Text to Speech Voice Registry Locations in Windows

The Windows Registry stores active configurations and installed models for system-wide Text-to-Speech (TTS) engines across distinct registry keys depending on whether the system uses the modern Windows OneCore speech platform or the legacy Microsoft Speech API (SAPI 5). These configurations determine the default voice tokens, pitch, speed, and installed language packs used by Windows Narrator, accessibility tools, and third-party software.

Modern OneCore Speech Voices (Windows 10 and 11)

Modern Windows applications, the Windows Settings app, and the current Narrator utilize the Speech_OneCore platform. The system-wide and user configurations are located in the following registry paths:

• System-Wide Installed Voices: HKEY_LOCAL_MACHINE\SOFTWARE\Microsoft\Speech_OneCore\Voices\Tokens
• Active User Default Voice: HKEY_CURRENT_USER\SOFTWARE\Microsoft\Speech_OneCore\Voices Within this key, the DefaultTokenId string value points directly to the active voice token in Tokens.
• 32-Bit Subsystem (for 32-bit apps on 64-bit Windows): HKEY_LOCAL_MACHINE\SOFTWARE\WOW6432Node\Microsoft\Speech_OneCore\Voices\Tokens

Legacy SAPI 5 Speech Voices (Desktop Applications)

Classic Windows applications and older software rely on the legacy SAPI 5 speech architecture. The configurations for these voices are stored separately:

• System-Wide Installed SAPI Voices: HKEY_LOCAL_MACHINE\SOFTWARE\Microsoft\Speech\Voices\Tokens
• Active User Default Voice: HKEY_CURRENT_USER\SOFTWARE\Microsoft\Speech\Voices The string value DefaultTokenId sets the default active voice path.
• 32-Bit SAPI Voices on 64-Bit Windows: HKEY_LOCAL_MACHINE\SOFTWARE\WOW6432Node\Microsoft\Speech\Voices\Tokens

How the Active Voice is Configured

Under both modern and legacy keys, individual voice profiles are stored as subkeys within the Tokens directory (for example, ...\Tokens\MSTTS_V110_enUS_DavidM).

The active voice is established by the DefaultTokenId registry value located in HKEY_CURRENT_USER\SOFTWARE\Microsoft\Speech\Voices (or Speech_OneCore\Voices). The data of this string value contains the full registry path to the desired voice token, instructing the operating system and speech-enabled applications which profile to load by default.