
Use a photo where the head is clear
Face lit and unobstructed, hair shape visible. Everything below the neck is going to be simplified anyway, so the head is the only part of the source that really matters.
One photo · Head about a third of the height · Gloss resin on a base
The whole joke of the format is that the head stays recognisable while everything else is thrown away. Get the ratio wrong and it is just a small statue; keep the face too detailed and it stops being funny.
All three are the same object made for different shelves. Look at where the head meets the collar first — that gap and seam is the detail that most conversions forget, and without it the figure reads as a single moulded statue. Every output is AI-generated, so treat these as what the conversion aims at rather than results you will get back identically.



Three things make one of these read as a real object, and two of them are about manufacturing.
One of these is a moulded resin figure in two pieces: a heavy head sitting on a spring inside a hollow neck, and a body on a weighted base. That construction shows. There is a visible gap and seam where the head meets the collar, the base is a simple disc because it has to be stable, and the paint is applied by hand over a moulded surface so edges are soft and small areas get missed.
On top of that, moulding forces simplification — fingers merge, clothing folds become shallow ridges, small props disappear. What survives is the head, which is exactly the point. One honest limit worth saying: this produces a photograph of a figure, not a printable model.
Creative engine
Upload a photo, name the base and setting, then check the neck joint and the paint edges.

A sports base, an office desk, a boxed collectible in packaging. Each one changes the body styling and the base, and it gives the object a reason to exist.

A beard, a cap, glasses, a particular hairline. At this scale one strong feature does more recognition work than a carefully rendered face.
The prompt handles the ratio and the resin. These are about your subject.

Face lit and unobstructed, hair shape visible. Everything below the neck is going to be simplified anyway, so the head is the only part of the source that really matters.

Sports figure on a pitch base, desk figure with a blank nameplate, boxed collectible, or a worn vintage piece. This sets the body styling and the base in one clause.

Say which detail must survive — the beard, the cap, the glasses. At this scale you get one or two, so choose rather than hoping.
Four jobs, all of them affectionate.
The format is inherently fond rather than mocking, which is why it works for colleagues and coaches where a caricature might not.
The native habitat of the format. A player on a pitch base with simplified kit is instantly legible to anyone who follows the sport.
Settling head ratio, base design and paint scheme before commissioning a sample. Far cheaper to iterate here than through a factory.
A figure sitting on a desk with a blank nameplate is a ready-made template for personalised gifts and office humour alike. Desk-scale objects come in softer materials too — a plush toy is the same idea in fabric.
The questions worth answering before the first run.
About a third of the total height is the sweet spot and it is what the default prompt asks for. Larger stops looking like a manufacturable object; smaller and you have a normal figurine, which is a different format entirely.
If the head is clear in the source, usually yes. All the recognition happens above the neck, so name the one feature that identifies the person — beard, glasses, hairline — and it will be preserved.
No. This produces a photograph of a figure, not a mesh. It works well as a brief for someone modelling one, because it settles the proportions and the paint scheme.
Because the paint is too perfect. Hand-painted resin has slightly soft colour boundaries, visible brush texture in flat areas, and uneven gloss. Ask for those explicitly and the object stops reading as a render.
Proportions and intent. A figurine keeps realistic proportions and tries to hold detail; this format deliberately throws the body away to make the head the joke. Different results from the same photo.
It goes over HTTPS, lands in your own history and nowhere else. Nothing is used to advertise. That matters a little more here than elsewhere, because the source is usually a photo of someone who is not you.
Credits are shared across everything Kavel runs. You see the cost before each run, and a failed job is never charged.
What's included
Secure checkout by Stripe. Card details never touch our servers.
What's included
Secure checkout by Stripe. Card details never touch our servers.
What's included
Secure checkout by Stripe. Card details never touch our servers.
What's included
Secure checkout by Stripe. Card details never touch our servers.
Charges appear as “KAVEL AI” on your card statement. You can cancel any time from Settings → Billing; cancellation takes effect at the end of the period you already paid for. Operator details, the full model list, and the refund window are on the about page.
Use a photo with a clear head, pick the format, and name the one feature that has to survive — or keep the proportions and make a figurine.