Identify
Photo first. You get a general summary. Then voice-describe it, pick a main category, and fix anything by hand.