AI Caricature from Photo: ChatGPT & Gemini Prompts

👁 120 reads

Creating an AI caricature from a photo takes about a minute: you upload a photo to ChatGPT or Gemini, paste a prompt, and get a drawing with an exaggerated head, profession details, and props. In early 2026, the “caricature of me and my job” trend exploded, where ChatGPT adds details from what it remembers about you. We tested 4 prompts on one photo and provide them below for you to copy, along with three more, plus pricing comparisons with human caricature artists.

Where the “Caricature of Me and My Job” Trend Started

In February 2026, the prompt “Create a caricature of me and my job based on everything you know about me” spread across Facebook, Instagram, LinkedIn, and TikTok. The Tab called it the first AI trend of the year on February 4, and Forbes (February 6, 2026) noted that while caricatures themselves aren’t new, what’s novel is how well ChatGPT “knows” the user from previous conversations. A TechRadar reviewer described how the model correctly identified his profession, drew books and coffee, and even included a desk mentioned in an old conversation.

Step-by-Step in ChatGPT and Gemini on Mobile

1. Take or choose a photo: face forward, good lighting, ideally head and shoulders. The model transfers glasses and hairstyle, so don’t choose a photo with a hat.
2. In ChatGPT, tap “+”, add the photo, and paste the prompt. In Gemini, select image creation (Nano Banana), add the photo, and paste the same prompt.
3. If the face “drifts,” write in the same conversation: “preserve facial features, eye shape, and smile from the original.”
4. Download the image. For printing, request higher resolution or upscale the file (see below).

Free limits are small and variable: 2026 reports cite around 20 images daily in Gemini and 2-3 in free ChatGPT. Details are in the article about free AI image generators.

Our Test: One Photo, Four Caricatures

The woman in the examples is AI-generated and does not exist. We first generated a regular “phone-style” photo, then pasted it into Nano Banana 2 (Gemini’s image model) with four prompts from the list below, tested in Polish exactly as written. Each generation via OpenRouter cost $0.068.

What we noticed: face, freckles, glasses, and hair color transferred in all four versions. In the profession caricature, the model added a mug with text reading “VETERINARIAN” (Polish for veterinarian), even though we asked for an image without text; the text is correct, but it shows that text bans don’t always work. The pencil version most resembles work from a street artist. The 3D version came out like an animated film character with large eyes but without the oversized head, so if you want a caricature, add “heavily exaggerated head on small body.”

7 Prompts for AI Caricature from Photo

We tested the first four ourselves; we didn’t test the next three, but they’re built on the same structure. All work in ChatGPT and Gemini after adding a photo. Replace items in square brackets with your own details. Note: we ran these tests in Polish, a heavily inflected language, so results may vary in English.

Prompt to copy (1, caricature with job, tested):

Create a caricature of the person from the attached photo and their job. They work as [veterinarian in a small clinic]: around them [dog, cat, stethoscope, coffee mug, and stack of patient files]. Style: colorful digitally drawn caricature, large head on small body, slightly exaggerated smile and glasses, but face must be recognizable. Light background, no text.

Use case: trend version without ChatGPT memory, you describe the profession yourself. Watch out for: model may add text on objects, check spelling.

Prompt to copy (2, pencil like street artist, tested):

Draw a caricature of the person from the attached photo in street caricaturist style: black pencil and charcoal on white paper, large head, slightly exaggerated glasses and curls, small body. Preserve recognizable facial features. No signature and no text.

Use case: classic caricature for framing. Watch out for: replace “glasses and curls” with what distinguishes your face.

Prompt to copy (3, 3D animation, tested without head note):

Transform the person from the attached photo into a 3D animated character in family film style: large eyes, soft lighting, heavily exaggerated head on small body, same hairstyle, glasses, and clothing. Background: [warm kitchen] blurred in background. No text.

Use case: avatar and card. Watch out for: without the head note you get a regular 3D portrait (that’s what happened to us, we added the note after testing).

Prompt to copy (4, for birthday, tested):

Create a caricature for the [50th] birthday of the person from the attached photo: exaggerated head on small body, holding a huge cake with candles shaped like the number [50], surrounded by balloons and confetti, style of colorful illustration like a greeting card. Face must be recognizable. The only text in the image is the number [50].

Use case: gift for 18th, 40th, or 50th birthday. Watch out for: the number came out correctly for us, but add longer wishes later in an editor.

Prompt to copy (5, couple):

Create a joint caricature of two people from the attached photos: first person on left, second on right. Both with large heads on small bodies, holding hands and [riding together on a tandem bicycle]. Preserve facial features, hairstyles, and glasses of each person. Colorful illustration, light background, no text.

Use case: anniversary, wedding. Watch out for: upload two separate photos, not one shared photo, and only with the other person’s consent.

Prompt to copy (6, work farewell):

Create a caricature of the person from the attached photo for a work farewell. They're leaving the [accounting] department for [retirement]: sitting in a beach chair, with a stack of binders and an old calculator nearby. Style: warm, kind caricature, large head, funny but not mean facial expression. No text.

Use case: card from the team. Watch out for: the word “kind” limits exaggeration of features someone might not like.

Prompt to copy (7, trend with ChatGPT memory):

Create a caricature of me and my job based on everything you know about me. Use the face from the attached photo and preserve its features. Colorful illustration, large head on small body, surrounded by objects related to my work and hobbies.

Use case: original trend. Watch out for: only works in ChatGPT with memory enabled, and the result publicly shows what the model knows about you.

More patterns for photo transformations are in the article about prompts for 1980s-style photos and in 6 prompts for cartoon-style photos. How to write your own prompts is explained in the guide to photo transformation prompts.

Caricature as Gift: Printing

Our images were 896 × 1200 pixels. At 300 dpi, that’s a print of about 7.6 × 10.2 cm, or postcard size. For A4 you need around 2480 × 3508 pixels, so generate in 4K (Nano Banana supports this, described in the article how to generate 4K images in Nano Banana) or upscale the finished file. For canvas you need less, since it’s viewed from greater distance. This is our opinion.

AI vs. Human Artist

AI wins on price and number of attempts. A human artist wins where an original on paper and capturing a person’s character, not just features, matters. This is our opinion.

Privacy: What “Everything You Know About Me” Reveals

Forbes (February 9, 2026) warned that with this trend, people give the tool quite a bit of data, and Jake Moore from ESET noted that people join trends faster than they think about what’s behind them. In ChatGPT, you can disable memory in Settings, in the Personalization section, under Memory; for one-time use, a temporary chat that doesn’t read memory and doesn’t go into history is enough. Disabling memory doesn’t delete saved memories, you must delete them manually. Only transform your own photos or photos of people who have consented. What else is better kept to yourself is described in the article what you can safely paste into AI.

Frequently Asked Questions

How do I make a caricature from a photo for free?

Upload a photo to free ChatGPT or Gemini and paste one of the prompts above. The free image limit is small, so immediately specify style and details in the prompt.

Is an AI caricature suitable for printing?

For postcard size, yes. For A4 and larger, generate in 4K or upscale the file to around 2480 × 3508 pixels.

Why doesn’t the caricature look like me?

Usually because of a side photo or one in shadow. Upload a front-facing photo and add “preserve facial features, eye shape, and smile from the original.”

Does ChatGPT need to know me to make a caricature with my job?

No. You just need to describe the profession and objects in the prompt, as in prompt 1.

Can I make a caricature of a coworker?

Only with their consent, especially if the image will go online or on a company card.

Kinetyra