The appetite triggers to prompt
Food reads as delicious through specific visual cues: glisten ("glossy glaze catching the light"), steam ("steam rising and curling"), pull and stretch ("molten cheese stretching between slices"), the pour ("honey ribboning down in slow motion"), and fresh texture ("water droplets on just-washed berries"). Every strong food prompt contains at least one of these in motion.
Backlight is the food photographer's cheat code — steam and glisten are nearly invisible front-lit and luminous backlit. "Warm window backlight, steam glowing" transforms an ordinary bowl shot.
Framing and motion for food
Go tighter than feels natural: macro framing that crops the plate makes texture the subject, and texture is what triggers appetite. The SLOW ZOOM preset creeping toward the dish is the genre's default move; a top-down LOCKED OFF frame suits flat-lay and preparation content; PUSH THROUGH past foreground ingredients adds restaurant-menu drama.
One dish, one hero element, one move per clip. The most common failure is prompting an entire table spread — appetite is specific, and the camera should be too.
Menus, delivery apps, and social
Restaurants use generated clips where static menu photos used to sit: a looping steam shot next to the signature ramen converts better than any still. Delivery platforms and ghost kitchens generate consistent hero clips across hundreds of menu items without staging a single shoot.
For food content creators, generation solves the b-roll problem: recipe videos need beauty shots of stages you can't re-cook — the rise, the sear, the first cut. Generate the beauty pass; film the instruction.