Key Findings
- The leopard recipe
gyaru fashion, leopard print top, denim mini skirtwas reproduced in 3/3 images. A fitted leopard-print long-sleeve top paired with a denim mini skirt appeared consistently - The Y2K recipe
y2k fashion, platform boots, mini skirtwas also reproduced in 3/3 images. A denim mini skirt, platform lace-up knee-high boots, a graphic-print short-sleeve/crop top, and a chain necklace appeared consistently - The hime gyaru recipe
hime gyaru fashion, frilly pink dresswas reproduced in 3/3 images. A pink multi-layer ruffle dress, bow-trimmed platform pumps, and hair accessories appeared consistently - Bare
gyaru fashiondid not produce the expected gyaru look, leaning instead toward a school-uniform/kogal style. All three images showed a patterned pleated mini skirt, and a necktie appeared in two of three, but no leopard print or skin-revealing outfits appeared - No hair color change was observed in any of the 15 images across all 5 conditions. Clothing-only keywords alone did not shift hair color
Experiment Design
| Item | Value |
|---|---|
| Model | z-image-turbo (6B, photorealistic distilled) |
| Steps | 8 |
| Sampler | euler |
| Scheduler | ddim_uniform |
| CFG | 1.0 |
| Size | 1024x1024 |
| Seeds | 42 / 43 / 44 fixed (shared across conditions) |
| Per condition | 3 images (3 seeds) |
| Total | 5 conditions = 15 images |
Base Prompt
We used a full-body standing pose rather than an upper-body shot so the entire outfit (top, bottom, shoes, accessories) could be observed at once.
Caveat: The Y2K, leopard, and hime gyaru conditions are multi-keyword “recipes,” not single-variable changes. Conclusions therefore apply to each recipe as a whole; individual keyword contributions cannot be isolated.
A00: control (no style keywords)
| seed 42 | seed 43 | seed 44 |
|---|---|---|
![]() | ![]() | ![]() |
Observations: All three images show dark-to-dark-brown hair with natural makeup. Outfits — a striped shirt dress, a black top with grey striped shorts, and a white t-shirt with beige shorts — are all office-casual to smart-casual. No gyaru elements appeared in any of the three.
A01: gyaru fashion (single keyword)
| seed 42 | seed 43 | seed 44 |
|---|---|---|
![]() | ![]() | ![]() |
Observations: All three images show a patterned pleated mini skirt, but hair stayed dark-to-dark-brown with no color change. A necktie appeared in 2/3 images (seeds 43, 44), and black socks with platform loafers also appeared in 2/3 (seeds 43, 44). Overall, rather than typical gyaru markers like leopard print or tanned skin, the outfits leaned toward a school-uniform/kogal silhouette.
A02: Y2K recipe
| seed 42 | seed 43 | seed 44 |
|---|---|---|
![]() | ![]() | ![]() |
Observations: 3/3 images show a denim mini skirt and platform lace-up knee-high boots together. A graphic-print short-sleeve top (cropped, with midriff exposed, in 2/3 images) and a chain or layered necklace also appeared in all three. Boot color varied — white in seed 42, black in seeds 43 and 44 — but the silhouette matched across all three. Logo text on the tops rendered as unreadable artifacts in some images.
A03: Leopard recipe
| seed 42 | seed 43 | seed 44 |
|---|---|---|
![]() | ![]() | ![]() |
Observations: 3/3 images show a fitted, long-sleeve leopard-print top together with a denim mini skirt. Wavy long hair and reddish lipstick, plus a necklace, also appeared in all three. Black lace-up platform shoes were visible in 2/3 images (seeds 43, 44); seed 42’s shoes are out of frame. This was the most consistently reproduced recipe of the five conditions.
A04: Hime gyaru recipe
| seed 42 | seed 43 | seed 44 |
|---|---|---|
![]() | ![]() | ![]() |
Observations: 3/3 images show a pink multi-layer ruffle dress (with an asymmetric hi-low hem) together with bow-trimmed platform pumps. A choker or necklace also appeared in all three, as did a hair accessory (a flower hairpin in seeds 42/43, a tiara in seed 44). Unlike the other four conditions, which showed neutral-to-serious expressions, all three images in this condition show a soft smile.
Note: The expression change was not specified in the prompt, but all three images shifted in the same direction. It’s possible the decorative dress keywords influenced overall pose and mood, but the cause cannot be confirmed.
Summary
| Condition | Hair change | Top | Bottom | Shoes | Success |
|---|---|---|---|---|---|
| control | - | - | - | - | - |
| gyaru fashion only | none | uniform-like (necktie in some) | patterned pleated mini skirt | platform loafers (2/3) | 3/3 (but not a typical gyaru look) |
| Y2K recipe | none | graphic-print top | denim mini skirt | platform lace-up boots | 3/3 |
| Leopard recipe | none | leopard-print top | denim mini skirt | black platform shoes (2/3) | 3/3 |
| Hime gyaru recipe | none | pink frilly dress (one piece) | pink frilly dress (one piece) | bow-trimmed platform pumps | 3/3 |
An abstract style name like gyaru fashion leaned toward a school-uniform/kogal look rather than the expected gyaru markers (leopard print, tanned skin, platform boots). By contrast, recipes that specified concrete clothing parts — pattern, material, item name, as in the leopard, Y2K, and hime gyaru conditions — all reproduced the intended outfit consistently in 3/3 images. This matches the “concrete part keywords beat abstract style names” conclusion from the gyaru makeup test and the gyaru hair/skin test.
Note: Clothing recipes alone did not shift hair color or skin tone. To target gyaru-style hair and skin, combine this article’s outfit recipes with the hair/skin keywords verified in the gyaru hair/skin test.
















![[Experiment] Prompts for Gyaru Makeup | 8 Conditions, 24 Images](/tips/gyaru-makeup-test/a05_false_lashes_s2_0_0000_7295072554507705269.webp)
![[Verified] Image Generation Prompt Best Practices](/tips/prompt-best-practices/cover_0_0000_4517457392071889496.webp)