Short answer: Attach a clear photo of yourself and ask for "a boxed collectible action figure of the subject from the reference photo", then describe five things in plain sentences: the figure, the packaging, the accessories, what stays recognisable (face, hair, outfit) and the scene. Leave the box blank and add your name afterwards, or give the exact words in quotes.
The action figure picture is everywhere: a person shrunk to a six-inch toy, sealed in a plastic blister with a coffee cup and a laptop beside them. Asked for in one short line, most image models give you a generic toy that looks nothing like you, on a box full of made-up letters. A longer prompt fixes both.
What is the prompt?
This is Prompt Gallery's tested Your Action Figure prompt. Attach your photo, paste it, and change the accessories to your own:
Design a boxed collectible action figure of the subject from the reference photo.
Figure: a detailed six-inch articulated figure with visible joints and fabric costume, standing in the tray in a neutral heroic pose, wearing the same costume as the reference.
Packaging: a clear plastic blister sealed onto a printed backing card. Beside the figure, a moulded accessory tray holding a camera, a guitar and a coffee cup.
Layout: leave a blank nameplate strip across the top of the card and a blank badge in one corner. Add no text, no letters, no numbers, no brand marks and no barcodes anywhere: every printed area is left as flat colour for type to be set later.
Identity: Keep the subject recognisable from the reference photo: the same face shape, skin tone and hair, translated into this style; the sculpt must read as this person: keep the hairstyle, hair colour, wardrobe and any distinctive feature.
Scene: on a collector shelf of other boxed figures, thrown out of focus, clean three-point studio lighting with a soft key, with real plastic sheen on the blister and believable contact shadow.
The card is photographed standing on the shelf, not held in a hand. Output: vertical 4:5 framing, sharp, high detail on the sculpt and the packaging texture.
Keep it original: do not depict any celebrity or real public figure other than the subject of the reference photo, and do not copy any existing painting, artwork, film or game character, mascot, logo or trademarked design.Each paragraph does one job, which is what Google's own guide for its image models asks for: "The more specific you are, the more control you have over the results."
Which photo gives the best likeness?
The model can only keep what it can see. A good reference photo has:
- One person, with the face large in the frame: a head-and-shoulders or waist-up shot, not a group or a distant full-body photo.
- The face toward the camera, straight on or slightly turned, in even daylight. Hard shadows and sunglasses cost likeness.
- The outfit you want on the toy. The prompt says "wearing the same costume as the reference", so the jacket in the photo becomes the figure's jacket.
If the face still drifts, add one line: "Preserve the exact facial identity, bone structure, skin tone and hair of the reference." Then ask again in the same chat; a second try with the same photo is often closer than the first.
Should the name be printed on the box?
You have two ways, and they suit different models.
Leave it blank (the prompt above). Every model can do this well, and you set your name afterwards in any photo editor, in a font you choose.
Write it in the prompt, in capitals and quotes, with where it goes. The gallery's Action Figure In Blister Pack asks for the product name "THE NIGHT ARCHITECT" in a large display typeface across the top and the line "URBAN LEGENDS · SERIES 1" as a smaller banner:
Large words given in quotes come out right; small print you did not spell out (a warning label, a barcode line) is invented and often garbled. So either give every word you want, or ask for "no other text".
How do you change the accessories and the pose?
Change only the line that holds them, and keep the rest:
| You want | Change this line to |
|---|---|
| Your job on the box | "a moulded accessory tray holding a laptop, a headset and a tiny whiteboard" |
| A hobby | "...holding a pair of running shoes, a medal and a water bottle" |
| A pet | "...holding a small articulated figure of the dog from the second reference photo" |
| A different pose | "standing in a confident pose, arms crossed" in place of "neutral heroic pose" |
| A square post | "Output: square 1:1 framing" |
Three accessories read best; with six or more the tray gets crowded and the objects shrink. For a pet, attach its photo as a second image and say which photo is which.
Does it work in ChatGPT and Gemini?
Yes: both take a photo with the prompt. OpenAI's image guide says its GPT Image models can "generate new images using other images as a reference", and ChatGPT takes an attached photo the same way; Google's guide for Gemini says: "Provide an image and use text prompts to add, remove, or modify elements, change the style, or adjust the color grading." In Prompt Gallery this prompt was tested in Nano Banana, Flux, Midjourney and Seedream.
Two things differ. Gemini adds an invisible SynthID watermark to every image it makes ("All generated images include a SynthID watermark"). And a model that does not accept a photo (a text-only setting) cannot keep your face: it invents a person from the words.
What else works the same way?
Swap the packaging paragraph and keep the identity one. The gallery's Collectibles group has 13 of these, each tested with a result next to it:
- Boxed Vinyl Figure: the big-headed collectible in a window box.
- Capsule Mini Me: a tiny chibi figure inside a capsule toy, held between two fingers.
- Snow Globe: you as a resin figurine in a glass globe, mid-shake.
- Holographic Trading Card: a foil card with your name, a rarity stamp and stats.
Keep the "Keep it original" paragraph in all of them. It stops the model from copying a real toy brand's logo or box design, and from putting a celebrity's face on the figure instead of yours. Use your own photo, or one of someone who said yes.
Where can you check this?
- Google AI for Developers: image generation with Gemini
- OpenAI: image generation guide (editing with reference images)
Where do you find more prompts like this?
Prompt Gallery has 367 tested image prompts, each with the real result next to it, ready to copy. Open People & Characters, then Collectibles, for the action figures, vinyl toys, trading cards and sticker sheets; copy one, attach your photo and change the details in plain words.
