By Denys MedvedievTutorial
Voice typing in Word: Win+H, Dictate, or better
Two built-in ways to dictate in Word, one system-wide upgrade. Where each lives, how to turn it on, and where they stop.
Last updated: June 2026
Voice typing in Word works two ways out of the box: press Win+H to open Windows voice typing in any text field, or click the Dictate button on Word's Home tab. Both need an internet connection to transcribe. A local tool such as Whisper adds offline dictation that works in Word and every other app.
A 700-word status report took me 19 minutes to type last month and a little under 5 to dictate at 145 words per minute. Same document, same Word window, same author. The only thing that changed was which input Microsoft's most famous app was listening to. Word has had ears for years; most people just never found the two buttons that turn them on. (I found them the embarrassing way, by pressing Win+H to mute a call.)
Here is the problem: Word users find one of the two built-in dictation tools, assume it is the only one, and judge voice typing by whichever they hit first. Right now Windows 11 ships voice typing on Win+H and Microsoft 365 ships Dictate inside Word itself, and they are different tools with different requirements. This guide covers turning on both inside Word, where each one stops, and how a system-wide tool covers what they miss.
By the end you will have voice typing running in Word in under a minute, plus an offline setup if you need one. Most of the dictation questions in our support inbox boil down to "which of these am I using". I read those emails, so I can say that with a straight face.
Word has two voice typing paths, not one
Microsoft built both of them, which is why nobody can tell them apart. Windows voice typing is part of the operating system. It opens with Win+H, floats above whatever app has focus, and types wherever the cursor sits. It works in Word the same way it works in Notepad or a browser form. Dictate is part of Microsoft 365. It lives on Word's Home tab, only exists inside Office apps, and needs an active Microsoft 365 subscription.
The boring truth: for plain paragraph dictation in Word, they produce near-identical results, because both route your audio through Microsoft's online speech services. The differences that matter are scope and requirements. Win+H comes with Windows and follows you across every app. Dictate is tied to your Microsoft 365 plan and knows Word-specific tricks like formatting commands. If you have a Microsoft 365 subscription and live inside Word all day, start with Dictate. If you write in Word, Outlook, Slack and a browser, start with Win+H. Neither costs extra on top of what you already have.
Win+H: voice typing that lives in Windows, not in Word
What you need before step 1: a Windows 11 PC (Windows 10 has the older dictation toolbar on the same shortcut), a working microphone, and an internet connection, because Windows voice typing transcribes through Microsoft's online speech recognition. No connection means no typing. Total setup time is under a minute.
1. Open your document and click where the text should go
Voice typing needs the cursor in a text box; with nothing focused, the pill opens but nothing types.
Check: the cursor is blinking inside your Word page.
2. Press Win+H
The voice typing pill appears near the top or bottom of the screen.
Check: it shows "Listening…" before you start talking.
3. Talk in phrases, not single words
Full sentences come out cleaner than word-by-word dictation.
4. Turn on automatic punctuation
Click the gear on the pill and enable auto punctuation, or speak it: "period", "comma", "question mark".
Check: ending a sentence and pausing produces a period without you saying one.
5. Stop with your voice or the same shortcut
Say "stop listening", or press Win+H again.
Win+H also takes editing commands: "delete that", "select that", "press Enter". They work, with the patience of someone negotiating homework: clear wins on simple requests, occasional surprises on compound ones. For longer corrections, keyboard and mouse remain undefeated. If the pill opens but nothing happens, our Win+H not working fixes cover the usual suspects, and Microsoft's voice typing documentation lists every command and the 40+ supported languages.
One more wrinkle: on Copilot+ PCs, a feature called fluid dictation cleans up grammar and spelling as you talk, and switches itself off in password fields.
The Dictate button, briefly
Dictate is the microphone button on Word's Home tab. Click it, wait for it to show it is listening, talk. It needs a Microsoft 365 subscription, a microphone and a reliable internet connection, and it works across Word for Windows, Mac, the web and mobile. On a Mac, Option+F1 toggles it. If you write in Word on a Mac all day, Dictate plus that one shortcut may be all the voice typing you need.
Its advantage over Win+H is Word-awareness. Auto-punctuation is available in most supported languages (a toggle in its settings), and it understands formatting commands: "bold that", "create bullet list", "delete last word". Its weakness is reach. The moment you leave Office, Dictate stays behind. I keep this section short on purpose. We have a full guide to Word's built-in Dictate feature covering every command, language and setting, and Microsoft's Dictate documentation is the canonical reference. If Dictate is greyed out or missing, the Word dictation troubleshooting guide walks through the fixes.
