AI Image to Custom Figurine: Prepare a Clear Design Brief

2026-11-15

An AI picture can make a figurine idea easy to explain: a familiar face, a particular outfit and a scene that captures someone’s interests. Turning that picture into an order takes a different kind of preparation. If you want to turn an AI image into a custom figurine, start by separating three things: photographs of the real person, the generated design concept, and the details a maker has actually agreed to produce.

Keep the AI image as inspiration. Supply real face references separately, identify which invented details you want, and ask about anything the picture leaves unresolved. An attractive image is not confirmation that its pose, accessories, packaging or proportions are available as a physical product. This guide explains how to turn those uncertainties into a clear, answerable enquiry before ordering.

What can an AI concept contribute to your enquiry?

A concept can communicate a direction that is awkward to describe in one sentence. Perhaps you want a relaxed portrait with a jacket, a seated pose and one meaningful object. Showing those elements together can help establish the intended mood. The image becomes useful when you also explain what each element means and which parts should be ignored.

For example, “use the jacket colour and the relaxed stance” is more precise than “make this exact picture.” The second request leaves the maker to interpret whether you also want the generated face, oversized hands, background furniture, printed box and imaginary lighting. Those choices may have very different implications for the scope of a commission.

A documented Adafruit AI figure project separates creating an image from producing a printable model and then making an object. That is one maker’s workflow, not evidence of THJORE’s production method. Its relevance here is limited but useful: having an image and having a specification for a physical object are different stages. A commission needs its own discussion of what can be made.

Start with the real person, then describe the imagined design

If the gift represents a real person, decide which references establish their appearance. A generated portrait might be the style you like while still showing a different jawline, hairstyle, age or expression. Sending only that portrait makes it unclear whether those changes are intentional.

THJORE’s fully customizable single-person figure listing describes a figure based on face, clothing and pose references. It asks for clear, front-facing photographs without filters and for face and outfit references to be labelled separately. The listing does not establish that an AI image can replace those face photographs or that every generated design will be accepted.

Create two clearly named sets: “real person” and “design inspiration.” In your notes, say which image controls the face and which controls the outfit or pose. A helpful instruction is: “Use the real photographs for the person’s appearance; the generated image shows my requested clothing and composition only.” If you deliberately want a stylised change, name that change rather than expecting the maker to infer it.

For detailed photo selection, use our custom bobblehead photo preparation guide. The extra task here is assigning a purpose to each file. More images do not necessarily resolve uncertainty if they disagree and none is identified as the reference to follow.

Annotate what you want to keep, change and disregard

Before submitting a generated image, examine it as a collection of requests. You might like the overall scene but dislike the invented footwear. The AI may have added a badge, a ring, an extra pocket or words that were never part of your idea. Circle or label important areas on a copy, while keeping the original reference available.

  • Keep: elements you intentionally want considered, such as the jacket colour or the position of an arm.
  • Change: elements that should follow another reference, such as real glasses instead of the generated pair.
  • Disregard: background scenery, lighting effects, invented text or objects that are not part of the requested figure.
  • Ask: details whose physical form or inclusion needs the maker’s answer.

Use these labels selectively. A page covered with arrows can be harder to understand than a short numbered list. Match each number to one request, and describe the visible detail rather than an impression such as “make it more premium.” “Plain dark jacket, without the pictured badge” gives the recipient of your enquiry something specific to confirm.

Text needs its own line in the brief. If a generated nameplate shows approximate or distorted lettering, type the intended wording separately. The illustration should not decide spelling, punctuation or capitalisation. Treat a decorative nameplate in the image as a request to discuss, not proof that its particular shape or finish is included.

Resolve conflicting views before adding more references

Different generated views may appear to describe the same figure while showing different details. In one image the jacket closes on the left; in another it closes on the right. A prop switches hands, a strap disappears, or the shoes change. The practical issue is not which image looks better in isolation, but which design you are asking someone to follow.

Choose a main concept and list exceptions beside it. For instance: “Image A controls the pose and clothing. Use the glasses in real photo B. Ignore the bag in image A.” If you genuinely want a combination, describe that combination directly. Do not leave several equally authoritative versions in the upload set.

A generated rear view is also a proposed design, not a photograph of an existing object. Compare it with the front before submitting it. If you have no preference for an unseen area, say so and ask the maker what information is needed. “Back of jacket not specified; please discuss” is more honest and useful than pretending an uncertain view has already resolved the design.

Translate the scene into physical questions

Size and proportions

An image does not establish height, width or depth. A figure can look compact against an invented background while needing more space because of outstretched arms or a large accessory. State the product size you are considering and identify any maximum space available for the finished object. Use a number and unit rather than “small enough for a shelf.”

Ask how the chosen size relates to the requested design, including its base and projecting details. Do not derive dimensions from image pixels or assume every element can keep the same visual proportion at every size. If a particular detail matters more than the overall composition, include that priority in the enquiry so a feasibility discussion has a clear starting point.

