ChatGPT image prompts.

80 copy-paste prompts for portraits, profile pictures, couple photos, devotional art, illustration and product shots. Plus the line that stops ChatGPT quietly turning your face into someone else's.

GPTPrompts.AI Editorial

Prompts written around the controls that actually change the output: lens, light and texture · Last updated August 2026

Why most photo prompts disappoint

Nearly every prompt list is a wardrobe and a location: put me in a suit, put me on a beach. That produces a generic image with a face that is nearly yours and skin like polished plastic.

The prompts here add the three controls a photographer would actually reach for. What must not change about you. Which direction the light comes from and how hard it is. And an explicit refusal of skin smoothing, which is the fastest way for an image to announce that it was generated.

Portraits from your own photo

Attach your photo first, then paste one of these. Each begins with the preservation instruction, which is the line most prompt lists leave out and the reason their results look like someone else.

Prompt 1

Using the uploaded photo as reference, keep the facial features, skin tone and hairstyle exactly as they are. Create a photorealistic portrait: 85mm lens at f/2.0, soft window light from the left with gentle falloff into shadow on the right, plain warm grey background slightly out of focus, natural skin texture with visible pores and no smoothing, catchlights in the eyes. 4:5.

Prompt 2

From the uploaded photo, preserve the face precisely. Environmental portrait: standing near a large window in a quiet room, 35mm lens, overcast daylight, muted palette, relaxed posture, looking slightly away from camera, fine grain, no retouching. Photorealistic. 4:5.

Prompt 3

Using the uploaded photo, face unchanged. Cinematic portrait with a single warm key light from the upper right and deep shadow opposite, dark background, 85mm, shallow depth of field, calm serious expression, subtle film grain. No text. 9:16.

Prompt 4

From the uploaded photo, keep the likeness identical. Golden hour outdoor portrait, low sun behind creating a rim light on the hair, warm haze, blurred greenery behind, 135mm compression, natural skin with slight shine. Photorealistic. 4:5.

Prompt 5

Using the uploaded photo and preserving the face exactly, create a black and white portrait: plain dark top, dramatic side light, deep blacks, high contrast, visible skin texture, direct eye contact. Monochrome, photorealistic, 85mm. 1:1.

Profile pictures and headshots

For LinkedIn, work profiles and anywhere you need to look like yourself on a Tuesday. The brief is deliberately unglamorous.

Prompt 1

Using the uploaded photo and keeping my face exactly as it is, create a 1:1 profile picture: head and shoulders, facing camera, soft even frontal light, plain mid-grey background with a gentle gradient, natural skin texture with no beauty filter, relaxed slight smile, sharp focus on the eyes. 1:1.

Prompt 2

From the uploaded photo, preserve the face precisely. Corporate headshot: mid-chest crop, well-fitted navy blazer over a plain shirt, soft light from the front left, plain light background, 85mm at f/2.8, approachable expression, natural skin retained. 1:1.

Prompt 3

Using the uploaded photo with the face unchanged, create a modern office headshot: standing in a bright workspace with blurred glass and greenery behind, natural daylight, smart casual dress, confident relaxed posture. 85mm, shallow depth of field. 4:5.

Prompt 4

From the uploaded photo, face identical. Outdoor professional portrait in soft open shade, blurred city background, smart jacket, no harsh shadows, friendly expression, suitable for a company website. 85mm, photorealistic. 4:5.

Couple and family photos

Harder, because models blend features between adjacent faces. Name each person by position and say explicitly that both faces must hold.

Prompt 1

Using the uploaded photo of two people, keep both faces exactly as they are: the person on the left and the person on the right must each retain their own facial features, skin tone and hairstyle with no blending between them. Photorealistic portrait standing close together, soft golden hour backlight, blurred open field behind, coordinated neutral clothing, natural and unposed. 85mm at f/2.0. 4:5.

