By Denys MedvedievComparison
Best Windows speech to text, ranked
The best speech to text app for Windows depends on where you type. Win+H is free and built in but needs the internet and only works in a text box. Whisper runs offline, works in every app, and pastes at the cursor.
Last updated: June 2026
The best speech to text app for Windows depends on where you type. Windows Voice Typing (Win+H) is free and built in, but needs the internet and only works in a text box. Whisper by Remskill runs offline, works in every app, and pastes at the cursor. It is the stronger pick for real writing.
Last weekend my younger daughter dictated a 90-word email to her grandmother in Ukraine in about 90 seconds. No keyboard. She is seven. Twenty-five years ago, the same machine in the same family took fifteen minutes to dictate one paragraph of a holiday letter, badly. Windows speech to text has quietly stopped being a science project. The honest answer is that you have more good options than you think, and one or two famous names you can skip.
Most "best speech to text for Windows" lists rank twelve tools and recommend whichever one paid for the link. That is not this. Windows in 2026 ships a free dictation feature most people never find, and the paid market has split into tools that record meetings and tools that help you write. This guide ranks the speech-to-text apps worth installing on Windows 10 or later, by four things that matter: accuracy, offline support, where it works, and price. Then it tells you, in plain terms, when the free built-in option is all you need. By the end you will know which one to install today, and which famous name to leave on the shelf. I read the support email for the app I build, so most of what I know about where these tools break comes from people writing in to tell me they broke.
A quick word on how to read what follows. I build one of the tools on this list, so I am not pretending to be a neutral referee, but I will name the cases where a free option or a competitor is the right call, because that is the only honest way to write this. The ranking is by use case, not by a single winner, because there is no single winner. The person who fires off three Slack messages a day and the person who drafts a 3,000-word report by voice want different tools, and a list that pretends otherwise is selling you something.
The short answer: Win+H is free, but here's where it stops
Windows · Win + H
Press the Windows logo key and H in any text box, and a small toolbar appears at the top of your screen. Start talking. That is Windows Voice Typing, and it is free and built into Windows 11. For a quick message in a chat window, it is hard to beat free and already-installed.
Two things stop it. First, it needs the internet. Voice Typing uses online speech recognition powered by Microsoft's Azure Speech services, so with no connection there is no dictation. Second, your cursor has to be in a text box for it to work at all. That rules out a lot of the places you want to talk: a file rename, a spreadsheet cell that fights you, a desktop app that Windows does not consider a "text box."
A smaller third thing shows up only after a few weeks. Win+H is a feature, not a workflow. It keeps no history of what you dictated, gives you no place to fix a word that always comes out wrong, and offers no way to bend the punctuation to how you talk. You get a microphone and a transcript and that is the whole deal. For a one-line reply that is exactly right. For the fourth email of the morning, you start to want more.
So the real question is not "is there a free option." There is, and you should try it first. The question is what to reach for when free runs out: on a train with no signal, in an app Win+H refuses to enter, or when the same job comes around for the fortieth time and you want it to remember.
How I picked: accuracy, offline, works-everywhere, price
I ranked these on four things, in this order, because this is the order they bite you.
Accuracy. Everything else is noise if the transcript is wrong. The catch is that nobody publishes a fair head-to-head number. Nuance advertises up to 99% recognition accuracy for Dragon from first use. In local mode, our own published range for Whisper is 95% to 99%, with the larger models scoring higher. Those are both self-reported, so treat them as claims, not a leaderboard. No primary source publishes a verifiable head-to-head accuracy percentage for Win+H, Google Docs Voice Typing, or Otter at all, so any list that hands you one for every tool is making it up. The boring truth is that a $20 USB microphone moves your accuracy more than any model swap. If you want to spend money on better transcription, spend it on the microphone first and the software second.
Offline. Can it work with the Wi-Fi off. Win+H cannot. Google Docs Voice Typing cannot. A tool that runs on your own machine can, which matters for a salary spreadsheet, a legal draft, or an email to your kid's school that has no business sitting in a vendor's logs. It also matters in the dumb, ordinary way: trains, planes, basements, the back seat of a car. Cloud dictation is only as reliable as the worst bar of signal you will hit today.
Works everywhere. Does it type into every app, or just one. This is the line that splits the field. A system-wide hotkey pastes wherever your cursor is; a browser feature only works in that browser tab. The test I use is mundane on purpose. Can it rename a file, fill a search box in a desktop app, or write a commit message. Most of the "free" options fail that test the moment you leave the browser, and you will leave the browser more than you think.
Price. Flat numbers, no "starting at." Free where free is real. I will not pad a ranking with a tool that costs more than a Windows license to sound thorough, and I will tell you when the free built-in option is all you need, which, for a lot of people, it is.
The Windows speech-to-text apps worth your time
Whisper by Remskill — best for writing in any app
Full disclosure: I build this one, so weigh that. What it does that the built-in option does not: it uses a system-wide global hotkey that works in any application where you can type, and pastes your transcribed text wherever the cursor is. The default Windows hotkey is Ctrl+Space: hold it, talk, let go, and the text lands.
