AI Action Figure from Photo: Step-by-Step Guide + Prompts

👁 129 reads

The AI action figure trend lets you upload your photo to ChatGPT or Gemini and get yourself as a toy: an action figure in a collector’s box, a desk statue, or a small chibi figurine. It came back in September 2026 because Nano Banana 2 and ChatGPT Images 2.5 hold faces well and render text correctly. We tested three prompts on one photo, and below we share 8 prompts you can copy. All current trends are collected in our AI photo trends overview.

Where the Action Figure Trend Came From

The trend has had two waves. In April 2025, the action figure in a box (“starter pack”) spread across LinkedIn and TikTok, made in ChatGPT with GPT-4o - Fast Company covered it. In September 2025, a second wave arrived with realistic 3D desk figurines after the Nano Banana (Gemini 2.5 Flash Image) launch. Storyboard18 wrote on September 13, 2025 that politicians and celebrities posting their figurines drove the trend.

In 2026, the format didn’t disappear - it became a template people repeat with each new model. The TikTok page for “action figure starter pack AI” was updated September 28, 2026, and September trend roundups list “toyification” among the most shared styles. We didn’t find data showing this is a new viral phenomenon on the scale of Hotel Lobby - it’s more of a steady, seasonally refreshed trend. That’s our assessment.

How to Make an AI Action Figure Step-by-Step

1. Choose a photo: face forward, good lighting, ideally full body or at least waist-up. The model transfers clothing, so dress how you want the figure to look.
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. In the prompt, replace the name on the box and accessories in square brackets. Short text (1-3 words) comes out correctly; long sentences on the box often fail.
4. If the face “drifted,” write in the same conversation: “keep facial features, glasses and hairstyle from the original.”
5. Download the image. For print or wallpaper, ask for higher resolution.

General principles for transforming photos that work in every trend are described in our guide on how to transform a photo with AI.

Our Test: One Photo, Three Figurines

The man in the examples is AI-generated and does not exist. First we generated a regular “phone” photo, then pasted it into Nano Banana 2 (Gemini’s image model) via OpenRouter with three prompts from the list below, tested in Polish. Each generation cost $0.068 and took about 15 seconds.

What we noticed:

• Boxed figure: face, beard, glasses, navy hoodie and white shirt transferred unchanged. The model added an autumn park from my photo as background inside the box, looking like cardboard printing. The text “TOMEK” and “Edycja limitowana” (Limited Edition) are correct.
• Desk figurine: most realistic, with acrylic base and 3D model on monitor. Downside: English text “1:7 SCALE FIGURE” and “THE DEVELOPER” appeared on the box, which we didn’t request. If you want a clean box, add “no text on the box.”
• Chibi: large head, round glasses and beard are immediately recognizable. Best variant for avatars and stickers.

8 AI Action Figure Prompts

We tested the first three ourselves. The rest follow the same pattern but weren’t tested. English versions are useful in Higgsfield and tools that understand English better. Insert your own data in square brackets.

Prompt to copy (1, action figure in box, tested in Polish):

Turn the person in the attached photo into a realistic 1:6 scale action figure packaged in a collector's box with a transparent plastic window and cardboard backing. The figure has the same face, hairstyle, beard, glasses and clothing as in the photo. Next to the figure, in separate compartments: [coffee mug, laptop, backpack and bicycle]. At the top of the box text "[NAME]", below it smaller "Limited Edition". Product photo on bright background, soft studio light.

Use for: classic “starter pack” for posts. Watch out: maximum 4 accessories, with more the model starts merging them.

Prompt to copy (2, desk figurine, tested in Polish):

Create a 1/7 scale collectible figurine of the person in the attached photo, realistic style. The figurine stands on a computer desk on a round transparent acrylic base with no text. The monitor screen shows a 3D model of the same figurine in modeling software. Next to the monitor stands a cardboard box in collectible toy style with a printed illustration of this character. Keep the face, glasses and clothing from the original.

Use for: version from the September 2025 Nano Banana wave. Watch out: add “no text on the box” - English text appeared in our test.

Prompt to copy (3, chibi, tested in Polish):

Turn the person in the attached photo into a small vinyl figurine in chibi style: large head, small body, matte plastic, same hairstyle, beard, glasses and [navy hoodie]. The figurine stands on a wooden shelf next to books, blurred room in background, warm lamp light. No text.