Prompt 2

From the uploaded couple photo, preserve both faces precisely and separately. Warm candid moment: the two standing forehead to forehead, eyes closed, gentle smiles, soft window light indoors, muted tones, shallow depth of field, tender and quiet mood. Photorealistic, 85mm. 4:5.

Prompt 3

Using the uploaded couple photo and keeping each face identical, create an evening portrait on a city rooftop, blurred lights behind, warm practical lighting, smart eveningwear, intimate mood, shallow depth of field. Photorealistic. 9:16.

Prompt 4

From the uploaded family photo, keep every face exactly as in the original. Everyone in coordinated neutral clothing on a home terrace at warm evening light, candid rather than posed, natural expressions. Photorealistic, no text. 4:5.

Devotional and festival art

For Janmashtami, Diwali and similar occasions. These generate a scene rather than editing your face, which is both more reliable and more appropriate for devotional imagery.

Prompt 1

Create a serene devotional illustration of Lord Krishna as a young cowherd playing the flute beneath a kadamba tree at dawn, peacock feather in the hair, soft blue skin tone in the classical style, warm golden light filtering through leaves, calm and reverent mood, rich traditional colour palette. Painterly illustration, no text. 4:5.

Prompt 2

Generate a photorealistic Janmashtami still life: a small ornate cradle with a peacock feather and a flute resting beside it, marigold garlands, a lit diya, soft warm light from a single source, deep shadows, quiet devotional atmosphere. No text. 1:1.

Prompt 3

Create a traditional Indian miniature-style painting of Krishna with a herd of cows at sunset by a river, fine detailed linework, flat rich colours with gold accents, decorative border, classical composition. Illustration, no text. 4:5.

Prompt 4

Generate a warm festive scene for Diwali: a doorway framed with marigold garlands and rows of lit diyas, rangoli on the threshold, soft evening light, no people, inviting and peaceful. Photorealistic, no text. 9:16.

Art styles and illustration

For images that are not meant to look photographic. Naming a medium and its physical properties works far better than naming an artist.

Prompt 1

Create a loose watercolour illustration of a quiet coastal town at dawn, visible paper texture, colours bleeding softly at the edges, limited palette of blue-grey and warm ochre, generous white space, gentle and unhurried mood. No text. 4:5.

Prompt 2

Generate a bold flat-design vector illustration of a city skyline at sunset, simplified geometric shapes, limited warm palette, clean edges, modern editorial style, empty space in the upper left for a headline. No text in the image. 16:9.

Prompt 3

Create a detailed pen and ink illustration of an old library interior, dense cross-hatching, dramatic light from a high window, high contrast, engraved book-plate style, monochrome. No text. 4:5.

Prompt 4

Generate a soft pastel children's book illustration of a small fox reading under a glowing mushroom at night, gentle textures, warm cosy palette, bokeh fireflies, whimsical and safe mood. No text. 1:1.

Prompt 5

Create a retro risograph-style poster illustration: two-colour print in coral and deep blue, visible misregistration and grain, simplified shapes, bold composition, 1970s design sensibility. No text in the image. 4:5.

Prompt 6

Generate an isometric pixel art scene of a tiny cluttered workshop, limited 16-colour palette, crisp pixel edges with no anti-aliasing, warm lamp light, lots of small readable details. No text. 1:1.

Product, mockup and social images

Practical output. The trick is leaving deliberate empty space, because you will almost always add text over it later.

Prompt 1

Create a studio product shot of [product] on a soft gradient background, gentle reflection below, crisp rim lighting, generous clean negative space at the top for a headline. Photorealistic commercial photography. 1:1.

Prompt 2

Generate a flat lay of a workspace from directly above: laptop, notebook, coffee cup and a plant on a light wooden desk, soft even daylight, muted natural palette, empty space in the lower third. Photorealistic, no text. 4:5.

Prompt 3

Create a lifestyle product image: [product] held in natural hands in soft window light, blurred home interior behind, warm authentic feel rather than studio-perfect, shallow depth of field. Photorealistic. 4:5.

