Ashi with a realtime voice loop.
This page uses browser speech recognition for fast mic input, a FastAPI WebSocket for session flow, and backend Ashi plus ElevenLabs for spoken replies. It is the clean bridge between text chat and true streaming voice.
Session is preparing.
Controls
Chrome and Edge work best for `SpeechRecognition`. If your browser does not support it, use the typed fallback and keep the same websocket loop.
Sign In
Sign in here so `ashi.dttc.io` can bind this live voice session to your real Ashi memory instead of only the browser session.
Anonymous session active right now. Sign in to restore your durable user memory online.
Advanced Token
If you already have a MEGATON bearer token, paste it here and reconnect. The sign-in form above is the normal path.
Typed Fallback
Transcript
Ashi Reply
Audio Playback
When ElevenLabs is ready, Ashi will attach MP3 output here and the page will try to autoplay it.
Status Feed
Booting the live voice foundation.