π¬ Just speak β e.g. "What can you help me with?". Pause briefly when done.
Record ~10s or upload a clean speech file. The assistant can then speak in this voice. Longer / multiple clips improve the clone.
Add more clips per person to improve their clone, or delete to reset.
Enrolling/deleting voices needs the admin key; talking works with any valid key.
Self-hosted LLM the assistant answers with. Leave blank to use the server default.