Prompt 4

Generate a minimal social media background: soft abstract gradient in muted teal and sand, subtle grain, gentle light falloff toward the corners, entirely empty centre for text to be added later. No text in the image. 9:16.

More ChatGPT Images 2.5 prompts

These additional prompts are deliberately specific about what to preserve, what to change and what to leave for a normal design tool. Replace bracketed details, attach a reference when the prompt says supplied or uploaded, and inspect the result before publishing it.

Prompt 28

Using the uploaded photo, preserve my face and hair exactly. Change only the outfit to a cream linen shirt, place me beside a bright window, natural skin texture, 4:5.

Prompt 29

Using the uploaded image, keep the person unchanged and replace only the background with a quiet library, matching the original light direction and perspective. 4:5.

Prompt 30

Create a realistic passport-style headshot from the uploaded photo: face unchanged, neutral expression, even light, plain pale background, no retouching. 1:1.

Prompt 31

Create a warm candid family photograph from the supplied references. Preserve every face separately, natural expressions, soft evening light, no invented children. 4:5.

Prompt 32

Create a clean ecommerce image of [product] on white with a soft contact shadow, accurate proportions, no label, no extra objects. 1:1.

Prompt 33

Create a premium product image of [product] on dark slate with one narrow rim light, realistic material texture and empty space on the right. 4:5.

Prompt 34

Create a flat-lay recipe image with [ingredients], top-down camera, clear separation, natural shadows and an empty lower band for typography. 4:5.

Prompt 35

Create a bright social graphic background in coral and navy with an empty centre, subtle paper grain, no letters, numbers or logos. 1:1.

Prompt 36

Create a simple editorial illustration of a team planning around a table, diverse adults, clear gestures, limited palette, no text. 16:9.

Prompt 37

Create a paper-cut collage of a moonlit garden, layered edges, restrained blue and cream palette, no text. 4:5.

Prompt 38

Create a cinematic still of a fictional detective in a rainy station, one practical light source, realistic wet surfaces, no readable signs. 16:9.

Prompt 39

Create a friendly classroom illustration with a teacher and adult learners, clear whiteboard left blank, warm daylight, no text. 16:9.

Prompt 40

Restore the supplied black-and-white photo by removing scratches only. Preserve grain, faces, clothing, crop and age; do not add colour. 4:5.

Prompt 41

Colourise the supplied archival photo subtly while preserving faces and original clothing patterns. Keep the result documentary rather than glossy. 4:5.

Prompt 42

Replace only the sky in the uploaded landscape with a soft overcast sky. Preserve buildings, trees, shadows and horizon exactly. 16:9.

Prompt 43

Turn the supplied room into a calm reading nook by changing only the wall colour and adding one lamp. Preserve furniture placement and scale. 4:5.

Prompt 44

Create a transparent-background sticker of a cheerful fictional cat, bold clean outline, limited colours, no words. 1:1.

Prompt 45

Create a realistic botanical study of a hibiscus flower on a plain background, accurate petals and leaves, soft daylight, no labels. 4:5.

Prompt 46

Create a minimal architectural illustration of a small courtyard house, clean perspective, trees for scale, muted palette, no signage. 16:9.

Prompt 47

Create a retro travel-poster composition of a fictional coastal town, simplified shapes, two-colour print texture, no words or flags. 4:5.

Prompt 48

Create four consistent storyboard frames of a courier leaving home, cycling, delivering a parcel and waving. Same clothes, bag and face throughout. 16:9.

Prompt 49

Create a close-up portrait of a fictional older woman with natural wrinkles, soft side light and a plain background. No beauty filter. 4:5.

Prompt 50

Create a realistic portrait of a fictional man in open shade, visible skin texture, relaxed posture, 85mm look, no logos. 4:5.

