Disable Online Speech Recognition in Windows 10

Windows provides both a device-based speech recognition feature (available through the Windows Speech Recognition Desktop app), and a cloud-based speech recognition service in those markets and regions where Cortana is available.  Microsoft may use the voice data collected on your computer to help improve their speech services.

To use speech recognition, the option Getting to know you (the privacy setting under Speech, inking & typing) must be turned on because speech services exist both in the cloud and on your device. The info Microsoft collects from these services helps to improve them. Speech services that don’t rely on the cloud and only live on your device, like Narrator and Windows Speech Recognition, will still work when this setting is turned off, but Microsoft won’t collect any speech data.

When your Diagnostic and usage data setting (Settings > Privacy > Diagnostics & feedback) is set to Full, your inking and typing input data is sent to Microsoft, and the company use this data in the aggregate to improve the inking and typing platform for all users.

To disable Online Speech Recognition in Windows 10, do the following.

1. Open the Settings app.
2. Go to Privacy -> Speech.
3. On the right, turn off the toggle option under Online speech recognition.
4. The feature is now disabled.

Alternatively, you can apply a Registry tweak.

Disable Online Speech Recognition with a Registry tweak

1. Download the following ZIP archive: Download ZIP archive.
2. Extract its contents to any folder. You can place the files directly to the Desktop.
3. Unblock the files.
4. Double click on the Disable Online Speech Recognition.reg file to merge it.
5. To undo the change when needed, use the provided file Enable Online Speech Recognition.reg.

You are done!

The Registry files above modify the Registry branch

HKEY_CURRENT_USER\Software\Microsoft\Speech_OneCore\Settings\OnlineSpeechPrivacy

Tip: See how to go to a Registry key with one click.

They change the 32-bit DWORD value named HasAccepted.

• HasAccepted = 1 - Online Speech Recognition is enabled.
• HasAccepted = 0 - Online Speech Recognition is disabled.

Note: Even if you are running 64-bit Windows you must still create a 32-bit DWORD value.

Also, Starting with Windows 10 build 17063, the OS has got a number of new options under Privacy. These include the ability to control usage permissions for your Library/data folders, microphone, calendar, user account info, file system, location, contacts, call history, email, messaging, and more. Also, there is a new layout for privacy settings shown above.

Finally, you can disable Online Speech Recognition from the privacy page of the Windows Setup program when installing the OS from scratch.

That's it.

Articles of interest:

• Unlock Extra Text to Speech Voices in Windows 10
• Windows 10 features new text to speech voices for Narrator and Cortana