模块化美食病毒图

Prompt

GPT Image 2.5. A modular prompt for food shots. It's not my favorite structure but fun experiment. 

2x2 grid, 16:9, do this for 4 clever viral subjects: INPUT ::= PERSONA_SEED + CUISINE_LANE + FRAME_TYPE    TASTE_1 :: read the cuisine  - what the cooking does to the room: wok steam, flour dust, smoke curl, citrus mist  - the hero texture of this lane: pull, crackle, drip, char, lamination  - honest kitchen evidence: sauce-splashed stove, stained apron, knife mid-board  - hands tell the résumé: small burn scars, practiced grip, flour in the nail beds    TASTE_2 :: read the frame type  - POV hands = 35mm overhead-diagonal, persona's hands + forearms only, dish 60% of frame  - portrait-with-dish = 50mm, persona mid-laugh or mid-taste, dish lifted into shared focus  - tablescape = 24mm slightly high, persona reaching in from the edge, abundance sprawl    APPETITE :: physics of delicious  - steam needs backlight to exist — place the window or lamp behind the food  - glisten on fats, matte on breads, condensation on cold glass  - controlled mess: a torn edge, a drip caught mid-run; sterile plates kill hunger  - persona's reaction is the seasoning: eyes on the food, never on the camera    STAGE :: warm frame  - 4:5, shallow depth (f/2.2), warm-neutral grade, no orange oversaturation  - kitchen or table is mid-use, not styled dead  - the unforgettable feature survives the crop even in the POV hands variant    BAR :: the viewer's mouth reacts before their brain does, and the hands could plausibly have cooked this for ten years

Attribution

Author

Gdgtify

Published

2026-09-09

Original post

View original ↗

Collected from

xianyu110/awesome-gpt-image2.5 · 2097620275940737326 · MIT