What Microsoft 365 Dictate actually is
Microsoft ships a built-in dictation feature across the Microsoft 365 family — Word, Outlook, OneNote, PowerPoint, and Teams. You click the microphone button or press Alt+` in Word and start speaking. The feature is officially called Microsoft 365 Dictation, often referred to as Dictate in Word or Office voice typing. Behind the scenes it streams your audio to Microsoft's Azure-backed speech service, returns the transcript, and types it into the document.
This is great when it fits the brief: an internet-connected M365 license, content you don't mind uploading to Microsoft, and a workflow that lives entirely inside Office. Outside that envelope it falls down quickly.
The four real limitations of Microsoft Word voice dictation
1. Cloud-only — your audio leaves your machine
Every spoken word in Word, Outlook, or Teams Dictate is uploaded to Microsoft's cloud. For most users that's fine. For lawyers drafting privileged client memos, doctors recording patient notes, finance analysts speaking about earnings before they're public, journalists with sources to protect, or engineers narrating proprietary code — it isn't. There is no on-device fallback in Office.
2. Locked inside Office
Dictation in Microsoft 365 only works inside Microsoft 365. The moment you switch to Slack, Notion, VS Code, a Chrome tab, your terminal, Photoshop, a CRM, or even Microsoft Edge's address bar, the built-in dictate button is gone. You're back to typing.
Power users want one hotkey that works everywhere — not five different dictate buttons in five different apps, half of which don't have one.
3. Requires an active Microsoft 365 subscription & internet
Dictate is gated behind a live M365 subscription. If you've installed Office via a perpetual licence (Office 2019, 2021, LTSC), Dictate isn't available. If your internet is slow, flaky, or absent, Dictate stops working. On long-haul flights, in airplane mode, on a co-working VPN that throttles audio, or in a secure room with no external connectivity, Office voice typing simply isn't there.
4. No AI rewrite — it's transcription only
Microsoft 365 Dictate transcribes what you say. If you ramble, ramble shows up on the page. There is no built-in path to say "make this more concise" or "rewrite this in a professional tone" without copying the text into Copilot, paying for Copilot, and pasting back. That's a separate product, separate cost, and another break in the workflow.
AirTypes vs Microsoft 365 Dictate — side-by-side
| Capability | Microsoft 365 Dictate | AirTypes |
|---|---|---|
| Audio leaves device | ⚠️ Yes — sent to Azure | ✅ No — local Whisper |
| Works offline | ❌ No | ✅ Yes |
| Works in Word / Outlook / Teams | ✅ Yes | ✅ Yes |
| Works in Slack, browser, IDE, CRM | ❌ No | ✅ Any app |
| Requires M365 subscription | ⚠️ Yes | ✅ No (standalone) |
| Languages | ~60 | 100+ (Whisper) |
| AI rewrite / prompt mode | ❌ Copilot only (paid extra) | ✅ My Agent (BYOK) |
| Per-user cost | Bundled in M365 (~$12.50/mo) | $3.99/mo (locked) |
| Linux support | ❌ No | ✅ Yes |
| macOS support | ✅ Yes | ✅ Yes |
| Windows support | ✅ Yes | 🔜 In development |
Dictating in Word — the offline workflow
The mental model is the same as Office Dictate, but with one global hotkey instead of clicking the microphone button. Open your Word document, place your cursor, hold the AirTypes hotkey (default Ctrl+Alt+D), speak naturally, release. Whisper transcribes on your CPU and the text is injected at the cursor — exactly as if you typed it.
Crucially: Word never knows there was audio. There is no plugin, no add-in, no cloud upload. From Word's perspective, a keyboard typed the characters. This means it works on any Word version — M365, perpetual Office 2021, Word 2019, Word for Mac, Word in Citrix or VDI sessions, even Word running inside a sandboxed VM with the audio device passed through.
Outlook — write emails by voice without sending audio to Microsoft
Outlook is one of the highest-value targets for voice typing. People write hundreds of emails a week, most are repetitive, and most are faster spoken than typed. With AirTypes, the workflow is:
- Open a new email or reply.
- Hold the hotkey, speak the email body.
- Release. Text is in the compose pane.
- Optionally hold the My Agent hotkey and say "Email, polish this and add a friendly opener" — the AI rewrites in place.
For people in regulated industries — finance, legal, healthcare, government, defence — this matters because the body of an email is often the most sensitive content in your day. Sending that audio to Microsoft for transcription, even with their compliance certifications, is not always allowed by internal policy. Local-only dictation removes the question entirely.
