One idea → a short film
Make a movie

Prompt library · GPT Image 2.5

The best ChatGPT image prompts, tested on GPT Image 2.5

GPT Image 2.5 reads a prompt the way a photographer reads a brief: what the image is for, what is in it, how it is framed and lit, and what must not appear. These eleven prompts follow that order across the jobs people actually use ChatGPT images for: candid photos, food, thumbnails with text, posters, interiors and logos.

  • 11 tested prompts
  • Made with GPT Image 2.5
  • Real outputs, unedited

Every example was generated with GPT Image 2.5 (Sunburst) on MakeInfluencer.ai. People shots use one photo of our own AI character as the reference; food, poster, interior and logo shots are pure text-to-image. Paste the prompts into ChatGPT to get similar results. MakeInfluencer.ai is not affiliated with OpenAI.

Full-body street-style photo of a woman in a camel coat crossing a city street
Candid phone photo of a woman laughing while stirring pasta in a small kitchen
Beauty close-up of a woman's cheek with terracotta blush and visible skin texture

Photo prompts with people

Candid, street and beauty shots. Upload a photo of yourself and keep the first paragraph so the face stays yours.

Full-body street-style photo of a woman in a camel coat crossing a city streetNo. 01GPT Image 2.5 · photo edit

Street-style crosswalk

Use the uploaded photo only for the woman's identity. Keep her face exactly as in the photo: face shape, eyes, eyelids, eyebrows, nose, lips, jawline, hairline, hair colour and skin tone. Do not beautify, slim, smooth or retouch her face; keep its natural asymmetry and texture.

Create a photorealistic editorial street-style photograph of the woman walking across a city crosswalk in late afternoon, mid-stride, glancing just past the camera. She wears an oversized camel wool coat over a white ribbed knit top, straight-leg dark jeans and white leather trainers, and carries a plain black leather shoulder bag with no visible logo. Full body visible, feet included, shot from slightly below eye level, 35mm look, pedestrians and buildings softly out of focus. Low warm sun from the side, long shadows on the asphalt. 35mm film look, subtle grain.

Real skin texture: visible pores, fine vellus hair, slightly uneven natural tone. Unretouched, no skin smoothing, no beauty filter, no glamorization. No text, no logos, no watermark.

Upload one clear photo of yourself, then paste the prompt.

Candid phone photo of a woman laughing while stirring pasta in a small kitchenNo. 02GPT Image 2.5 · photo edit

UGC kitchen moment

Use the uploaded photo only for the woman's identity. Keep her face exactly as in the photo: face shape, eyes, eyelids, eyebrows, nose, lips, jawline, hairline, hair colour and skin tone. Do not beautify, slim, smooth or retouch her face; keep its natural asymmetry and texture.

Create an unedited back-camera phone photo of the woman, hair in a messy bun, laughing while stirring a pot of pasta in a small apartment kitchen. A jar of tomato sauce and a wooden spoon on the counter, a few dishes in the sink behind her. Taken by a friend from the doorway, slightly tilted, waist-up. Warm overhead kitchen light plus blue evening light from the window. Phone look: slight noise, natural colour, not graded. A little steam shine on her forehead.

Real skin texture: visible pores, fine vellus hair, slightly uneven natural tone. Unretouched, no skin smoothing, no beauty filter, no glamorization. No text, no logos, no watermark.

Upload one clear photo of yourself, then paste the prompt.

Beauty close-up of a woman's cheek with terracotta blush and visible skin textureNo. 03GPT Image 2.5 · photo edit

Texture-honest beauty close-up

Use the uploaded photo only for the woman's identity. Keep her face exactly as in the photo: face shape, eyes, eyelids, eyebrows, nose, lips, jawline, hairline, hair colour and skin tone. Do not beautify, slim, smooth or retouch her face; keep its natural asymmetry and texture.

Create a photorealistic beauty close-up of the woman showing a sheer terracotta cream blush blended high on the cheekbone and a clear balm on the lips. Tight crop from brow to chin, face turned slightly to the left. Soft side window light from the right, raking across the cheek. Skin texture remains visible under the blush: pores, fine vellus hair, slight shine on the nose. Macro-lens look, shallow depth of field.

Real skin texture: visible pores, fine vellus hair, slightly uneven natural tone. Unretouched, no skin smoothing, no beauty filter, no glamorization. No text, no logos, no watermark.

Upload one clear photo of yourself, then paste the prompt.

