Build 26278082

v2.4.15 Patch Notes

This update focuses on spoken output: the keyboard can read aloud on its own, anything that can't be spoken now says why, and short sentences come out right.New

• The keyboard's speak-on-send button now works on its own — you no longer need to switch on "Enable Speech Synthesis (TTS)" in Voice Translation first

• The wrist panel now shows a hint line at the bottom of the voice page, so an unsupported reading language or a failed cloud synthesis tells you why right in VR

• Cloud synthesis failures now come with a reason you can act on (invalid API key, unknown model, unreachable server, no audio returned) instead of a raw English error

Improvements

• The Speech Synthesis (TTS) card on the Voice Translation page now warns when the current target language can't be spoken by the local model, and lists the ten languages that work

• Switching speech synthesis off now frees the memory and VRAM it used right away, when the keyboard's speak-on-send is off as well

Fixes

• Cloud synthesis is no longer skipped for languages the local model doesn't support — Vietnamese, Thai and others now play with multilingual cloud voices

• Reading a short sentence without closing punctuation no longer adds an extra sentence of its own

• Fixed local synthesis occasionally going silent for a whole sentence, and a crash that could happen when speech synthesis was switched off mid-sentence