Models come directly from your OpenAI-compatible local server. Your choice is saved in the persistent data folder.
Allow microphone permission once to show Windows input-device names.
Face and emotions
Stored on this device so each screen can have its own look.