Woman with a clear umbrella in a narrow wooden lane in Kyoto at dawn after rainNo. 04GPT Image 2.5 · photo edit

Travel photo

Use the uploaded photo only for the woman's identity. Keep her face exactly as in the photo: face shape, eyes, eyelids, eyebrows, nose, lips, jawline, hairline, hair colour and skin tone. Do not beautify, slim, smooth or retouch her face; keep its natural asymmetry and texture.

Create a photorealistic travel photograph of the woman in a narrow Kyoto lane with wooden machiya houses at dawn after rain. She wears a beige trench coat, holds a clear umbrella and looks back over her shoulder at the camera, medium distance. Wet stone paving in the foreground reflecting paper lanterns, houses in the midground, a pagoda roofline in soft haze behind. 35mm look, eye level. Cool blue dawn light with warm lantern glow. Film look with fine grain. No other people.

Real skin texture: visible pores, fine vellus hair, slightly uneven natural tone. Unretouched, no skin smoothing, no beauty filter, no glamorization. No text, no logos, no watermark.

Upload one clear photo of yourself, then paste the prompt.

Woman at the top of a deadlift in a garage gym with chalk on her handsNo. 05GPT Image 2.5 · photo edit

Fitness action shot

Use the uploaded photo only for the woman's identity. Keep her face exactly as in the photo: face shape, eyes, eyelids, eyebrows, nose, lips, jawline, hairline, hair colour and skin tone. Do not beautify, slim, smooth or retouch her face; keep its natural asymmetry and texture.

Create a photorealistic candid photograph of the woman at the top of a deadlift in a garage gym, standing tall with the barbell at hip height, gripping it just outside her thighs. One identical black rubber plate on each end of the bar, placed symmetrically at equal distance from her hands. Chalk on her hands and on the bar. She wears a grey T-shirt and black leggings, hair in a high ponytail with flyaways. Full body visible, feet included, slightly low angle, 28mm look. Hard daylight from an open garage door on the left. Sweat on her temples, flushed cheeks, focused expression. Not a glossy fitness advert.

Real skin texture: visible pores, fine vellus hair, slightly uneven natural tone. Unretouched, no skin smoothing, no beauty filter, no glamorization. No text, no logos, no watermark.

Upload one clear photo of yourself, then paste the prompt.

YouTube thumbnail of a surprised woman next to a laptop with the text I TRIED ITNo. 06GPT Image 2.5 · photo edit

YouTube thumbnail with text

Use the uploaded photo only for the woman's identity. Keep her face exactly as in the photo: face shape, eyes, eyelids, eyebrows, nose, lips, jawline, hairline, hair colour and skin tone. Do not beautify, slim, smooth or retouch her face; keep its natural asymmetry and texture.

Create a 16:9 YouTube thumbnail, photorealistic. The woman, large and close on the left third, with a genuinely surprised expression, mouth open, one hand on her cheek. On the right third, an open laptop showing a simple rising green line chart. Headline text "I TRIED IT" rendered exactly once in heavy white sans-serif with a thick black outline, top right, large and legible at small size. High contrast, simple deep-blue gradient background, soft key light from the front-left. No other text.

Real skin texture: visible pores, fine vellus hair, slightly uneven natural tone. Unretouched, no skin smoothing, no beauty filter, no glamorization. No logos, no watermark.

Upload one clear photo of yourself, then paste the prompt.

Nine-frame 3x3 contact sheet of candid flash photos of a woman at home, each ruined by a classic amateur mistakeNo. 07GPT Image 2.5 · photo edit

Almost-perfect photo dump

Use the uploaded photo only for the woman's identity. Keep her face exactly as in the photo: face shape, eyes, eyelids, eyebrows, nose, lips, jawline, hairline, hair colour and skin tone. Do not beautify, slim, smooth or retouch her face; keep its natural asymmetry and texture.

Create a vertical contact-sheet collage of nine candid amateur photos in a clean 3x3 grid with thin black dividers: near-miss snapshots that almost became great portraits. All nine show the woman, long glossy dark hair worn loose, in the same outfit and the same cosy city apartment, shot by a friend just before a night out. Outfit: a fitted charcoal ribbed tank top, a high-waisted black denim mini skirt, delicate layered necklaces and canvas sneakers. The apartment has a neutral fabric sofa, a bookshelf, a warm glowing floor lamp and a full-length mirror. Look: direct on-camera flash, the slightly grainy colour of a compact CCD digital camera, subtle noise, occasional motion blur. Each frame has a different pose and a different classic mistake: 1) she turns mid-step and her face smears with motion blur; 2) half-body, leaning on the wall with a hand on her hip, blown out by a flash fired too close; 3) sitting on the sofa arm with legs crossed, laughing and blinking at the exact moment; 4) the autofocus locks on the picture frames behind her, leaving her soft; 5) full-body standing shot with the camera tilted about 20 degrees so the door frame leans; 6) glancing back over her shoulder while the photographer's blurry thumb covers the lower right third; 7) the flash didn't fire: dark, moody and noisy, only her seated silhouette readable; 8) aimed too low, cropping her head at eye level; 9) standing and smiling while the big lampshade behind her lines up like a giant hat. Authentic and effortless, never staged. No timestamps, no camera interface.

