Turn a Photo into a Cartoon

How it works

Your photo is scaled so its total area is around one megapixel, keeping the original proportions. An image-editing model reads that copy alongside an instruction describing a clean illustrated cartoon: simplified shapes, expressive but faithful features, controlled outlines, and a coherent background, with the subject identity and composition preserved.

A second instruction rules out distorted identity, extra limbs, malformed hands, unreadable text, watermarks, and unrelated objects.

The model redraws at high strength. Each run begins from fresh noise, so a second attempt gives a different illustration rather than the same one again.

Simplifying and exaggerating are opposites

The interesting thing about cartooning is that it does two contradictory things at once, and only one of them is asked for here.

Simplification removes. Fewer tones, fewer edges, no skin texture, no small detail. It is what makes a cartoon readable at any size and it has a cost: every face has its unusual measurements averaged down along with everything else, so simplification on its own pushes every subject towards the same generic face.

Exaggeration does the reverse. A caricaturist works out how your face departs from an average one — a longer jaw, wider-set eyes, a stronger brow — and pushes each of those departures further. The result is measurably less accurate than a photograph and frequently faster to recognize, because the identifying information has been amplified rather than merely retained.

Real cartooning does both, and the exaggeration is what pays for the simplification.

This workflow is asked for simplified shapes with faithful features, and the negative instruction explicitly bars distorted identity. Both of those block the amplifying half. What remains is the subtractive half, running alone.

That is a defensible default — nobody wants an automatic caricature of themselves or their family — and it explains the characteristic result. Clean, appealing, competent, and slightly generic.

Cartoon is not one style

The other reason results feel non-specific is that the target is unusually vague.

Anime is a definable tradition with shared conventions. Comic art has a recognizable set of printing-derived signatures. Cartoon covers a century of work that agrees on almost nothing: newspaper strips, mid-century animation, modern television, editorial drawing, children's books, vector mascots. These share a family resemblance and very little else.

A model asked for a cartoon has no narrow target to hit, so it produces something near the center of everything it has seen labeled that way. The output tends to be a clean, contemporary, vector-adjacent illustration, because that is what the middle of the distribution looks like.

If you want a specific tradition, this page will not aim at it, and the reason is the word rather than the model.

Line weight is the missing craft

There is one concrete, visible thing that separates these results from drawn work.

In a drawing, line thickness carries meaning. The outer silhouette gets the heaviest line. Interior forms get lighter ones. A shadowed edge gets weight; an edge in light gets almost none. The eye reads that hierarchy without noticing it, and it is a large part of why a drawing feels authored.

Generated outlines are largely uniform. Every edge arrives at roughly the same weight, so nothing is emphasized, and the picture flattens into an even field of outlined regions.

This is the difference people are noticing when they say a result looks like clip art without being able to say why. It is not the colors or the shapes. It is that nothing in the line is telling you where to look.

Publication gate

This page ships once the workflow has been run against a portrait with a strong expression, a neutral portrait, a group photograph, and a subject with unusual facial proportions, with likeness compared to the source in each case.