# Why "a wolf" gets you the average wolf

> Generic output isn't a settings problem or a talent problem. A model returns the average of what it was named, and most briefs name a category — here's what to name instead.

August 16, 2026 · https://justog.club/stories/why-a-wolf-gets-the-average-wolf

"Slop" became Merriam-Webster's word of the year in late 2025, defined as low-quality digital
content produced in quantity by machine. The complaint is real, the images being complained about
do exist, and pretending otherwise is not worth anyone's time.

But the usual diagnosis is wrong twice over — and both wrong answers keep people stuck.

## Why do AI images look generic?

Because a model returns the average of what it was given, and most briefs name a category.

Ask for "a wolf" and the honest response *is* the average wolf: the one that sits at the centre of
every wolf that has ever been drawn. It isn't a failure of the tool. It's an accurate answer to a
question with no subject in it.

Genericness, in other words, is almost never a rendering problem. It's a naming problem — and it
survives every technical fix because the technical fixes never touch it.

## Isn't it a settings problem?

That's the first wrong answer, and it's most of what's published on this.

The standard advice is to lower the guidance scale for variation, add negative prompts against
*smooth, plastic, airbrushed*, ask for film grain and natural skin texture, or switch to a
differently fine-tuned model. All of that works — on **surface**. You get a less plastic average
wolf.

There's one piece of that advice worth keeping, because it points the right way: piling forty
adjectives into a line produces generic results, and two or three strong constraints beat a
paragraph of vibes. The reason is the same as everything else here — adjectives describe the
manner of a thing, and the thing was never specified.

## Isn't it just that most people aren't creative?

That's the second wrong answer, and it's worse, because it's unfalsifiable and it stops the
conversation.

The observation underneath it is fair enough: given the ability to make anything, a lot of people
ask for the most familiar subject in the most familiar style. But "be more creative" is not an
instruction anybody can act on, and it misdescribes what's actually missing — which isn't
imagination, it's a **subject**.

That's a solvable, teachable thing, and it takes one sentence rather than a personality change.

## So what do you name instead?

The metaphor where the subject meets the people who'd wear it.

Not the animal — what the animal *means* here. A wolf for a gym is pack discipline, or the lone
hunter's persistence. That's a different image from a wolf for a hiking club, which is a different
image again from a wolf on a black metal sleeve. Same animal, three unrelated designs, and the
difference was decided before any style was chosen.

Some worked examples of the same move: a leaping koi carries going-against-the-current. A skull
carries memento mori — not "edgy". The point of naming it is that everything downstream then has
something to be faithful to, and the additions stop being decoration.

## Is there a test for whether it worked?

Yes, and it's blunt: **a line that could sit on any other subject is a failure.**

If the slogan on the wolf shirt would work equally well over an eagle, a bear, or a barbell, it
isn't voicing anything — it's filler that happens to rhyme with the picture. Same for the props,
the palette, the texture: each should be traceable to the specific idea rather than to the general
category.

This is the rule the design layer here runs on, and it's deliberately harsh, because filler is
what genericness is made of. It's also the honest answer to the slop charge — not a defence of the
tool, and not a claim that the critics are wrong about what they've seen.

## Do adjectives help at all?

Barely, and this is measurable rather than a matter of taste.

In our own production testing, concrete noun props were delivered close to one-for-one while
adjectives like *hellish* or *aggressive* did essentially nothing. Asking for a fiercer wolf
returns the same wolf. Asking for a snapped chain on a spiked collar, smoke coming off the fur,
no pupils — that returns a different animal.

So intensity is a list of things, not a stack of modifiers.
[What to change when a result is wrong](/stories/when-the-image-comes-out-wrong) goes
through the rest of that.

## Where does style come in?

After the core, as a register — not as a substitute for it.

Once the idea has a subject, the same subject can be rendered as vintage americana engraving, as
sumi-e ink wash, as 90s bootleg airbrush, as Y2K chrome, as a varsity mark, as an anatomical
plate. Those are genuinely different worlds and choosing between them is real design work.

What style can't do is supply the missing subject. A generic idea in an interesting register is a
generic idea with a filter on it, which is most of what gets posted and most of what people mean
when they say something looks machine-made.

## What if it's a photo of my own dog?

Then the specificity is already there, and the failure mode inverts.

A photograph of your dog is the least generic input possible — it's one animal, with a broken
stripe over one eye and a folded ear tip. What goes wrong is a rendering that returns a handsome
dog *of that general description* instead of him. The fix is to name what makes him identifiable
so those features are held while everything else is free to change.

The dog stays your dog. The rendering is what varies.

## How do I get one made?

Describe the idea at [JustOG](/) — including who it's for, which is the part that does the work —
and the directions come back as a set. Pick one, drag the crop frame, see it composited on the
real garment, and it's made to order and shipped.

Designs other people have published are in the [shop](/shop).

The model isn't averaging your idea. It's averaging the category you handed it instead of one.