Prompt 51

Create a quiet still life of a notebook, fountain pen and tea on a wooden desk, morning window light, empty upper third, no text. 4:5.

Prompt 52

Create a professional presentation background with abstract blue shapes at the edges and an empty white centre. No letters, numbers or icons. 16:9.

Prompt 53

Create a clean app onboarding illustration of a person checking a calendar, simple geometry, high contrast, accessible colours, no words. 4:5.

Prompt 54

Create a fictional street-food stall at dusk with warm lights and steam, believable surfaces, no brand signs or readable menus. 16:9.

Prompt 55

Create a cosy winter interior with a chair, blanket and reading lamp, realistic scale, quiet composition and no people. 4:5.

Prompt 56

Create an abstract gradient with a clear diagonal light path and generous negative space, subtle grain, no text. 1:1.

Prompt 57

Create a macro photograph of a green leaf after rain, crisp droplets, natural veins, soft background blur and no fantasy elements. 4:5.

Prompt 58

Create a realistic ceramic mug mockup with a blank white surface, three-quarter angle, soft studio light and transparent background. 1:1.

Prompt 59

Create a square album-cover concept for an instrumental record using geometric shapes and deep violet, but no letters or numbers. 1:1.

Prompt 60

Create a hand-drawn map illustration of a fictional park with paths and empty label boxes, no real place names. 4:5.

Prompt 61

Create an infographic-style illustration of a recycling workflow using empty circles and arrows, clean spacing, no text. 16:9.

Prompt 62

Create a natural outdoor portrait of a fictional couple, each face distinct, late-afternoon light, candid posture, no wedding clichés. 4:5.

Prompt 63

Create a fashion editorial image of a fictional model in a structured black coat, plain studio, directional light and realistic fabric folds. 4:5.

Prompt 64

Create a product comparison layout with three blank cards and one object in each, aligned on a neutral background, no labels. 16:9.

Prompt 65

Create a soft children's-book illustration of a small fox crossing a stream, visible brush texture, calm colours and no words. 4:5.

Prompt 66

Create a photorealistic city park scene at dawn with one empty bench in the foreground, mist, natural light and no signage. 16:9.

Prompt 67

Create a high-contrast black-and-white architectural photograph of stairs and shadows, precise geometry, no people and no text. 4:5.

Prompt 68

Create a simple three-panel comic about planting a seed, with empty speech bubbles and consistent character design. 16:9.

Prompt 69

Create a realistic overhead desk scene for a project manager: notebook, calendar, pen and coffee, tidy but lived-in, no readable text. 4:5.

Prompt 70

Create a clean logo exploration sheet using three abstract shapes for a fictional bakery, no letters, no trademarked symbols, white background. 1:1.

Prompt 71

Create a believable rainy-window portrait of a fictional person, face visible through soft reflections, natural skin and subdued colour. 4:5.

Prompt 72

Create a print-friendly botanical border around an empty centre, consistent leaves, two ink colours, no words and no gradients. 1:1.

Prompt 73

Create a realistic photo of a handmade soap bar on a stone tray, soft side light, visible texture, blank wrapper and no brand. 4:5.

Prompt 74

Create a calm sunrise landscape with layered hills, one winding path and soft atmospheric depth, natural colours, no text or landmarks. 16:9.

Prompt 75

Create a realistic close-up of a woven basket with oranges and linen cloth, directional morning light, tactile texture and no labels. 4:5.

Prompt 76

Create a modern editorial portrait of a fictional chef in a plain kitchen, face natural, hands visible, soft side light, no logos or readable signs. 4:5.

Prompt 77

Create a simple geometric pattern for wrapping paper in sage and cream, even spacing, clean edges, no letters, numbers or symbols. 1:1.

Prompt 78

Create a realistic image of a bicycle leaning against a painted wall, late-afternoon shadow, accurate spokes and no brand marks. 4:5.

Prompt 79

Create a cosy café interior with empty tables, warm pendant lights and a clear foreground, believable proportions, no menus or signage. 16:9.

