AI Art Style Prompts: Turn One Photo Into 5 Illustration Styles With ChatGPT

By Clayton H · Published 7 November 2026 · Tools: ChatGPT, HeyGen

You can turn one portrait into a completely different art style by uploading it to ChatGPT and pasting a style prompt that tells the image generator what to keep (framing, pose, expression, hand gestures) and what to redraw (face, clothing, background, palette). This guide gives you five prompts: Superflat pop art, a 1920s Constructivist poster, 80s fashion illustration, Matisse paper cut-outs and Persian miniature.

To use them, open ChatGPT, upload your original photo, paste one prompt as it is and generate. Run two or three variations and keep the one where the style reads best. If you want the image to talk, upload the styled frame to an avatar tool such as HeyGen and add your script.

What is AI style transfer?

AI style transfer means asking an image model to redraw an existing photo in a new art style while keeping its composition, pose and expression. The difference between a good result and a “photo with a filter” is the prompt: it has to give the model permission to redesign everything except the pose, and it has to describe the style in concrete terms (shapes, line quality, palette, texture) rather than only naming it.

What you get

• Five copy-paste prompts, one per style, each written to work on any portrait.
• Exact color palettes with hex codes inside each prompt, so results stay on style.
• A fill-in outfit line in each prompt, so you can keep the suggested look or swap in your own.
• Photo tips for the cleanest results, plus how to animate a styled frame.

What you need

• ChatGPT with image generation. OpenAI says ChatGPT Images is available on all tiers; usage limits depend on your plan.
• One clear portrait photo. A talking-head or half-body shot works best.
• Optional: an avatar video tool such as HeyGen, if you want the styled image to speak.

How to turn a photo into an illustration with ChatGPT

1. Open a new chat in ChatGPT.
2. Upload your original photo. Use the original every time, never a result from another style, or the styles start to blend.
3. Copy one full prompt from the style library below. Keep the outfit line as it is or replace the text inside the brackets with your own outfit.
4. Paste the prompt and generate.
5. Generate two or three variations and keep the one where the style is strongest and the face still reads clearly.
6. Start a new chat for the next style, with the original photo again.

How to choose the right photo

• Clear face, facing the camera. The prompts keep the gaze and expression, so they need to be readable.
• Visible hands. An open, natural gesture gives every style a stronger shape to redraw.
• Simple pose and background. The background gets redrawn anyway; a cluttered one only confuses the model.
• Mouth closed, if you plan to animate. Talking-photo tools lip-sync more reliably from a relaxed, closed mouth.

The 5 AI art style prompts

1. Superflat pop art

Japanese pop art: rounded cartoon features, flat candy colors and a dense field of blossoms under a rainbow.

Transform the provided photo into a wildly colorful Japanese superflat pop-art illustration. Keep the framing, camera angle, pose, gaze, facial expression and hand gestures of every person in the photo, and keep each person’s hairstyle, hair color and eye color recognizable. Keep any props in the photo in the same place, redrawn in the same style. Freely redesign faces, clothing, objects and surroundings; the style matters more than photographic likeness, but every face must stay readable. Turn each person into a rounded, charming cartoon character with enlarged glossy eyes, a tiny nose, clean outlines and thick sculpted cartoon hair. Dress them in [OUTFIT: their original clothing in the same cut, covered in an oversized graphic cherry-blossom print with a chunky bow or bold collar]. Fill the background with a dense, layered field of original cartoon cherry blossoms, stars and puffy clouds at different scales under a bold rainbow arc, with original designs that copy no existing artist’s characters. Keep a calmer area around each face so it reads instantly. Palette: hot pink #FF4FA3, sunshine yellow #FFD400, royal blue #2450FF, turquoise #22C7C7, white, and deep navy outlines #1A1F4D. Smooth flat color planes, almost no shading or depth. Keep the aspect ratio of the source photo. No realistic portrait, painted skin, shadows, 3D rendering, text or logos.

2. Constructivist poster

1920s graphic poster: a huge red circle, black diagonals, angular figures and stencil-printed color.

Transform the provided photo into a bold 1920s Constructivist graphic poster. Keep the framing, camera angle, pose, gaze, facial expression and hand gestures of every person in the photo, and keep each person’s hairstyle, hair color and eye color recognizable. Keep any props in the photo in the same place, redrawn in the same style. Freely redesign faces, clothing, objects and surroundings; the style matters more than photographic likeness, but every face must stay readable. Build each person from angular silhouettes and a few hard-edged shadow planes. Simplify faces into strong black, cream and red shapes while keeping the eyes and mouth expressive. Dress them in [OUTFIT: a sharp double-breasted coat with monumental, exaggerated shoulders and a high collar]. Replace the surroundings with a huge red circle behind the main figure, black diagonal beams crossing the frame, cream triangular cut-outs and simplified industrial forms at the edges. Palette: vermilion #E03C1F, charcoal black #151515, warm cream #EFE6D2 and small areas of gray #8C8C8C. Aggressive diagonals, flat stencil color, slight ink wear and rough printed edges. Keep the aspect ratio of the source photo. No photographic face, painterly shading, halftone dots, 3D graphics, text, letters or logos.

3. 80s fashion illustration

Bold 1980s fashion-magazine drawing in the manner of Antonio Lopez: marker and gouache, high-shine features, power shoulders.

