Windows Speech Recognition Registry Locations

Windows Speech Recognition stores its configurations, voice profiles, acoustic training data, and user dictionaries across specific keys in the Windows Registry Editor (regedit). These settings are divided between user-level customizations in HKEY_CURRENT_USER and system-wide engine configurations in HKEY_LOCAL_MACHINE. This guide outlines the exact registry paths where Windows Speech Recognition and modern OneCore speech settings are saved.

User-Specific Settings (HKEY_CURRENT_USER)

Most user-specific speech recognition settings, including custom profiles, vocabulary additions, and audio preferences, are stored under the current user hive.

1. Main Speech Recognition Profile and Data

• Path: HKEY_CURRENT_USER\Software\Microsoft\Speech
• Purpose: Serves as the primary root for user-level speech data.

2. User Speech Profiles

• Path: HKEY_CURRENT_USER\Software\Microsoft\Speech\CurrentUserProfiles
• Purpose: Contains registry tokens for the active speech recognition profile, including training data, acoustic models, and adaptation metrics.

3. Speech Dictionary and Lexicon

• Path: HKEY_CURRENT_USER\Software\Microsoft\Speech\CurrentUserLexicon
• Purpose: Stores user-added words, custom pronunciations, and vocabulary adjustments made through the Speech Dictionary tool.

4. Audio Input Configuration

• Path: HKEY_CURRENT_USER\Software\Microsoft\Speech\AudioInput
• Purpose: Contains user preferences for the selected microphone, input levels, and audio device mappings specific to speech recognition.

Modern Speech and Privacy Settings (OneCore)

On Windows 10 and Windows 11, modern speech features (such as cloud-based speech recognition and dictation) utilize the “OneCore” speech subsystem.

1. Modern Speech Settings

• Path: HKEY_CURRENT_USER\Software\Microsoft\Speech_OneCore
• Purpose: Contains configuration settings for newer Windows dictation features, voices, and recognizers.

2. Online Speech Recognition Privacy

• Path: HKEY_CURRENT_USER\Software\Microsoft\Speech_OneCore\Settings\OnlineSpeechPrivacy
• Key Name: HasAccepted (DWORD)
• Values:
  • 1 = Online speech recognition enabled
  • 0 = Online speech recognition disabled

System-Wide Engine Settings (HKEY_LOCAL_MACHINE)

Global engine definitions, installed recognition engines, and default audio endpoints are stored at the machine level.

1. Installed Recognizers

• Path (64-bit systems): HKEY_LOCAL_MACHINE\SOFTWARE\Microsoft\Speech\Recognizers
• Path (32-bit components on 64-bit systems): HKEY_LOCAL_MACHINE\SOFTWARE\WOW6432Node\Microsoft\Speech\Recognizers
• Purpose: Lists all installed speech recognition engines, supported language packs, and engine-specific capabilities.

2. Global Audio Formats and Devices

• Path: HKEY_LOCAL_MACHINE\SOFTWARE\Microsoft\Speech\AudioInput
• Purpose: Defines the default system audio input interfaces and driver tokens recognized by the speech subsystem.

How to Access These Settings in Regedit

1. Press Windows Key + R to open the Run dialog box.
2. Type regedit and press Enter.
3. Copy any of the paths listed above and paste it into the Registry Editor address bar at the top, then press Enter to jump directly to the key.