It runs offline. All local transcription is pure-Rust and stays on your device; nothing leaves the machine once the model is downloaded. You pick how it runs. There is no single "recommended" model the app pushes on you, because the right pick depends on your machine and your languages.
Local Whisper ships eight models: four English-only, from Base at about 140 MB up to Medium at roughly 1.5 GB, and four multilingual that go up to Large v3 at about 3 GB. The multilingual ones cover 99 languages; the English-only .en builds do English and only English. The smaller models load fast and run on modest hardware; the larger ones want more RAM and reward you with cleaner transcripts. A second engine, NVIDIA's Parakeet at about 600 MB, runs 5 to 10 times faster than Whisper on a CPU and handles English plus 24 European languages, with no Asian languages and no translate-to-English. If you mostly speak English and want speed, Parakeet is the fast lane. If you need 99 languages, translation, or finer control, Whisper is the one to load.
If you want the latest OpenAI quality and web answers in the same tool, Cloud mode uses your own OpenAI key, so you pay OpenAI directly and we take no cut. You can even ask it to look something up out loud. Say the activation phrase "Hey whisper" and it will run a web search and paste the answer (I named the activation phrase after the product, which my wife pointed out is the kind of thing only a person who builds software would find clever). The whole thing is light: the app itself is about 25 MB, no dedicated GPU required, because local transcription runs on the CPU. It installs on Windows 10 or later. We walk through the first-run setup in detail in our voice to text on Windows guide.
Best for: anyone who writes by voice across more than one app and wants it to work on a flight.
Windows Voice Typing (Win+H) — best free, built-in
Already on your PC, no install, no account. Press Win+H, talk into a text box. The trade is the internet requirement and the text-box-only limit. For short messages where you are online anyway, it is the correct first choice precisely because it costs nothing. Microsoft has added automatic punctuation to Win+H, and it is right there in every fresh Windows 11 install. Most people who "don't have speech to text" already do and have never pressed the key combination. Start here. If it covers your day, you are finished and you spent nothing.
Best for: quick dictation in a browser or chat box, online, no setup.
Dragon Professional — best for specialized vocabulary
The old king of Windows dictation. Nuance still advertises up to 99% accuracy from first use, optimized for Windows 11 and backwards-compatible to Windows 10. If you dictate dense medical or legal terms all day, Dragon's custom vocabulary depth is its reason to exist, with decades of work behind handling jargon that trips up general-purpose models. One honest caveat: as of this writing the Nuance store was mid-migration and I could not confirm a current price, or whether the current version is a one-time purchase or a subscription, from a primary source. Historically Dragon was a one-time license, but I will not print a number I cannot stand behind, so check before you buy. The other caveat is the one I have an opinion about, and I will get to it below: the era of training your dictation software is over for almost everyone except this narrow professional case.
Best for: professionals with heavy domain jargon and a budget.
Wispr Flow — best polished cloud option
A direct competitor and a slick one. Wispr Flow advertises 100+ languages with automatic detection and runs natively on Windows, Mac, iOS, and Android. The auto-detect is the headline feature. It says it moves between your languages mid-sentence the way you do. I could not confirm from its homepage whether transcription runs on-device or in the cloud, so if offline matters to you, ask them directly before you commit. Polished, multi-platform, worth a look if cloud is fine for you and you want the same tool on your phone and laptop. We laid out the head-to-head in our Wispr Flow alternative piece if you are deciding between the two.
Best for: people who want a refined, cross-device experience and don't need offline.
Otter.ai — best for meetings, not dictation
Otter is excellent at the thing it does, which is not this. It is a meeting transcriber that joins Zoom, Microsoft Teams, and Google Meet, with a free Basic tier capped at 300 transcription minutes a month. It does not type into your apps; it records and transcribes conversations, then summarizes them. Different category entirely. If your problem is "summarize my meetings," Otter is a strong answer; if it is "write faster," it is the wrong tool no matter how good it is. It also runs as its own app and web service rather than a system-wide hotkey, so it is not something you reach for to fill in a form field. We dug into that line in our Otter.ai alternative piece.
Best for: capturing and summarizing meetings, not writing-by-voice.
Quick comparison: where each tool works
I left accuracy out of this table on purpose. Only Dragon and Whisper publish a number, and both are self-reported. Win+H, Google Docs Voice Typing, and Otter publish no verifiable figure at all. Any "Tool X is 96.3% accurate" leaderboard across all five is a guess formatted as a fact. So the columns below are the ones you can check: where it runs, whether it works offline, whether it reaches every app, and what it is built for.
| Tool | Pricing model | Local/Cloud | Works offline | Works in every app | Languages | Best for |
|---|---|---|---|---|---|---|
| Whisper by Remskill | Free local, paid cloud (see /pricing) | Local or cloud | Yes, local mode | Yes, system-wide hotkey | 99 on multilingual models | Writing by voice in any app |
| Win+H Voice Typing | Free, built in | Cloud (Azure) | No, needs internet | No, text box only | Multiple, set in Settings | Quick online dictation |
| Dragon Professional | Paid, check Nuance store | Check vendor | Check vendor | Within supported apps | Check vendor | Specialized vocabulary |
| Wispr Flow | Paid, see vendor | Unconfirmed | Unconfirmed | System-wide on Win/Mac | 100+ advertised | Polished cloud dictation |
| Otter.ai | Free tier, paid above (see vendor) | Cloud | No | No, its own app | Check vendor | Meeting transcription |
Read the table left to right and the field sorts itself. If "works in every app" and "offline" both need to be yes, the list is short. If you are happy online and inside one app, the free built-in options handle it and you can stop reading here. The decision is rarely about which tool is "best" in the abstract. It is about which row matches the way you work.
