Style tags this model ignores

Are there tags that simply do nothing when you type them?

There are. And until you know that, you assume you wrote the prompt wrong.

Each of the four below failed to do what its name says. pixel art is the sharp case: it scored high when measured — it changes the image, just not into pixel art. A clear example of a number and an effect diverging.

Every cut below shares one seed, one base prompt and one sampler setting, and differs only by the tag marked in amber.

Example image generated with the pixel art tag — Pixel art

pixel artPixel art

Example image generated with the sketch tag — Sketch

sketchSketch

Example image generated with the lineart tag — Lineart

lineartLineart

Example image generated with the monochrome tag — Monochrome

monochromeMonochrome

What each cut shows

pixel art: The figure is an ordinary anime render; the only pixel feel is the background foliage and shadows breaking into square blocks. Real pixel art needs a dedicated LoRA or post-processing.
sketch: No pencil, no unfinished feel — colour thins slightly and that is all.
lineart: If you expected line only, this misses. The result is fully coloured.
monochrome: It does not turn black and white; the colour only thins a little. greyscale strips the background completely, though in the art-medium comparison some colour stayed on the figure.

How these were made

Shared prompt

1girl, solo, mature female, adult woman, short black hair, brown eyes, white blouse, blue jeans, standing, park path, <tag slot>

Tap a tag to carry it over to the builder

Negative prompt

school uniform, pleated skirt, blazer, serafuku, child, loli, teenage, bad quality, worst quality, worst detail, censor, jpeg artifacts, watermark, signature, text, extra limbs, bad hands, bad anatomy

Model

rinFlanimeIllustrious_v30 · euler_ancestral/normal · steps 28 · CFG 5 · 896×1152 · seed 777003

Takeaway

When a tag seems inert, check whether it works on your model before rewriting the prompt around it. It saves the slot.

How far this result reaches

All four may work elsewhere. sketch and lineart in particular are well understood by models that read natural language.

Measured on a single Stable Diffusion family checkpoint. The same tags can behave differently on NovelAI, Midjourney or another checkpoint. This is a reference point, not a verdict.

Related comparisons

Medium — what it looks painted with

How far do medium tags like watercolour and oil carry on an anime model?

Line and colouring

Can a tag change how the lines and the colouring are handled?

Rim light — which tag actually makes it

Does typing rim light actually put light on the outline?

The guide this comes from →