Prompt 80

Create a clean presentation illustration of a bridge between two islands, three neutral colours, bold silhouette, empty space for a later caption. 16:9.

Prompt 81

Create a studio photograph of a folded cotton shirt, natural creases, soft grey background, accurate seams, no logo and no text. 1:1.

Six rules that decide the result

Say what must not change

For any prompt that edits your own photo, an explicit instruction to preserve facial features, skin tone and hairstyle is the difference between a portrait of you and a portrait of someone who resembles you.

Forbid skin smoothing

Ask for natural skin texture with visible pores and no beauty filter. Over-smoothed skin is the single clearest signal that an image was generated, and it is the default behaviour.

Name a lens and an aperture

85mm at f/2.0 for portraits, 35mm for context, 135mm for compressed background blur. This gives the model something concrete to act on, unlike the word cinematic.

Describe light by direction and quality

Soft window light from the left. Hard midday sun overhead. Warm rim light from behind. Direction plus quality produces depth; naming a mood produces flat, evenly lit images.

Ban text unless you need it

Letters come out garbled, especially in non-Latin scripts. Generate the image clean and add wording afterwards where you control the typeface.

Refine with one change at a time

When it is close, reply with the single adjustment rather than rewriting the prompt. The model keeps the rest, so you do not accidentally lose the parts that worked.

Frequently asked questions

Can ChatGPT edit a photo of me?

Yes. Attach the photo and describe the change in plain language, then keep refining across turns. The instruction that matters most is telling it what to leave alone: keep the facial features, skin tone and hairstyle exactly as in the uploaded photo, change only the clothing, lighting and background. Without that line the model treats your face as one more thing it is free to restyle.

Is there a Nano Banana prompt for ChatGPT?

Nano Banana is Google's image model inside Gemini, not an OpenAI model. You can adapt the same plain-language brief for ChatGPT Images, but the output is produced by ChatGPT's current image system. If you need the Gemini model, use Gemini and name the current model only when the product or API actually exposes it.

How do I make a good AI profile picture in ChatGPT?

Ask for a square crop, front-facing, with soft even light and a plain background, and explicitly forbid skin smoothing. Try: using the uploaded photo and keeping my face exactly as it is, create a 1:1 profile picture, head and shoulders, soft frontal light, plain mid-grey background, natural skin texture with no beauty filter, relaxed slight smile, sharp focus on the eyes.

Why do my ChatGPT images look obviously AI-generated?

Usually three things at once: over-smoothed skin, symmetrical lighting with no shadow direction, and a background with no depth. Fix all three in the prompt. Ask for natural skin texture with visible pores, name a single light source and its direction, and specify a lens and aperture such as 85mm at f/2.0 so the background falls off naturally.

Why is the text in my generated image gibberish?

Image models are unreliable at rendering letters, and non-Latin scripts fare worse. Add this to any prompt: do not render any text, letters or numbers anywhere in the image. Then add wording afterwards in a normal editor, where you control the typeface and the spelling.

Can I use ChatGPT-generated images commercially?

Usually yes under OpenAI's terms for content you create, but there are real limits worth knowing: images of identifiable real people, recognisable brands and copyrighted characters carry separate legal considerations that the tool's terms do not settle for you. For anything printed, sold or advertised, check OpenAI's current usage terms and think about the subject matter separately.

What aspect ratio should I ask for?

Name it or you will get a square. Use 1:1 for a profile picture, 4:5 for an Instagram feed post, 9:16 for a story or reel, and 16:9 for a banner or presentation slide. Adding the ratio to the end of any prompt on this page is enough.

How many attempts should I expect?

Three to five for anything you care about. Image generation is non-deterministic, so the same prompt produces different results each run. Generating a few and picking the best is a more effective use of your time than trying to perfect a single prompt.

Keep going

Don't stop here

What to read next

Hand-picked guides our readers explore right after this one.