2. Dragon Professional, the accuracy benchmark that is now coasting

Verdict first: Dragon still earns its accuracy reputation, and it is a hard sell for a new buyer in 2026 anyway. It has been the name in serious dictation for years, its accuracy after training is the bar others are measured against, and it runs offline with strong voice commands for long-form work. I did not run a fresh test of it here, so I am going by the current docs and the consumer buyer's experience rather than pretending otherwise. What the docs and user reports agree on is the friction: hours of reading training passages before it settles in, a high one-time price, and, most telling, no major mainstream consumer release in roughly three years while Nuance steers the brand toward healthcare enterprise and retired the standalone Dragon Home edition. You would be paying a premium for software in maintenance mode.

Verdict: Dragon still leads on trained accuracy, but the stalled development and the high price make it a tough recommendation for new buyers. Our full Dragon NaturallySpeaking review walks through the current price and what Nuance did with the brand.
My take: The reputation is real and, on the current lineup, so is the coasting. For a professional who already owns and lives in it, it is still excellent. For someone buying fresh, I would not spend that money on a product this quiet.

3. Wispr Flow, slick AI cleanup with a privacy asterisk

Wispr Flow is the one that feels the most modern, and it is also the one that gave me the most pause. It leans hard into AI: as you talk across your apps on Windows and Mac, it strips filler words and adjusts tone, and for quick messages it is genuinely pleasant. Then you read the history. A developer documented the app capturing screenshots and audio and sending them to the cloud without clear disclosure, and the company's first move was to ban him before the CTO apologized. It needs an internet connection to work at all, because the whole engine is remote, and reliability complaints after the trial period are a recurring theme. The polish is real. So is the fact that your voice is someone else's traffic.

Verdict: Capable and modern, and hard to recommend for sensitive work given the cloud dependency and the privacy history.
My take: The cleanup impressed me and the backstory stopped me cold. For a throwaway message it is fine. For anything I would not post in public, I do not want it as the pipe my voice runs through.

4. Windows Voice Typing (Win+H), free, built in, and cloud-bound

Free and already on your machine, so why buy anything? That is the fair question, and Win+H answers part of it. Press Win+H in any text field on Windows 11 and you get dictation with nothing to install and no account to create. For a short note it does the job. The ceiling shows up fast once you lean on it. It needs an internet connection, and users report a long-running "0x800704cf" error that falsely claims you have no connection. There is no custom vocabulary, so technical and medical terms come out mangled with no way to teach them, and the toolbar does not follow you cleanly as you move between apps. If most of your dictation happens in one editor, our walkthrough of voice to text in Word covers the built-in route there.

Verdict: A fine free starting point, and the cloud requirement plus missing vocabulary push serious users elsewhere.
My take: I dictated a quick paragraph and it was fine, then hit the wall the moment I tried a product name it had never seen and could not be taught. Good for a one-line note, thin for a working day. We break down where the ceiling sits in our Windows voice typing review.

5. superwhisper, on-device dictation that is Mac-first

superwhisper is the closest thing to our own pitch, with one catch for this list: it grew up on the Mac. It runs Whisper-family models locally, so a Mac user gets on-device dictation with the audio kept on the machine, a lifetime option alongside the subscription, and a clean system-wide paste. On Windows the story is thinner, since the Windows build is newer and still catching up to the Mac app I actually tested. I include it because it is the honest on-device answer for anyone reading this on a Mac, and because it shows the offline approach is not ours alone. If you are on Windows, though, it is not the finished tool that the Mac version is yet.

Verdict: The on-device pick for Mac users, and a work-in-progress on Windows compared with the Mac app.
My take: On the Mac it did quietly what I want dictation to do, keep the audio local and paste anywhere. On Windows I would wait for the newer build to mature before betting a workday on it.

