feat(voice): on-device voice tier — Gemma 4, Moonshine and Kokoro in the browser

.env.example

@@ -72,6 +72,11 @@ OPENAI_REALTIME_TRANSCRIBE_MODEL=gpt-4o-mini-transcribe
 # Set to 1 to keep them while debugging; delete the log when you are done.
 GEV_VOICE_LOG_CONTENT=
 OPENAI_HUD_SUMMARY_MODEL=gpt-5-nano
+# On-device voice (Cloud / On-device choice in the mic control) runs models
+# in the browser. Its natural voice downloads an open-source speech component
+# (GPL) to the browser at runtime; set to off to leave natural voice out and
+# speak with the browser's on-device voices (or text) instead.
+# GEV_NATURAL_VOICE=off
 # Optional: opt-in per-IP rate limit for the OpenAI cost endpoints
 # (/api/realtime/token + /api/openai/hud-summary), in requests per minute.
 # DEFAULT IS UNLIMITED — unset (or 0) means no throttling, unchanged behavior.