ChatGPT photo prompts.
35 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.
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.
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.
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.
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.
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.
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.
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.
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.
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.
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.
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.
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.
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.
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.
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.
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.
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.
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.
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.
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.
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.
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.
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.
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.
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.
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.
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.
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 ChatGPT Nano Banana prompt?
Nano Banana is Google's image model inside Gemini, not an OpenAI one, so there is no Nano Banana mode in ChatGPT. The prompts people share under that name are ordinary image prompts and most of them work in ChatGPT too, they will simply be rendered by OpenAI's image model instead. If you specifically want the Nano Banana look, use the Gemini app.
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.