Dictation software: how to choose one
Dictation software turns what you say into typed text on your computer. This page is about choosing one — what the category actually contains now, what separates the options, and how to test them before you pay.
Written by the people building one of them, so read it that way. VoiceSnap Pro appears in the table below in the same format as everything else, marked as what it is: unreleased. Where a competitor is better for your job, the comparison pages say so.
What dictation software actually is
Dictation software listens to your voice and writes down what you said. The useful modern version does it system-wide: you put the cursor where the text belongs — a reply in your mail client, a Slack thread, a commit message, a form in a browser tab — hold a key, speak, and the finished sentence appears in that field. There is no separate window to dictate into and no transcript to go and copy out of afterwards.
That is worth naming, because a lot of software filed under "dictation" is not that. Meeting transcribers record a conversation and hand you a document later. Speech-to-text models and APIs turn an audio file into a string and leave the rest to you. Both are useful. Neither types for you while you are working, which is the thing most people mean when they go looking for dictation software.
How AI dictation differs from what you remember
If you last tried dictation in the Dragon era, you learned a command language. You said “comma”, “new paragraph”, “cap that”. You trained the software on your voice before it was any use, and it was unforgiving about words it had never heard.
Modern AI dictation transcribes a whole utterance at once and works out from context what you meant — where the sentence ended, whether that was a question, which of two identical-sounding words fits. Punctuation arrives without being spoken. Restarts and “ums” can be stripped on the way through. The practical difference is that you can talk the way you talk, instead of performing a dictation.
It is not magic. Everything in this category still degrades with a poor microphone, background noise, an accent the model has heard little of, or vocabulary nobody has taught it. Any product that implies otherwise — ours included — should be tested on your own voice before you believe it.
What to look for
These six questions separate the options far more than any headline accuracy figure will. Ask all six of anything you are considering, including the free ones.
System-wide insertion, or a window you copy out of
The biggest single difference in how a dictation tool feels to use. System-wide means the text lands in whatever field already has your cursor. The alternative is an app you dictate into and then copy from, which adds two steps to every thought and is the reason most people quietly stop using it.
How to check Dictate into somewhere awkward — a browser form, a terminal, a code comment. If the words appear there, it is system-wide.
Punctuation and paragraphs you do not have to speak
Some tools still expect you to say the punctuation out loud. Some insert it for you and get commas roughly right. Few break paragraphs where you paused. Whatever it does not do, you will do by hand afterwards — and that cleanup pass is where the time you saved goes.
How to check Dictate three sentences and a question at normal speaking pace without saying a single punctuation mark, then read back what you got.
A vocabulary you can teach
Every job has words the model has not met: colleagues' names, an internal tool, an acronym your industry uses and nobody else does. If there is no editable word list, you will be correcting the same five words for as long as you own the software.
How to check Look in the settings for a word list you can add to, not just a general claim about accuracy.
Licence model
This category prices in three shapes: free with the operating system, a subscription, or a one-time purchase. None of them is automatically the right one. A subscription funds continuous development and is cheap to walk away from. A one-time purchase stops costing money and cannot be repriced from under you. Free is free, and for short bursts it is often enough.
How to check Work out what two years costs rather than what one month costs, then ask honestly whether you will still be dictating in two years.
Privacy and data retention
Most cloud dictation sends your audio to a server to be transcribed, which for many people is fine and for some is disqualifying. What varies is what happens next: how long the audio and the text are kept, whether either is used to train models, and whether you can delete them. Running the model on your own machine sidesteps the question, usually at some cost in accuracy or speed.
How to check Find the sentence in the vendor's privacy policy about training and retention. If there is not one, that is your answer.
Whether it works where you actually work
Platform, offline behaviour, and the specific applications you live in. A Mac-only tool is no help on the work PC. A cloud tool is no help on a plane. And some apps — remote desktops, certain secure fields, a few Electron editors — refuse injected text no matter what any vendor claims.
How to check Trial it in the three apps you use most, on the machine you use most, and do it before the refund window closes.
Dictation software on Windows
Windows 11 has voice typing built in. Press Win + H, a small bar appears, and what you say is typed into the field you were already in. It costs nothing, it is already installed, and for a message or a search box it is genuinely fine.
Its limits are the ones you would expect from something free with the OS: it punctuates, but paragraph breaks need saying out loud; there is no word list you can edit; one language is selected at a time; and nothing is kept — the text is typed and then it is gone. Paid Windows dictation competes almost entirely on those points, from the professional suites built for medical and legal work at one end to the newer AI dictation apps at the other.
If you specifically want to know whether Win + H is enough for you, the head-to-head below goes through it properly rather than repeating it here.
Dictation software on Mac
macOS ships with dictation too, on a keyboard shortcut you can rebind. On Apple silicon it runs on the device, so it keeps working with no network at all — something no cloud dictation tool can match, at any price.
The shape of its limits matches Windows': one language at a time, no vocabulary you can teach it, nothing kept afterwards. Mac users do have more third-party choice than Windows users, though, and it is the platform where local-model dictation apps are most established — worth knowing if sending audio to a server is the part you object to.
