Windows 11 voice typing: how Win+H works and where it stops
Press the Windows key and H in any text box on Windows 11 and a small toolbar appears. Talk, and text shows up at your cursor. That is voice typing, and for a lot of people it is all the dictation they will ever need. This page explains how to set it up properly, what it can do, the three places it stops, and what to reach for when it does.
The short answers
Does Windows voice typing work offline? No. Microsoft's own page lists the requirements as "be connected to the internet, have a working microphone, and have your cursor in a text box", and states that voice typing "uses online speech recognition, powered by Azure Speech services". Your audio is sent to Microsoft's servers for transcription. If you turn Online speech recognition off in Settings, voice typing stops working; Microsoft's privacy page says that then "only device-based features (such as Narrator) are available".
Is my voice recorded? Microsoft says voice data "is sent to Microsoft only to provide the service and create text transcriptions" and that it does not "store, sample, or listen to voice recordings without your permission". Contributing voice clips to improve the service is a separate, optional setting.
Which languages? Around fifty, including all the large European ones, Chinese, Japanese, Korean, Hindi and Turkish. The full list is on Microsoft's voice typing page. Voice typing listens in the language of your current keyboard layout, one at a time.
Setting it up in two minutes
- Click into any text field, then press Windows key + H. On a touch keyboard, tap the microphone button instead.
- Click the gear on the toolbar and turn on automatic punctuation. It is off by default, which is why first attempts produce a wall of unpunctuated text. With it on you can still say "period", "comma" or "question mark" when you want to be explicit.
- In the same menu, decide about the profanity filter, which replaces swearing with asterisks when on.
- Optionally turn on the voice typing launcher, a floating microphone button that follows you between text fields so you do not need the shortcut.
- To dictate in another language, switch your keyboard layout with Windows key + Space first. Voice typing follows the layout, not what you are saying.
Useful commands once it is running: "delete that", "select that", "press Enter", "press Backspace", "new line", "new paragraph", and "stop listening" to end the session. The exact set varies by language.
What it does well
It is free, already installed, and needs no download. English accuracy in a quiet room with a decent microphone is good, and it handles ordinary business vocabulary without training. Because it is a system feature rather than an app, it works in almost any text box: browsers, Office, chat apps, the search bar. For dictating a paragraph into an email a few times a day, it is hard to argue with.
Where it stops
The internet, and the round trip. Everything you say travels to Microsoft and back. On a train, on a plane, on a locked-down office network, or simply when the connection is bad, voice typing degrades or refuses to start. There is no local fallback for voice typing; the on-device engine in Windows is used by other features, not this one.
One language at a time, tied to the keyboard. If you write in English and Polish in the same hour, you switch layouts every time. Mix two languages in one sentence, as bilingual households and immigrant families do constantly, and it transcribes the half it was told to expect. There is no automatic detection.
No memory of you. There is no custom vocabulary, so your surname, your product names and your clients' names are guessed fresh every time. There is no way to feed it a document, an audio file or a meeting recording; it only listens live. And the transcript exists only in the field you dictated into. Close the tab before copying and it is gone.
A fourth limit is smaller but bites daily: the toolbar is where recording starts and stops. You cannot bind it to a mouse button or a key you hold in your hand while walking around the room, and it does not pause your music while you talk.
When voice typing is not working
The same handful of causes explain nearly every report of a greyed-out Win+H toolbar or a mic icon that does nothing:
- Nothing has keyboard focus. Voice typing needs the cursor blinking in a text field before you press the shortcut.
- No internet, or Online speech recognition is switched off under Settings, Privacy & security, Speech.
- The keyboard layout is set to a language voice typing does not support.
- The microphone is disabled for desktop apps under Settings, Privacy & security, Microphone, or another app holds it exclusively.
- The window is running as administrator. Standard-user features such as the voice typing toolbar cannot type into elevated windows.
What to use when it is not enough
You want offline, or you want your audio to stay on the PC. Two paths. Voice Access, also built into Windows 11, downloads a language pack and runs on the device; it is designed for controlling the PC by voice and also takes dictation, in seven languages across fifteen regional variants. Or a local dictation app that runs a Whisper model on your graphics card, which is what AlphaDictate does: press a key of your choice to start, the same key to stop, text pasted where your cursor was, in 99 languages with automatic detection, and nothing uploaded. It is $39 a year after a free hour. Handy is a free open-source app in the same local category with fewer features.
You want the best possible accuracy and do not mind the cloud. Wispr Flow at roughly $15 a month is the polished option, with a large language model cleaning up the transcript after the fact. The audio leaves your machine, as with voice typing, but the results are better than Microsoft's.
You need to operate the computer by voice. Voice Access is the right tool, not voice typing. It clicks buttons by name, shows numbered overlays and corrects text by voice. If you came from Dragon, our note on what happened to Dragon covers that path.
We make one of the apps above, so weigh the recommendation accordingly. The honest summary is that voice typing is the right default for occasional dictation in one language with a good connection, and that its limits are architectural rather than bugs: it was built as an online service for a single keyboard language, and no setting changes that.
Frequently asked
Does Win+H work without internet? No. Voice typing uses Microsoft's online speech recognition and does not start offline. Voice Access and local dictation apps do work offline.
How do I turn on punctuation in Windows voice typing? Press Windows key + H, click the gear icon on the toolbar, and switch on automatic punctuation. It is off by default.
Does Windows voice typing send my voice to Microsoft? Yes. Audio is sent to Microsoft's servers to produce the transcript. Microsoft states it is used only to provide the service and is not stored or reviewed without your permission.
Why is voice typing greyed out or not working? Usually no text field has focus, there is no internet connection, Online speech recognition is turned off in Settings, or the keyboard layout is set to an unsupported language.
Can voice typing detect which language I am speaking? No. It listens in the language of the current keyboard layout. Switch layouts with Windows key + Space before dictating in another language.
What is the difference between voice typing and Voice Access? Voice typing types what you say into a text field and needs the internet. Voice Access controls the whole PC by voice, also takes dictation, runs on the device after a one-time language download, and supports fewer languages.