6. Otter.ai, excellent for meetings but not for drafting

One line settles Otter: it is a meeting tool, not a dictation tool. It transcribes calls, tags who spoke, and plugs into Zoom and MS Teams, with a usable free tier, and it is genuinely good at that. Sit down to dictate a document into your own apps, though, and it fights you, because it is built around recordings and it runs entirely in the cloud. I tried using it to draft rather than to capture a call, and it was the wrong shape for the job. If capturing the call itself is the real goal, our guide on how to record a Teams meeting as a participant handles that separately.

Verdict: Reach for Otter when you need notes from a Zoom call, and reach for a real dictation tool when you need to write.
My take: Brilliant at the job it was built for, and the wrong tool the second I tried to draft an email with it.

7. Google Docs Voice Typing, free if you live in Google Docs

I opened a blank Google Doc, clicked Voice Typing, and it worked well enough to be tempting, right up until I tried to use it anywhere else. That is the whole story. Google's Voice Typing is free and surprisingly usable, but it only runs inside Google Docs, with no desktop app, no offline mode, and no system-wide capability. It cannot type into your email client or a Word document. For drafting straight into a Doc it is a no-cost choice. For anything outside Docs it is a dead end.

Verdict: A fine free tool if your writing already lives in Google Docs, and useless outside it.
My take: Pleasant inside the Doc, a brick wall the moment I wanted the text in Word instead.

8. Talon, the hands-free option built for RSI and coding

Talon caught me off guard, because it is not really a dictation app at all, it is a full hands-free control layer. It was built for people who cannot use a keyboard comfortably, and it turns your voice into both text and commands, so you can navigate, edit, and even write code without touching the keys. It runs on Windows, Mac, and Linux, keeps its recognition local, and is free to start with a paid beta for the better engine. The trade is a steep learning curve, since you are learning a command grammar, not just talking. For someone with RSI or an accessibility need, that investment is exactly the point.

Verdict: The pick when hands-free control matters more than casual dictation, especially for coding or an accessibility need.
My take: I expected a dictation tool and found a control system. For plain prose it is more than most people need. For someone who cannot type, it is a different category of useful.

Accuracy and recognition quality

Raw word accuracy across the best voice dictation software is high enough now that it is rarely the sole deciding factor. Where tools separate is the messy stuff. Punctuation is a frequent complaint. On a long Apple Community thread, users describe dictation writing "question mark" as literal words and capitalizing at random, and the same class of problem shows up on Windows. Custom vocabulary matters just as much. If you dictate medical terms, code identifiers, or unusual names, a tool that lets you teach it those words beats a more accurate engine with no way to learn them, which is why the built-in options frustrate technical users. As for the raw engine, the model behind our own tool gives a useful anchor for what "accurate out of the box" means in 2026.

6.34% WER

Source

Offline and privacy: where your voice goes

This is the dividing line that splits the list in half. Six of the eight run in the cloud. Windows Voice Typing and Google Docs, Wispr Flow and Otter, and the Windows side of the newer cloud players all send your speech off to a server, which means your words travel to a third party. Sometimes those words are a contract clause or a patient note. One offline-focused review puts it plainly: most cloud dictation services store recordings, sometimes indefinitely, to train their models. For a professional dictating sensitive material every day, that is real exposure, not a hypothetical. Only three tools here keep recognition on the machine itself: our own SayOnce and Dragon on Windows, and superwhisper on the Mac. If privacy is part of your job, treat on-device as a hard requirement, not a nice-to-have.

Subscription versus one-time purchase

The pricing split tracks the privacy split, and it is the tension buyers raise most often. Cloud tools like Wispr Flow run on subscriptions, and even a modest monthly plan adds up to real money by December, every year. A one-time license flips that math: you pay once and own it. Dragon takes the one-time route but at a high price and with stalled updates. Our own SayOnce sits in the affordable one-time slot, which is the niche the subscription crowd has largely abandoned on Windows. We are not going to quote exact prices here, since they move and our own price lives on the product page. What is stable is the model. If you dictate regularly, do the annual math before you sign up for a monthly plan, because the cheaper-looking option is often the more expensive one by the end of the year.