by OpenAI

GPT Image 2.5 Prompts

OpenAI's GPT Image 2.5 — up to 50% faster than Images 2, with pinpoint local edits that leave the rest of the frame untouched and reference subjects that stay recognizable across repeated rounds.

image

GPT Image 2.5

Create an irresistibly captivating, ultra-realistic close-up portrait of a naturally beautiful young woman designed to create an instant “STOP SCROLLING” reaction. The camera is extremely close to her face, capturing the mesmerizing warmth of her deep brown eyes as the undeniable focal point. Her eyes must look alive and intensely realistic, with intricate iris patterns, tiny natural veins, wetness along the waterline, crisp catchlights, and individual eyelashes sharply resolved. Her beauty is raw, authentic and imperfect—visible freckles scattered naturally across her cheeks and nose, realistic pores, subtle peach fuzz, tiny beauty marks, natural skin texture and delicate variations in tone. Do NOT make her skin artificially flawless. Above one eyebrow is a thin, naturally healed scar, while a subtle healed scar curves along her cheek, giving the portrait a mysterious, unforgettable character. Her long, glossy black hair falls naturally around her face, with a few delicate strands crossing her forehead and cheek. Her eyebrows are full, naturally shaped and individually detailed. Soft, naturally pink lips remain slightly relaxed, creating an effortless expression that feels candid rather than posed. THE VIRAL HOOK: Capture the exact split-second where she looks directly into the camera as if she has noticed the viewer. Her expression is calm, mysterious and emotionally magnetic—subtle vulnerability mixed with quiet confidence. The viewer should feel like they accidentally captured a private, once-in-a-lifetime moment. Lighting is soft cinematic daylight entering from one side, producing luminous highlights across her eyes, cheekbones and skin while maintaining realistic shadows. Add an extremely subtle atmospheric glow around the edges without making the image dreamy or artificial. Composition: extreme facial close-up, eyes positioned near the visual center, face filling almost the entire frame, slightly unconventional tilted camera angle, intimate perspective, natural asymmetry, realistic facial proportions. Camera realism: full-frame professional camera, 85mm portrait lens, f/1.4, razor-sharp focus on both eyes, extremely shallow depth of field, natural optical bokeh, realistic lens rendering, subtle filmic micro-contrast. Visual quality: hyper-photorealistic, cinematic editorial photography, 8K-level detail, HDR, physically accurate skin, realistic reflections, lifelike eyes, natural color grading, premium fashion-magazine aesthetic, authentic human imperfections, breathtaking micro-details. ABSOLUTELY AVOID: plastic skin, excessive retouching, doll-like face, CGI appearance, uncanny eyes, exaggerated makeup, fake freckles, over-saturated colors, excessive HDR, artificial skin smoothing, distorted facial anatomy, extra eyelashes, asymmetrical eyes, blurry eyes, watermark, text, logo. Overall feeling: beautiful enough to stop the scroll, realistic enough to feel like a photograph, mysterious enough to make people stare twice.

Details

GPT Image 2.5

Create an extreme close-up underwater portrait of Abraham Lincoln after the subject has just jumped into a swimming pool or clear shallow water. Keep the mood calm, weightless, and ethereal, emphasizing light reflections. No struggle or urgency. 16:9

Details

GPT Image 2.5

Ultra-realistic iPhone creative portrait, identity preserved exactly from reference image. Young woman standing indoors in front of a glass storefront or reflective wall, photographed from a slightly high front angle. She holds a smartphone horizontally directly in front of her face, covering her eyes and upper face in real life. The phone screen faces the camera and clearly shows a live camera/photo view of her face. Composition: The real woman is visible behind the phone, but her face is mostly hidden by the horizontal phone. On the phone screen, her face appears clearly centered, smiling softly and winking. The phone acts like a frame within the frame. Her hand grips the top edge of the phone, fingers visible across the top. Pose: One hand holds the phone horizontally across her face. Her other hand is lower near her waist making a peace sign or casual gesture. She stands relaxed, facing camera. Outfit: White fitted zip-front sleeveless top, loose light-wash baggy jeans, brown belt, playful accessories, rings, bracelet, red-and-white shoulder bag, small plush/keychain accessories hanging from bag or belt. Phone screen details: The phone screen shows a clean live-camera image of her face with a wink and soft smile. Add cute sticker-style doodles around the screen/photo area: small colorful sparkles, stars, flower sticker near the face. The stickers should look like playful phone-camera decoration, not random floating objects. Environment: Indoor mall/storefront area with glass doors, reflections, polished floor, soft daylight mixed with indoor light, casual streetwear snapshot vibe. Camera + lighting: 0.5x iPhone wide-angle photo, close candid framing, realistic reflections on glass and phone screen, slight distortion on hands, raw social-media feel. Important: Match the phone-covering-face composition, horizontal phone screen showing her face, wink expression on screen, hand gripping phone, playful stickers, white zip top, baggy jeans, accessories, and indoor glass background closely. Natural skin texture, realistic screen reflections, no beauty filter, no smoothing. 4K vertical 9:16.

Details

Related models

GPT Image 2

OpenAI's top-ranked image model — it plans the layout, reasons, and self-checks before it draws, nailing complex multi-part instructions.

image

233 prompts

GPT-6 Astra

OpenAI's GPT-6 Astra — the frontier reasoning model inside ChatGPT that plans first, then directs image and video generation from a single conversational prompt.

imagevideo

8 prompts

Nano Banana

Google's fast, free conversational image generation and editing model.

image

67 prompts

Seedream 5

ByteDance's high-fidelity bilingual image model — the best quality-to-cost ratio for high-volume pipelines.

image

37 prompts

Browse AI Image Prompts All models