The original Aicom English pronunciation engine with a reconstructed sound model. The original DSP firmware remains unavailable.
Install the add-on, restart NVDA, and select Accent Messenger. Both 32-bit and 64-bit engines are included. No separate Python or Visual C++ Redistributable installation is needed.
The package is now named accentMessenger. If an experimental
package is installed, a startup dialog asks whether to remove it. Choose Yes,
then restart to finish. No leaves it installed and asks again next startup.
Both packages supply the same synth module and must not remain together.
Your existing Messenger voice settings are retained.
This version includes the approved straighter onset for isolated E/e, with or without a final period. It preserves the louder output, voice controls, consonant improvements and responsive cancellation of earlier builds.
Leave "Read numbers as individual digits" unchecked for "one hundred" or check it for "one zero zero." It defaults to unchecked. English decimals and ordinals are supported through the number parser from Outspoken.
A separate Accent Messenger SAPI installer provides the same engine for SAPI applications, with its own settings dialog and TOML preferences.
NVDA 2021.1 or later is the declared range. The native libraries target Windows 7 SP1 and later; use an NVDA version that supports your Windows version. Windows 7 and older NVDA still need real-machine testing. This release has passed native engine and simulated NVDA audio checks, not a full compatibility matrix.
English pronunciation, with ten voice characteristics: V0 is deepest, V9 sharpest, and V5 is the existing default. Other variants are experimental. Inflection uses five steps: 0 for monotone, 25/50/75 for reduced intonation, 100 for full intonation (default). Word spacing runs from 0 (shortest, default) to 9; it adds pauses and cannot shorten the existing minimum. Rate, pitch and volume are independently adjustable. Cancellation, pause, speech indexes and breaks are supported. Long passages are split into short requests, which can change pauses and intonation. Text is transliterated to ASCII. Character-mode commands and per-command capital pitch changes are not yet implemented. The missing original DSP firmware has not been recovered.
Try ordinary navigation, holding and releasing Tab, Ctrl to stop, pause and resume, and reading a paragraph. Please include your Windows and NVDA versions when reporting results. Errors are recorded in NVDA's normal log.
Maintained by Tamas Geczy, with development assistance from OpenAI Codex.
Original pronunciation software by Aicom. Native emulation uses Unicorn;
Windows 7 support uses YY-Thunks. The complete source is available separately as
accent-messenger-0.4.0-source.zip. Keep that companion
download alongside the add-on when sharing this version. See LICENSE and THIRD-PARTY-NOTICES.md for component terms.