Two of the guides on this site cover the built-in route end to end, if what you actually need is to get dictation working on the Mac in front of you today.
How the main options compare
Every product we have written a full head-to-head against, with its licence model in words rather than a figure — prices change constantly and a stale number on a page like this is worse than none at all. Follow a row for the detailed version, including what that product does better than ours.
| Product | Licence model | Available today | Comparison |
|---|---|---|---|
| Apple DictationThe dictation built into macOS, iOS and iPadOS — free with the operating system. | Free with the operating system | Yes, already on your Mac | Read the comparison |
| Windows Voice TypingThe voice typing built into Windows 11 — press Win + H and start speaking. | Free with Windows | Yes, built into Windows 11 | Read the comparison |
| DragonThe long-established professional dictation suite, strongest in medical, legal and accessibility work. | Subscription or per-seat licensing in current editions | Yes | Read the comparison |
| Otter.aiMeeting transcription and notes — it records conversations rather than typing for you. | Free tier plus subscription plans | Yes | Read the comparison |
| Wispr FlowA shipped, actively developed dictation app that types into any application on Mac and Windows. | Subscription-priced, with a free tier | Yes | Read the comparison |
| superwhisperA Mac-first dictation app that runs speech models locally on your own machine. | Subscription, with a one-time lifetime option | Yes | Read the comparison |
| OpenAI WhisperOpenAI's open-source speech recognition model — free, self-hostable, and not an application. | Free and open source | Yes | Read the comparison |
| Google Voice TypingVoice typing in Google Docs through Chrome, and Gboard's voice typing on Android. | Free with a Google account | Yes | Read the comparison |
| VoiceSnap ProThe app this site is about — system-wide dictation for macOS and Windows, sold once. | One-time purchase, no subscription | No — waitlist only | What VoiceSnap Pro is |
Last row on purpose. Everything above it is software you can install this afternoon. Ours is not, and putting an unreleased product at the top of a list it has not earned would be exactly the thing this page is meant to help you see through.
Where VoiceSnap Pro fits
Plainly: it has not shipped. There is no binary, no trial and no user you could ask. If you need dictation working today, install one of the products in the table — several of them are free, and the comparison pages will tell you which one suits you.
What it is being built to be is a system-wide dictation app for macOS and Windows that finishes the text rather than merely transcribing it: punctuation and paragraphs without speaking them, filler words removed, a vocabulary you can teach it, and every dictation kept in a library you can search months later. It is sold once, for $39, with no subscription. That part of the design is settled and is not going to change.
The rest is a promise until you can run it yourself, which is the honest reason this page recommends other people's software above.
Try dictating before you pay for anything
The quickest way to find out whether you think well by talking is to talk. These run in your browser, free, with nothing to install — and where they recognise speech, they use your browser's own speech service rather than ours.
Further reading
Longer pieces on individual corners of this, in the guides and on the blog.
Questions people ask about dictation software
There is no single answer, and anyone who gives you one is selling something — this page included, so weigh it accordingly. The useful version: if you dictate a sentence or two at a time and want it free, the dictation already in macOS and Windows is hard to beat. If you dictate for hours in a specialist vocabulary, the professional suites still lead. If you write all day in ordinary language and want the text to arrive finished rather than raw, that is what the newer AI dictation apps are for. Ours is in that last group and has not shipped yet.
For most people, in a quiet room, with a decent microphone — yes, and it has been for a few years. Treat any accuracy figure a vendor quotes as a ceiling rather than a promise: those numbers are measured on clean recordings of general speech. Accents the model has heard little of, background noise and specialist vocabulary all pull it down, which is why every recommendation here ends with testing it on your own voice.
No. Windows 11 and macOS both include dictation that types into any field, at no cost, and for a lot of people that is the end of the question. Paid tools compete on what happens to the text after the words are recognised: punctuation, paragraph breaks, filler-word removal, a vocabulary you can teach, a searchable record of everything you dictated. If none of that is worth money to you, use the built-in one.
Direction and timing. Dictation is live, and it is for text you are composing now — it types as you speak, into the app you are already in. Transcription is after the fact, and it is for a recording someone else made: a meeting, an interview, a lecture. It produces a document you go and read. Otter and Whisper sit on the transcription side of that line, and the comparison pages cover both.
It depends on the product, and it is worth checking rather than assuming. Most cloud dictation uploads your audio to be transcribed. The questions that decide it are how long the audio and the transcript are kept, whether either is used to train models, and whether you can delete them. Tools that run the speech model on your own machine avoid the question altogether, usually trading some accuracy or speed to do it.
Some of it, not all. Built-in dictation stops at the operating system boundary by definition, and several of the better third-party apps are Mac-first. If you move between a Mac and a PC, check platform support before anything else on this page — it rules out more options than any other single criterion.
If the one-time purchase is the part you came for
VoiceSnap Pro is a single $39 purchase with no subscription, for macOS and Windows. It is not finished. Leave an email address and you will get exactly one message, on the day it is available — no newsletter, no countdown.
One launch email. No newsletter, no sharing your address. Privacy policy