Quick fix: Open Settings → Accessibility → Speech → Voice access and toggle On. Download the speech model when prompted (one-time, ~200 MB). Say “Voice access wake up” to start listening, then dictate by speaking naturally; say “Voice access sleep” to pause.

You want to type documents, emails, browser searches, or chat messages without using a keyboard — for accessibility reasons, repetitive strain injury recovery, or pure speed. Windows 11’s built-in Voice Access is much better than the old Speech Recognition: on-device processing, near-real-time transcription, and command vocabulary that handles formatting, editing, and app control.

Symptom: Need hands-free dictation in any Windows 11 app (Word, browsers, chat).
Affects: Windows 11 22H2 or later.
Fix time: ~15 minutes including initial setup.

What causes this

Voice Access is a system-level accessibility feature in Windows 11 that listens to your microphone and translates speech into text and commands. It runs locally (no cloud round-trip), supports US English plus a growing list of other languages, and works in any text field, system menu, or app control. It replaces the older Speech Recognition feature that needed extensive training and worked poorly in modern apps.

For most users, the only setup needed is enabling it once. After that, voice commands work everywhere — including login fields, the Start menu, web forms, and dialog boxes.

Method 1: Set up and start using Voice Access

The standard setup. One-time configuration.

Open Settings → Accessibility → Speech.
Toggle Voice access On.
Optionally toggle Start voice access after you sign in to your PC if you want it always available.
On first activation, Windows asks you to download a speech model (~200 MB). Click Download. Wait for completion.
The Voice Access toolbar appears at the top of the screen. Click Mic setup if needed to select your microphone.
Say “Voice access wake up” (or “Voice access start listening”). The toolbar status changes to Listening.
Open any text field (Notepad, web search bar). Speak normally — words appear as you say them.
To pause: “Voice access sleep” or “Voice access stop listening”.
To exit Voice Access: click the X on its toolbar or say “Voice access close”.

Within 30 seconds of starting Voice Access for the first time, you can dictate documents.

Method 2: Use voice commands for formatting and editing

For when basic dictation isn’t enough — adding formatting, navigating, and editing text by voice.

Open a Word document. Start Voice Access.
Dictate: “Hello world” — text appears.
For punctuation: say the name of the mark. “Hello world period” produces “Hello world.” Commands: comma, question mark, exclamation mark, colon, semicolon, open quote / close quote, new line, new paragraph.
For formatting: “Select last word”, then “Bold that”. Or “Select last three words” followed by “Italicize that” or “Underline that.”
To navigate: “Go to end of document”, “Go to beginning of line”, “Move down five lines.”
To delete: “Delete last word”, “Delete this sentence”, or “Undo that.”
To click buttons on screen: “Click File”, “Click Save,” etc. Voice Access shows numeric labels on clickable items via “Show numbers” — say a number to click that item.
For full command list: “What can I say” opens the help guide.

Practice with these commands builds enough vocabulary for hands-free workflow in ~30 minutes.

Method 3: Train Voice Access for your voice (deep customization)

For users with accents, soft voices, or specialized vocabulary.

Open Voice Access. Click the gear icon (Settings).
Go to Customize commands. You can add custom voice triggers for specific actions: “Open my project folder” → runs a path; “Send slack to John” → opens specific chat.
For better recognition: open Settings → Time & language → Speech.
Click Get started under Microphone. Speak the calibration prompts. This sets levels and reduces background noise tolerance.
Under Recognize non-native accents for this language: toggle On if English is not your first language. Improves recognition accuracy significantly for accented speech.
For specialized vocabulary (technical jargon, names): Voice Access learns from your usage. Use the word repeatedly; correct misrecognitions with “Correct that” — Voice Access shows alternatives and the chosen correction is remembered.
For privacy: all processing is on-device. Go to Settings → Privacy & security → Speech to verify cloud speech recognition is off if you want strict local-only processing.

This is the depth tier for users who plan to make Voice Access their primary input method.

How to verify the fix worked

Open Notepad. Activate Voice Access. Dictate a paragraph. Words should appear within ~500 ms of speaking.
Test commands: “New line,” “Select last sentence,” “Bold that.” All should execute correctly.
Open Edge. Say “Click address bar.” Cursor moves to URL bar. Dictate a search query. Press Enter via “Press enter.”

If none of these work

If Voice Access shows “I didn’t catch that” for most input, the microphone setup is the issue. Verify microphone selection: on the Voice Access toolbar, click the mic icon to switch to a different mic input. Built-in laptop mics often have poor pickup; a USB headset or external mic gives much better recognition. Adjust input level: open Settings → System → Sound → Microphone → Properties. Speak normally; the input meter should reach 60–80%. Adjust gain in the slider. Reduce background noise: close apps that play audio. For noisy environments, use a directional or headset mic. Check language: Voice Access defaults to English (US). If you speak British English, Spanish, German, French, or Chinese, install the matching speech pack via Settings → Time & language → Language & region → Add a language. Voice Access uses the speech pack matching your display language. Last resort: try the older Speech Recognition feature (Win + Ctrl + S). It needs more training but works on older Windows builds where Voice Access isn’t available.

Bottom line: Voice Access in Settings → Accessibility → Speech enables modern on-device dictation. Initial setup takes 5 minutes; learning the command vocabulary takes 30 minutes; productive hands-free workflow takes a week of regular use.