Teams — chat messages and meeting notes by voice
Teams chat is short-form, fast, and a perfect fit for hotkey dictation. Hold, speak, release — message typed. For meeting notes (whether in OneNote, a Loop component, or a private Word doc), AirTypes is especially useful because you can dictate during a Teams call without your microphone audio ever leaving your machine — you're using a hotkey-triggered second mic capture, not pushing the call audio through a transcription service.
Excel and OneNote — places Microsoft Dictate is weak
Microsoft 365 Dictate inside Excel is limited to comments and notes — you cannot dictate into a cell directly with their built-in dictation. AirTypes types into whatever has focus, so you can speak a cell value, a formula, or a long header into any cell on the sheet.
In OneNote, Microsoft Dictate works but is cloud-bound. AirTypes works the same way it does in Word, with the audio staying local — useful when you're capturing meeting notes that contain financials, names, or strategy you'd rather not stream to a third-party speech API.
My Agent — dictate the rough draft, your AI rewrites for Office tone
This is the part Microsoft 365 doesn't offer at this price point. AirTypes My Agent lets you bring your own API key (OpenAI, Anthropic Claude, Google Gemini, OpenRouter, Groq, or a local Ollama server) and use a separate hotkey to push your spoken text through a profile.
Example profiles people set up for Office work:
- Email — system prompt: "Rewrite the user's words as a clear, professional business email. Only return the email body, no commentary."
- Report — system prompt: "Rewrite the user's words as a paragraph for an executive report. Tighten phrasing, keep facts."
- Reply — system prompt: "Compose a short, polite reply to the email above the cursor based on what the user dictates."
- Bullet — system prompt: "Convert the user's words into a clean bullet list with parallel structure."
Activate the profile by saying its name first: "Email, tell the team the deadline moved to Friday and we'll send the new draft tomorrow." A clean, polished email body appears at the cursor. No browser tab, no Copilot subscription, no copy-paste.
Privacy & compliance — why offline matters for enterprise teams
Microsoft is SOC 2, ISO 27001, HIPAA, and GDPR-compliant. So is the dictation service, on paper. The question isn't whether Microsoft can handle the audio responsibly — it's whether your internal data classification policy allows sensitive audio to leave the endpoint at all.
For teams in highly regulated industries — automotive (BMW, Mercedes, Volkswagen), pharmaceutical (Pfizer, Roche), banking, defence, public sector — the answer is increasingly "no". When the policy is "no third-party audio processing", local-only dictation is the only option. AirTypes was built for exactly that constraint: Whisper runs on your CPU, the audio is never transmitted, and there is no account or telemetry tied to the transcription itself. The companion Cloud features (My Agent prompt routing) are entirely opt-in and bypass our servers — they go directly from your machine to your AI provider with your own key.
For more on the enterprise angle, see our deep dive on enterprise offline voice recognition for regulated industries.
Frequently asked questions
Does Microsoft Word dictation work offline?
No. Microsoft 365 Dictate streams your audio to Microsoft's cloud speech service for transcription. It will not work without an internet connection and is only available on subscription M365 — not on perpetual-licence Office editions.
Does AirTypes work inside Microsoft Word, Excel, and Outlook?
Yes. AirTypes types into whatever app has focus — there is no plugin or add-in. Word, Excel, Outlook, OneNote, Teams, PowerPoint all work the same way as any other desktop app: open the document, place the cursor, hold the hotkey, speak.
Is voice dictation faster than typing in Word?
For most people, yes. Average typing speed is around 40 words per minute. Comfortable speech is around 130–160 words per minute. Even allowing for editing time after dictation, professionals typically finish first drafts 2–3× faster by voice.
Can I keep using Microsoft 365 Dictate alongside AirTypes?
Yes — they don't interfere. Microsoft 365 Dictate is a button inside Office. AirTypes is a global hotkey across the OS. Many teams use the built-in Office button for non-sensitive work and the offline AirTypes hotkey for confidential drafts.
When will AirTypes ship for Windows?
The Windows build is in active development. Linux and macOS are available today; the Windows version follows the same offline architecture and global hotkey model. Sign up on the download page to be notified at launch.
Try the offline alternative to Microsoft 365 Dictate
7-day free trial. No credit card. Linux & macOS today, Windows in development.
Download AirTypes