Windows Dictation Not Working? Here's How to Fix It (And When to Just Switch)

Windows dictation tools break more often than they should. Win+H stops responding. Dragon throws a microphone error. Voice Access opens but doesn't transcribe. You say words and nothing happens.

This guide covers the most common causes and fixes for Windows dictation failures in 2026 - for both the built-in Windows voice typing and Dragon NaturallySpeaking. If the fixes don't work, there's a section at the end on switching to something that doesn't require constant maintenance.

Win+H (Windows Voice Typing) Not Working

Windows voice typing is triggered by pressing Win+H. If the bar doesn't appear, or it appears but doesn't transcribe, work through these in order.

Fix 1: Check Microphone Permissions

Fix 2: Check the Default Microphone

Windows voice typing uses the default input device. If you recently plugged in a headset or USB audio interface, the default may have changed to a device that isn't working or isn't physically connected.

Fix 3: Toggle Voice Typing in Settings

Fix 4: Restart the Speech Service

The Windows Speech Recognition service can hang without any visible error. Restarting it often clears the problem.

Fix 5: Check for Windows Updates

Voice typing bugs are occasionally introduced in Windows updates and fixed in subsequent patches. Check Settings > Windows Update and install any pending updates. If a recent update broke it, check the Microsoft Feedback Hub - others may have reported the same issue with a known workaround.

Fix 6: Re-register the Voice Typing Package

If nothing else works, the voice typing component may be corrupted. Open PowerShell as Administrator and run:

Restart after running this. It re-registers the speech package without affecting your other apps.

Dragon NaturallySpeaking Not Working

Dragon failures fall into a smaller set of categories. These are the most common ones.

Microphone Not Detected

Dragon is picky about audio devices. It doesn't always pick up system default changes automatically. Go to Dragon's Audio menu > Check Microphone. If your device isn't listed, disconnect any other audio interfaces and restart Dragon. USB microphones and audio interfaces sometimes require Dragon to be restarted after plugging in.

Transcription Stops Mid-Session

This usually means Dragon lost its focus on the active window. It happens most often with apps that have custom UI frameworks (Electron apps like Slack, VS Code, Discord). Dragon's accessibility hooks can lose the handle to the text field when the window repaints.

Workaround: click directly into the text field before dictating, not just into the window frame. Some Electron apps need the "Enable Dragon support in web views" option toggled in Dragon's advanced settings.

Dragon Running Slowly After Windows Update

Windows security updates occasionally conflict with Dragon's audio processing pipeline. Check Nuance's support site for the specific update version that broke things - they usually release a Dragon patch within a few weeks. Temporary fix: roll back the specific Windows update through Settings > Windows Update > Update history > Uninstall updates.

Activation or License Errors

Dragon's DRM and licensing server can cause activation failures, especially after reinstalling Windows or changing hardware. Contact Nuance support directly - this is not something you can fix without their help. Be ready to prove your purchase and provide your hardware ID.

The Honest Assessment: When to Stop Troubleshooting

Both Win+H and Dragon have the same structural problem: they're fragile. Windows updates break them. Audio device changes break them. App updates break them. You end up spending more time maintaining the tool than using it.

If you've spent more than 30 minutes troubleshooting Windows dictation and it's still broken, that's a signal worth listening to. The underlying technology is old. The maintenance burden is real.

Why dictate.app Doesn't Break the Same Way

dictate.app uses a fundamentally different architecture. Instead of hooking into Windows speech recognition or maintaining a complex local speech model, it:

• Records audio using standard Windows audio APIs
• Sends it to Groq's Whisper API (no local model to corrupt)
• Pastes the result using Windows clipboard APIs

There's no speech recognition engine to break. No user profile to corrupt. No licensing server to fail. The microphone setup takes 30 seconds - pick your default device and you're done.

Windows updates don't affect it because it doesn't rely on the Windows speech stack. Electron apps work fine because it pastes via clipboard rather than trying to hook into accessibility APIs. It's a simpler design that's harder to break.

Quick Diagnostic Checklist

Before concluding something is broken, run through this:

1. Is the microphone physically working? (Test in Sound settings)
2. Is it set as the default input device?
3. Do apps have microphone permission?
4. Is Windows fully updated?
5. Has the speech service been restarted?
6. Does the problem persist after a full system restart?

If you've checked all six and it's still broken, the tool is the problem - not your setup. Time to switch.

Questions? Email support@dictate.app or check the homepage for more on how dictate.app works.