Real skin texture: visible pores, fine vellus hair, slightly uneven natural tone. Unretouched, no skin smoothing, no beauty filter, no glamorization. No text, no logos, no watermark.

Upload one clear photo of yourself, then paste the prompt.

Food and interior prompts

Text-to-image prompts with no reference photo. The detail is in materials, props and light.

Bowl of tonkotsu ramen with chashu, soft egg and nori, steam backlit by a windowNo. 08GPT Image 2.5 · text-to-image

Food photo

Photorealistic 45-degree food photograph of a bowl of tonkotsu ramen: cloudy pork broth with fat droplets, chashu slices with charred edges, a halved soft-boiled egg with a jammy yolk, scallions and a sheet of nori, on a dark wooden counter with chopsticks resting on the rim. Steam rising, a few broth splashes on the counter. Backlight from a window behind the bowl making the steam glow. 50mm look, shallow depth of field. Natural colour, no oversaturation.

No text, no logos, no watermark.

No photo needed. Paste the prompt as is.

Open-plan loft kitchen and living room with exposed brick and steel windowsNo. 09GPT Image 2.5 · text-to-image

Real-estate interior

Photorealistic real-estate interior photograph of an open-plan kitchen and living room in a renovated loft: exposed brick wall, white oak floors, a matte black kitchen island with three stools, a linen sofa facing tall steel-framed windows. Wide lens, camera at chest height, vertical lines perfectly straight. Bright overcast daylight through the windows balanced with warm pendant lights over the island. Lived-in but tidy: a bowl of lemons, an open book on the sofa. Accurate proportions, no fisheye distortion. No people.

No text, no logos, no watermark.

No photo needed. Paste the prompt as is.

Text, poster and logo prompts

Where GPT Image 2.5 pulls ahead: layouts with exact words. Quote the text, give it a position and ban everything else.

Vintage-style travel poster of a pagoda at sunset with the word KYOTONo. 10GPT Image 2.5 · text-to-image

Vintage travel poster

Create an original vintage-style travel poster: a five-storey pagoda silhouetted against a large orange sun, layered blue hills and a few flying cranes, flat screen-print shapes with subtle paper grain. The word "KYOTO" rendered exactly once in tall cream Art Deco capitals across the bottom, spelled K-Y-O-T-O. Limited palette of deep teal, burnt orange and cream. No other text.

No logos, no watermark.

No photo needed. Paste the prompt as is.

Why these prompts work

Three rules behind every prompt on this page

  1. 01

    State the use first

    "Product listing photo", "YouTube thumbnail", "candid phone photo". The intended use sets the level of polish before any detail is read.

  2. 02

    Quote every word of text

    Put exact text in quotes, say where it goes and that it appears once, then add "no other text". Always proofread the output.

  3. 03

    End with exclusions

    No logos, no watermark, no extra people. GPT Image 2.5 follows negative instructions well when they come last and are specific.

How to write your own

How to write a ChatGPT image prompt that works first time

OpenAI's own guidance boils down to one order. Write one line per slot and skip the slots you do not need.

  1. 1

    Medium and use

    "Photorealistic candid photo for an Instagram post" or "flat vector infographic". This sets the look before anything else.

  2. 2

    Subject and action

    Who or what, with visible details, what they are doing and where they look.

  3. 3

    Setting and composition

    Place, time of day, two or three props, then shot size, angle and lens.

  4. 4

    Light and texture

    Light direction and colour, then materials: skin pores, fabric creases, glaze speckle, steam.

  5. 5

    Text and exclusions

    Exact words in quotes with position and count, then what must not appear: other text, logos, watermarks, extra people.

FAQ

Frequently Asked Questions

Run these prompts on your own photos

Sign up to use GPT Image 2.5, Nano Banana Pro and Seedance 2.5 in one studio, with your AI character saved for every shot.

Start creating in minutes · Pay as you go