Contact points and unsupported details

Look for the parts that appear to hold the composition together. Does each foot meet the base? Is a prop touching a hand, or floating beside it? Is a long object supported anywhere else? An illustration can leave these relationships ambiguous. Mark the ambiguity and ask how the pose could be represented physically.

A good question is: “Could this object be held in this position, or would the pose need to change?” Avoid specifying an unseen support system or assuming a slender connection will be suitable. The maker needs to assess the particular design. Your role in the brief is to explain the visual result you prefer and which changes you would consider.

Hidden sides and movement

Identify meaningful details hidden by the camera angle: the other hand, the back of a chair or the far side of an accessory. Decide whether you have a real reference, a clear preference or an open question for each. This prevents a front-facing concept from silently becoming a complete design.

Also separate a pose from movement. A bent arm in a picture does not establish a movable joint. THJORE’s single-person listing offers stationary-head and bobble-head selections; neither selection establishes articulated limbs. If movement is central to your idea, read our guide to bobbleheads and articulated action figures and confirm the actual product type before planning around it.

Product example from THJORE reunionJoy’s fully customizable single-person figurine listing
A genuine example from THJORE’s single-person product listing, shown as a product reference. This is not an AI concept and finished-object comparison.

Separate the figure from its presentation

A concept may show a portrait inside an illustrated retail box, surrounded by accessories, a display case and a dramatic backdrop. Make an explicit list of the physical items you are requesting. Otherwise, an image intended to communicate the figure’s outfit can appear to request an entire packaged set.

Write “figure only; ignore illustrated packaging” if that is your intention. If the box or display arrangement matters, ask about it separately. A packaging option on a product page does not confirm reproduction of a generated box design. Likewise, an image of several figures does not change a single-person listing into a group commission. Select an appropriate product or discuss a different arrangement.

Keep budget and scope in the same conversation. The current THJORE listing specifically asks buyers to discuss unusual accessories, complex scenes and changes beyond the listed options before ordering. A generated picture cannot confirm that those requests are included at the displayed price. Ask which elements can be considered and whether the agreed design changes the quotation.

Use a brief that separates requests from answers

A short brief should make it possible to answer individual questions without interpreting the whole image again. The template below is an editorial suggestion for organising an enquiry, not a claim about a required upload format. Replace each bracketed section with your own details and leave unanswered items clearly marked.

Subject: Enquiry about a single-person figurine using an AI concept as inspiration.

Real person: [Names of original face photo files.] These establish the person’s appearance. [Any intentional appearance change, or none.]

Main concept: [File name.] Please consider its [outfit, pose or specified details]. It is generated inspiration, not a photograph of an existing figure.

Other references: [File name and the exact element it controls.] If references disagree, follow [specific reference] for [specific detail].

Keep / change / disregard: [A short numbered list linked to the concept.] Exact requested text: [wording, or none].

Product and size: [Product link and option under consideration.] Available display space: [dimensions with units, if relevant]. Preferred head option: [choice or question].

Physical questions: [Unsupported prop, missing rear view, base arrangement or other uncertainty.] My priority is [detail]; I would consider changing [detail].

Included items requested: [Figure and any separately requested accessories or presentation items.] Please confirm what is feasible, what is included and any effect on the quotation before I order.

After receiving a response, add “confirmed,” “needs clarification” or “not included” beside the relevant points. Do not mark the entire concept confirmed because one question was answered. A reply agreeing to a jacket colour does not also establish the feasibility of the pictured prop or the inclusion of packaging.

An illustrative example: a portrait with an imagined hobby scene

Consider a hypothetical buyer whose generated image shows a relative wearing a green jacket, holding a fishing rod, with a fish floating beside the figure and a printed box behind it. This is an illustration of brief preparation, not a THJORE customer case or a confirmed available design.

The buyer wants the relative’s real face, the green jacket and the fishing theme. They do not want the generated beard or the box. Instead of uploading the image alone, they supply an original face photograph, label the jacket as inspiration and explicitly exclude the beard and packaging. They ask whether the rod and fish can be represented in a feasible arrangement.

Their priorities are the recognisable person and the jacket; the exact prop position is negotiable. If the maker proposes an alternative arrangement, the buyer can evaluate it against those priorities. The discussion has moved from “copy everything” to a set of decisions that can be accepted, adjusted or declined individually.

When is the concept ready to become an order?

You are ready for the next step when the real subject is identified, the chosen concept is unambiguous and the essential physical requests have answers. Unresolved details should remain visible in the conversation. Do not treat an uploaded file, a selected product option or a generated preview as agreement to every element in the image.

Later, if a proof is provided for your order, compare it with the agreed brief and actual reference photographs. Our custom bobblehead proof review checklist covers that separate stage. The AI concept remains a reference for the choices you retained; it does not override changes you subsequently agreed with the maker.

Explore the fully custom figure collection for a suitable starting product. If your idea depends on a generated scene, unusual accessory or another detail outside the listed options, send the concept and your questions to THJORE before ordering. A useful brief preserves the creative idea while giving both sides a clear record of what is requested, what is possible and what has actually been agreed.