Transform the provided photo into a bold 1980s fashion-magazine illustration in the manner of Antonio Lopez. Keep the framing, camera angle, pose, gaze, facial expression and hand gestures of every person in the photo, and keep each person’s hairstyle, hair color and eye color recognizable. Keep any props in the photo in the same place, redrawn in the same style. Freely redesign faces, clothing, objects and surroundings; the style matters more than photographic likeness, but every face must stay readable. Sculpt each face with a few high-shine marker planes: graphic cheekbones, a defined brow and crisp highlights. Draw the hair as glossy shapes with white highlight streaks. Dress them in [OUTFIT: a power-shouldered fuchsia jacket with sharp exaggerated shoulders, plus bold gold jewelry such as statement earrings or a chunky cuff]. Reduce the surroundings to energetic sweeping marker strokes and color blocks that still hint at the original setting. Palette: fuchsia #E6007E, cobalt #1F4BD8, sunflower #FFC20E, black and bright white paper. Confident marker and gouache strokes, crisp highlights, visible pencil underdrawing in places. Keep the aspect ratio of the source photo. No photographic skin, soft gradients, 3D rendering, text or logos.

4. Matisse paper cut-outs

Hand-cut, painted paper shapes in the manner of Henri Matisse’s late cut-outs: ultramarine, coral fronds and stars on white.

Transform the provided photo into a composition made entirely of hand-cut, gouache-painted paper shapes in the manner of Henri Matisse’s late cut-outs. Keep the framing, camera angle, pose, gaze, facial expression and hand gestures of every person in the photo, and keep each person’s hairstyle, hair color and eye color recognizable. Keep any props in the photo in the same place, rebuilt as cut paper. Freely redesign faces, clothing, objects and surroundings; the style matters more than photographic likeness, but every face must stay readable. Build each face from a single cut paper shape, with the eyes, brows and mouth as small, precise cut pieces so the expression stays clear. Make the hair one bold cut shape in the person’s hair color. Dress them in [OUTFIT: large, simple cut-paper shapes in ultramarine that keep the original clothing’s silhouette]. Fill the background with large seaweed fronds, stars and abstract coral shapes on white paper. Palette: ultramarine #2340A8, coral #F2694B, lemon #F7D23E, leaf green #2E8B57, black and white paper. Visible scissor-cut edges, slightly uneven painted surfaces, small overlaps and tiny pinholes. Keep the aspect ratio of the source photo. No outlines, shading, gradients, photographic detail, 3D rendering, text or logos.

5. Persian miniature

Illuminated manuscript painting: jewel-toned palace garden, intricate embroidery, flattened perspective and a gilded border.

Transform the provided photo into a richly detailed Persian miniature painting from an illuminated manuscript. Keep the framing, pose, gaze, facial expression and hand gestures of every person in the photo, and keep each person’s hairstyle, hair color and eye color recognizable. Keep any props in the photo in the same place, restyled as ornate objects that belong in the miniature. Freely redesign faces, clothing, objects and surroundings; the style matters more than photographic likeness, but every face must stay readable. Paint each person as a graceful manuscript figure with a delicate oval face, almond eyes, fine arched brows and carefully patterned hair. Dress them in [OUTFIT: a flowing silk robe with intricate floral embroidery, contrasting brocade cuffs and a small jeweled cap]. Replace the surroundings with a jewel-toned palace garden: pointed tiled arches, patterned walls, flowering branches, slender cypress trees and small birds, arranged in shallow stacked planes with flattened perspective. Palette: lapis blue #1F3C88, turquoise #3FB6B2, vermilion #D9472B, emerald #1E7B4F, warm ivory #F3EBD6 and luminous gold #D4A72C. Fine floral and geometric patterns, delicate ink contours, tiny gold accents and a narrow illuminated border, painted on aged manuscript paper. Keep the aspect ratio of the source photo. No photographic skin, realistic lighting, depth of field, readable script, text or logos.

How to animate your styled image

1. Open a talking-photo tool such as HeyGen and choose the option to create an avatar or talking video from a photo.
2. Upload the styled frame from ChatGPT.
3. Type your line or upload a voiceover, then generate.
4. Repeat for each style, keeping the same voice so the clips cut together cleanly.

HeyGen’s talking-photo page talks about photos rather than illustrations. Every illustrated frame made with these prompts animated without problems in testing, but test one image before you batch the rest.

Things worth knowing

• One style per generation. Mixing two prompts in one run gives you a muddy result that is neither.
• Likeness is deliberately secondary. The stronger the style, the less photographic the face. The prompts keep hairstyle, hair color and eye color so you still read as you.
• If it looks like a filter, run it again. The prompts push the model to redraw the whole frame, but some runs come back too close to the photo.
• Hands and small details can drift. Check fingers and props, and regenerate if something looks off.
• Text sometimes sneaks in. Every prompt says no text or logos; if letters still appear, regenerate rather than editing them out.
• Artist-inspired styles. Naming Antonio Lopez or Matisse steers the look; the results are new images, not copies of their work. If you plan to use the images commercially, check your image tool’s terms.

FAQ

Will the result still look like me?

Partly. The prompts keep your pose, expression, hairstyle, hair color and eye color, but they redraw the face in the style, so expect a stylized version of you rather than a portrait.

Do these prompts work on group photos?

Yes. Every prompt is written for “every person in the photo”, so each person gets the same treatment. They were tested on single portraits, so try one run before you batch a group shot.

Can I change the outfit?

Yes. Replace the text inside the [OUTFIT: …] brackets with your own description, or leave it as it is for the suggested look.

Do the prompts work in other image generators?

They were written and tested in ChatGPT. Other image models that accept a reference photo can follow them, but results vary, so test one style first.