Four boxes
Mic to text.
Text to the same agent loop as chat.
Tool or refusal.
Text to speech.
Label the speech boxes as separate from the language model.
Sources and go deeper
- MDN Web Docs, Web Speech API community last checked 2026-10
- Web Speech API community draft community last checked 2026-10
- W3C /TR/speech-api/ spec confirm This URL returned 404 on 2026-10. The community draft above is the page that opened. last checked 2026-10
- Speech synthesis — Wikipedia wikipedia last checked 2026-10
- Speech recognition — Wikipedia wikipedia last checked 2026-10