“Echo what you see.”
Sends camera frames to a vision model and returns a spoken description of what is visible.