Use for: avatar, sticker, card. Watch out: change the clothing color in brackets to yours.

Prompt to copy (4, figure with profession):

Turn the person in the attached photo into an action figure in a box in 90s toy style. Works as a [nurse], so in the box are: [stethoscope, coffee mug, ID badge and tablet]. On the box text "[NAME]" and "[Series: Hospital]". Keep face, hairstyle and glasses. Bright background, product photo.

Prompt to copy (5, couple in one box):

Make two action figures in one collector's box from two attached photos: person from first photo on left, from second on right. Each has their own face, hairstyle and clothing. Between them accessories: [two mugs, tandem bicycle, map]. At top text "[NAME AND NAME]". Bright background, soft light.

Use for: anniversary, gift. Upload two separate photos and only with the other person’s consent.

Prompt to copy (6, EN, blister pack):

Turn the person in the attached photo into a realistic 1:6 scale action figure sealed in a collectible blister pack with a clear plastic window and a printed cardboard backing. Keep the exact same face, hairstyle, glasses and clothing. Accessories in separate slots: [coffee mug, laptop, backpack, bicycle]. Title on top: "[NAME]", subtitle "Limited Edition". Clean product photo, soft studio light.

Prompt to copy (7, EN, desk figurine):

Create a 1/7 scale collectible figurine of the person in the attached photo, realistic style, standing on a computer desk on a round clear acrylic base with no text. The monitor shows the 3D model of the same figurine in modeling software. Next to it, a toy-style box printed with an illustration of the character. Keep the face, glasses and outfit from the original. No text on the box.

Prompt to copy (8, EN, chibi in windowed box):

Turn the person in this photo into a 3D chibi-style vinyl toy with oversized head and expressive eyes, same hairstyle, glasses and outfit, placed inside a pastel display box with a transparent window. Matte plastic texture, soft light, no text.

Which Tools and Free Options

Limits: Gemini help page (as of October 3, 2026) confirms Nano Banana 2 image generation works without a plan but no longer states the number of images - only writes that multimedia creation uses more of the general limit. In March 2026 the same table mentioned 20 images daily without a plan and 100 in AI Pro. From October 9, free Gemini chat switches to Flash-Lite model alone. Google mentions chat models but nothing about images, so we’ll check the image limit after the change. Details are in our articles on Nano Banana and GPT Image.

If you want to turn the figurine into a short video (hand picks up figurine from desk, rotation like in a toy review), Higgsfield has ready-made video effects for this. According to Higgsfield’s blog from September 21, 2026, plans cost $15 (Starter), $49 (Plus) and $129 (Ultra) monthly, and a Nano Banana 2 image is 2 credits, about $0.10. More about the platform is in our Higgsfield article.

Common Mistakes

• Too much text on the box: name and one subtitle line is the safe maximum.
• Side photo or wearing a hat: the figurine gets someone else’s face or the hat permanently.
• Real toy brand logos in the prompt: ChatGPT and Gemini often refuse or transform them into similar ones. Write “box in collectible toy style.”
• Others’ photos without consent: show a coworker’s figurine to them privately first. The line between fun and impersonation is described in our article on deepfakes.

Frequently Asked Questions

How to make an AI action figure from a photo for free?

Upload a photo to free Gemini or ChatGPT and paste prompt 1 or 3. Gemini has a larger daily image limit, so it’s easier to get several attempts there.

Which prompt is best for a boxed figure?

Prompt 1: 1:6 scale, transparent window, 4 accessories and short text. In our test it produced correct text and faithful face.

Can I 3D print an AI figurine?

Not from this image. It’s a flat graphic, and 3D printing needs a 3D model (e.g. STL file), which is made in other tools or commissioned from a modeler.

Why doesn’t the figurine look like me?

Usually due to a poor photo: shadow, profile, sunglasses. Upload a front-facing photo and add “keep facial features from the original.”

Does AI figurine work in other languages?

Yes. We tested all prompts in Polish and the text came out correctly, but the English prompts work just as well.

The link to Higgsfield is a partner link: if you buy through it, QUANTSLANT earns a commission and you pay the same price. It has no influence on the content, including where the coverage is critical.