Where both built-ins hit a wall
Three limits show up in practice, ranked by how often they bite.
First, neither works offline. Win+H transcribes through Microsoft's online services; Dictate states a reliable internet connection as a requirement. On a train, on a plane, on hotel Wi-Fi that drops every 40 seconds, both go quiet. Between you and me: dictation that cannot work offline is also a privacy decision someone else made for you. The email to your kid's school, the salary review, the legal draft — all of it routes through a vendor's servers because you wanted to type with your voice. Local transcription sends nothing anywhere, and for some documents that is not a preference, it is the requirement.
Second, the language ceilings differ. Win+H supports 40+ languages. Dictate supports about 15 languages and regional variants in full, with about 25 more in preview at lower accuracy. Dictating in Ukrainian or mixing languages mid-document gets uneven fast.
Third, scope. Dictate ends at Office's borders. Win+H goes everywhere but stays a Windows feature; there is no Mac equivalent of it. And I never found a place to teach either tool the project names, client names or product jargon I use every day — "Remskill" arrives as "rem skill" no matter how many times I correct it.
One hotkey that types in Word and everywhere else
In the late 1990s a relative of mine ran Dragon NaturallySpeaking on a Windows 98 desktop with 64 MB of RAM. Training it took 45 minutes of reading calibration text, dictation crawled with a 4-second delay per sentence, and one paragraph of a holiday letter took 15 minutes. The headset got thrown across the room. The headset survived; the experiment did not. I filed voice typing under "ideas whose time has not come" and did not touch the file for twenty years. This section is the reopened file.
Whisper is a desktop app that puts one hotkey on top of the whole system. Press Ctrl+Space in Word (or anywhere else you can type), talk, release, and the text lands at your cursor. The same key works in Outlook, Slack, a browser, a code editor. There is no toolbar to find and no per-app feature to enable, which is the structural difference from both built-ins.
Setup takes about ten minutes, most of it a download:
1. Install the app
It runs on Windows 10 or later and macOS 11 or later, and weighs about 25 MB.
Check: the Whisper window opens to Settings.
2. Sign in
The local pipeline — models, history, presets, the hotkey — needs only an account.
3. Pick a local model and download it
The default is the multilingual Small at ~480 MB. Parakeet v3 at ~600 MB runs 5–10× faster on CPU and covers English plus 24 European languages; its details live on the NVIDIA Parakeet model card. The Whisper model family goes up to Large v3 at ~3 GB and covers over 90 languages on the multilingual variants (the English-only builds are exactly that).
Check: the model shows as downloaded in Settings.
4. Put the cursor in Word, hold Ctrl+Space, talk, release
The transcript pastes where the cursor sits.
Check: your words are in the document, punctuation included.
After the model download, transcription happens on your machine with zero network activity; the same dictation works at 11,000 meters with the Wi-Fi off. Accuracy in local mode typically lands between 95% and 99% depending on model size, and a custom vocabulary teaches it the names the built-ins keep mangling. There is a cloud mode too, where you bring your own OpenAI key for the latest transcription models, but for Word dictation, local is the half I use. I dictated this section into the draft while a kettle negotiated with a teabag; the section needed two corrections, the tea needed none.
When the built-in tools are enough
If you dictate a paragraph a week, at a desk with good internet, only in Word, and in one of the ~15 languages Dictate supports in full, skip Whisper. Dictate is already in your Microsoft 365 plan and does that job well. The same goes for Win+H if your dictation never leaves Windows and the 40+ supported languages cover yours: it is one shortcut away and costs nothing extra. Third-party voice typing earns its place when you dictate daily, work offline, need custom vocabulary, or want one tool across every app — not before.
Twenty-five years separate the thrown Dragon headset from a Word document that types itself while you make tea. The tools caught up. The remaining work is picking which listener fits your week. Press Win+H today and dictate one real paragraph — a status update, a reply you have been putting off. If it holds up, you have a new default. If you hit the offline wall or the vocabulary ceiling, you know where the upgrade lives.
Try the same paragraph with one hotkey
Download Whisper, hold Ctrl+Space in Word, and watch the transcript land at your cursor.
Works on Windows 10 or later and macOS 11 or later. Local mode runs offline.
