What was built and checked
A managed Chrome test reproduced an unavailable system English voice when the visitor selected Read aloud. The November 8 update replaces that device-voice dependency with 318 bundled, short MP3 segments recorded ahead of time from the stock Kokoro af_sarah synthetic voice. The older twelve in-game WAV recordings remain available for exact transcript matches. Playback does not call browser text-to-speech, run an AI model, contact a cloud voice service or require a GPU. Read-aloud is opt-in, full text captions remain, and the Save & sound panel includes a Test voice control. Drawing the 3D scene pauses beneath open dialogue or map panels.
Verified in automated checks: 69 Node test groups passed, including nine new audio groups; all 505 authored dialogue/practice combinations resolved to bundled clips. The Byte and Bolt voice compatibility QA passed on commit 8c880ee, including isolated Chromium media-playback checks and the packaged source ZIP. GitHub Pages deployment succeeded for the same commit.
What remains unverified
These checks are not a human pronunciation/listening review and were not conducted on the user's laptop or other physical devices. No measured fan/noise, real laptop performance, offline file:// ZIP execution or universal browser/audio compatibility is claimed. This is a website game accessibility and compatibility improvement, not new native NewBrain learning; accepted Original12 remains 2/12. No completed Avatar Builder body or functional internal anatomy is implied.
Full dated game QA record → · Source build guide → · Development update → · Timeline →