GPT Image 2.5 人像提示词仍在调
作者分享一版尚未满意的写实编辑人像提示词,含构图、发型、服装与场景的完整英文描述。




Still refining this prompt, and I still can’t get a result I’m happy with…… Keep going! 😅 GPT Image 2.5 : Image Type: Create a photorealistic editorial portrait. Composition: Waist-up portrait, front view. Subject: The subject is one 20-year-old Japanese or Korean female portrait subject. full bust, narrow defined waist. young sweet pretty face, soft rounded charm, bright friendly eyes, gentle cute beauty, approachable youthful portrait look. deep side-parted long soft waves, defined face-framing layers, polished wave shape. sleek wet finish, defined damp sections, neat separated strands, controlled close-to-head shape. soft black-tea brown hair, muted brown-black salon tone. soft matte skin texture, refined pores, velvety smooth finish. delicate innocent expression, relaxed brows, softly open eyes, gently closed lips, relaxed cheek muscles. Wardrobe: She wears tight body-skimming upper-body fit, neon pink triangle bikini top, slim halter strings, minimal sliding triangle cups, smooth stretch swim fabric. ultra-low-rise waistband sitting very low on the hips, neon pink low-rise side-tie bikini bottoms, string-tied hips, minimal swimwear coverage, clean swim-bottom silhouette at the lower crop edge. Pose and Composition: She has her chin slightly raised, both hands resting naturally along the body or on a nearby support surface, fingers relaxed and loosely following the body line, and presents a relaxed upright posture. Scene: The portrait takes place in abandoned overgrown park edge, cracked pavement textures, rusted play structures, weeds pushing through concrete, broken bench. Lighting: rainy overcast ambience with damp air and cold muted daylight, grey-blue rain sky where visible. wet-surface reflected fill on the subject, soft upward bounce from nearby surfaces, cool glossy highlights along lower contours. Camera Look: vintage sharp rendering, rich micro-contrast, subtle warm highlights, clean 35mm color density, precise edge definition. Create a series of imperfect amateur smartphone selfies. Keep the same person, outfit, and setting described above; freely vary poses and camera angles. Generate 4 separate 9:16 photographs, not a collage, with no added text.