Once the Attunement Wizard is completed, you are ready to wake Yumi up and start your first conversation.
1. Wake Yumi Up
From your terminal, select "🌸 Wake Yumi Up" from the main dashboard or run the dedicated command:
uv run yumi wake-up
This starts the Uvicorn server in the background and automatically opens your web browser to:
http://localhost:8000/
2. Activate Audio Context
To comply with browser security rules (which prevent sites from auto-playing audio or capturing the mic without user interaction), you must click the large pink button:
[ Connect & Start Audio Context ]
- Microphone Permissions: Your browser will ask for microphone access. Allow it.
- WebSocket Handshake: The client establishes a binary WebSocket connection with
ws://localhost:8000/ws. - Avatar Render: Yumi's Live2D avatar compiles and centers on your screen.
3. Say Hello!
Start speaking naturally. You do not need to click any push-to-talk button!
- Speak a simple greeting: "Hello Yumi! Nice to meet you."
- Watch your terminal window. You will see
[Voice Activity Detected]when you begin speaking, and[Transcription: Hello Yumi...]when you finish. - Yumi's avatar will instantly react:
- She will nod or smile based on the emotional assessment.
- She will begin speaking her response, and her lips will open and close in perfect synchrony with the sound waves!
4. Test Her Senses
To see the depth of Yumi's capabilities, try these interactive verbal tests during your conversation:
- Switch Personalities: Speak clearly: "Switch to tsundere" or "become kuudere". Yumi will acknowledge the switch instantly and adjust her language, expressions, and tone!
- Barge-In: While she is in the middle of a sentence, speak loudly: "Wait, hold on!". Watch how her voice instantly cuts off, and she immediately closes her mouth and begins listening to your next thought.
Now that you have had your first conversation, learn more about how to set her up across different operating systems in our Installation section!