent free tier is more generous, a full trial is a more accurate test — but they answer different questions, and the question you want answered is "does this work in my apps", not "does the free part work in my apps".

The history, and what it is not

It is worth being clear about what this is, because MacWhisper genuinely has a document library and this is not one. There are no folders, no tags, no search across projects and no export to Markdown. It is the record of what you dictated, kept locally so that the durability argument has somewhere to land.

If what you want is a searchable body of transcribed audio you can organise and work with, that is MacWhisper's shape and not this one's.

Running both

There is a perfectly sensible configuration where you own both, and it is probably the most common outcome for anyone who does interview work.

MacWhisper handles the recordings — the interview, the client call, the episode. Coii VoiceInput handles the typing you do all day, in Slack and Mail and a terminal, where what matters is that the key is instant and the words cannot be lost. They do not fight: they are separate apps with separate hotkeys, and the only rule is not to give them the same one.

If you do try both, give each a fortnight rather than an afternoon. The thing you are testing is not accuracy on a clean sentence in a text editor — every app in this category passes that test. It is what happens in the three applications you actually spend your day in, on a normal day, with background noise and a sentence you have not planned.

The applications that make this hard

Underneath, all of these tools are doing something awkward: inserting text into an application that only expects it from a keyboard. How well that goes is a property of the receiving app more than the sending one, and it is where nearly every support request in this category comes from.

Ordinary Mac text fields are the easy case. The hard ones are Electron apps, which draw their own text fields and expose them inconsistently; terminals, which have their own rules about what a paste means; and password fields, some of which deliberately refuse programmatic input — correctly, and not something any of these apps should be defeating.

This app's approach is the accessibility API first, a synthetic paste as a fallback, and the clipboard restored afterwards. It covers most of the hard cases and not all of them, and where it fails the words are still in the history.

The overlap, examined properly

Since system-wide dictation is the one square where these two apps sit on the same board, it is worth spending a section on what "system-wide dictation" actually means in each, because the phrase covers a lot of different behaviours.

In a transcription application, dictation is an added surface. The app's centre of gravity is a window with a waveform and a transcript in it, and the system-wide mode is a way of getting that same engine to type somewhere else. That is a genuinely useful thing to have, especially if you already have the app open for other reasons, and it is free in MacWhisper's tier list.

In a dictation-only application, there is no window to be the centre of gravity. Everything the app does happens in the two seconds around a keypress, which means every design decision in it is about that: how fast the key responds, what happens if you release it early, where the text goes if you moved, what the app does when the sentence is long, and what state your machine is left in afterwards. There is nothing else for it to be good at.

Neither of those is the correct approach in general. They are correct for different people. The way to tell which one you are is to look at how often you would press the key. Somebody dictating five or six times a day is well served by the mode inside the bigger app. Somebody dictating sixty times a day — which is what happens once the habit forms and short messages start going through it — is going to notice every one of those small decisions, sixty times.

What a day of sixty presses looks like

The reason the small decisions matter is volume, and volume is the thing nobody expects when they start.

Most people install a dictation app intending to write something long with it: a chapter, a report, a difficult email. That is almost never where it sticks. Long-form writing by voice is a skill, it takes weeks, and the first drafts are worse than what you would have typed. Plenty of people give up at that point and conclude dictation is not for them.

What actually sticks is the short stuff. The reply in a thread you have been ignoring. The commit message. The note you write to yourself at a traffic light. The two sentences of context you owe somebody before they can start work. These were never worth opening a keyboard for properly, so they got written badly or not at all, and speaking them takes four seconds.

At that volume, a half-second delay before the key is ready costs you thirty seconds a day and, more importantly, breaks the reflex — you start pressing early and losing your first word. A target window that resolves on release sends one message in twenty to the wrong place. A clipboard that is not restored destroys something you had copied about once a week, always at a bad moment. And a sentence lost to a stalled engine happens rarely enough to be forgettable and often enough to be infuriating.

That is the entire argument for buying a dictation-only app, and it is only an argument if you are going to press the key a lot. If you are not, the free dictation inside a transcription app you already own is the better answer and this page should say so.

What neither app does

Neither controls your Mac. They put text in; they do not delete a paragraph, switch applications or click a button because you asked out loud. That is Talon's territory and Talon does it properly.

Neither replaces reading what you wrote. Dictation gets punctuation approximately right and proper nouns approximately wrong — surnames, product names, anything with an unusual spelling — and that is true of every product in this category regardless of what it claims about accuracy. Anything going to a customer still wants a pass with your eyes.

And neither of them will do a good job in a noisy room with a laptop microphone. That is not a software problem and no vendor has solved it. If you dictate in an open office, the cheapest accuracy upgrade available to you is a headset, not a different app.

Which one to buy

Buy MacWhisper if any part of your work involves audio that already exists: interviews, recorded calls, podcasts, lectures, video you need subtitles for. Buy it also if you want a free tier rather than a trial, if you write in a language ours does not cover, or if you would rather have one app than two.

Buy Coii VoiceInput if dictation is the only thing you want and you would like it to be the only thing the app does: a key that is ready the instant you press it, a target window that does not move while you talk, a clipboard that comes back, and a history that has your sentence in it before anything downstream gets a chance to lose it.

Buy both if you record interviews and also answer forty messages a day. They are not alternatives to each other in that case; they are two different tools that happen to share a technology.

For the rest of the field: superwhisper is the other local Mac app with a large feature set, BetterDictation is the closest one-time-purchase dictation-only comparison, Wispr Flow is the subscription case, and Aqua Voice is the cloud one. If you have not tried it yet, Apple Dictation is free and already installed.

The pricing is one line and the download is thirty days with no card.