Other tools worth knowing
The Windows field is wider than five names. A few you will see on other lists, with one honest line each so you can skip the ones that don't fit:
- Microsoft Word Dictate — built into Microsoft 365 Word, free if you already pay for Office, and good inside Word. Stuck inside Word. If most of your writing happens there anyway, it is a fine no-extra-cost option. We covered its quirks in speech to text in Word.
- Google Docs Voice Typing — free, but only inside Google Docs and Slides in a supported browser, where the browser controls the speech service. Not system-wide, and the moment you tab away from Docs it stops.
- OpenAI's speech-to-text API — whisper-1, gpt-4o-transcribe, and gpt-4o-mini-transcribe, with a 25 MB upload limit per file. This is a developer building block, not an app you install. It is what tools like Whisper's Cloud mode call under the hood, not something you point at and start talking to.
- Speechnotes — a free browser dictation pad. Fine for jotting notes, lives in a tab, does not leave the tab.
- Windows Voice Access — Microsoft's hands-free OS control, built for accessibility and full keyboard-free operation rather than fast writing. Worth knowing it exists if you need to drive Windows by voice entirely.
- OpenAI Whisper — the open-source model itself, which you can run locally without the command line. If you specifically want the Whisper model on a PC, see our walkthrough of OpenAI Whisper for Windows for the no-terminal way to install and use it.
If you want the longer rundown of the whole category, including the meeting tools and the cloud transcription services, we wrote a full transcription software guide. The short version for this page: most of these are either browser-bound or category-mismatched for the job of writing into any Windows app.
When to skip a paid app (and just use Win+H)
Windows · Win + H
Here is the part most of these lists leave out, because sending you to a free competitor does not pay for anyone's blog. If you are online and you only dictate the occasional short message into a browser or chat box, do not pay for anything, including us. Press Win+H and move on. It is free, built into Windows 11, and handles short bursts fine. If you live in Google Docs, its Voice Typing is free and right there inside the document. If your real need is recording and summarizing meetings rather than writing, Otter's free tier gives you 300 minutes a month. A paid dictation app starts earning its place at a specific line: when you write longer pieces, work somewhere with no signal, or need to talk into apps that Win+H cannot enter. Below that line, the built-in option is the right answer, and I would rather you kept your money and came back when you outgrow it.
Here is the one strong opinion I will plant in this article: if your dictation tool needs you to train it, it's 1999. Modern speech recognition works out of the box on most accents and languages. The 45-minute calibration ritual that defined the old Windows era was a workaround for scarce compute, not a feature. In the late 1990s a relative of mine ran Dragon NaturallySpeaking on a Windows 98 desktop with 64 MB of RAM. The training process took 45 minutes. You read a list of words to calibrate it. Then dictation worked, barely, at maybe 70% accuracy with a four-second delay per sentence. One paragraph of a holiday letter took fifteen minutes. The headset got thrown across the room. The headset survived; the experiment did not. Twenty-five years later a seven-year-old in the same family dictated a full email in 90 seconds, no training, no calibration. If a 2026 tool still asks you to train it before it will listen, that is a 1999 idea wearing a new icon.
Pricing without the runaround
Whisper's entire local pipeline is free for anyone with an account: local Whisper, Parakeet, on-device AI enhancement, dictation history, custom hotkeys, the lot, with no payment method required to sign up. That is the part worth sitting with: the offline, works-in-every-app dictation that this whole article is about costs nothing, and you do not hand over a card to get it. Whisper Pro adds the Cloud surface for OpenAI-powered transcription, cloud AI enhancement, and voice web search, for people who want the latest models and web answers in the same tool. Win+H and Google Docs Voice Typing are free too. Otter has a free 300-minute-a-month tier and paid plans above it. Dragon is a paid product whose current price I could not confirm from Nuance directly, so check their store. The exact Whisper Pro numbers live on the pricing page, flat, in writing, no "starting at," and we email before any renewal so the date and amount are never a surprise.
Open a text box right now and press Win+H. If that quick toolbar covers everything you do, you are done, and you saved a download. The moment you find yourself on a train with no signal, or talking at an app that refuses to listen, or writing something longer than a text message, that is the line where a tool you install starts to matter. My seven-year-old crossed it the day she wanted to write to her grandmother from the back seat of the car, where the moon kept disappearing and the Wi-Fi never showed up at all.
Want to see it on your Windows PC?
Download Whisper, hold Ctrl+Space, and watch the transcript land in any app. The local pipeline is free, no card at signup.
Free local dictation for every signed-in user. Pro adds the cloud features on a separate trial.
