![[Experiment] Bunny Girl Outfit Variations — How Much Does Impression Change with Color, Material, and Design? 48-Image Comparison](/tips/bunny-girl-variation-test/cover.webp)
Conclusion First
- Color specification (red / pink / gold / white) spread to the leotard, bunny ears, and bow tie in all conditions, with high reproduction accuracy
- Material specification (metallic silver / lace / velvet / latex) showed clear texture changes
- Leather specification showed some texture change, but color stayed black — relatively small difference
- Strapless specification caused the straps to disappear, confirming that design changes also work
- All 4 basic parts (bunny ears, bow tie, cuffs, leotard) showed high retention across all 12 conditions
- By prepending a modifier to
bunny girl outfit, outfit variation control is possible — that’s the confirmed finding - [Deep Dive] When two modifiers are combined (color × material, color × accessory, material × accessory), all 4 conditions reflected both specifications simultaneously
- [Deep Dive] The leather specification that showed little difference alone showed dramatically increased visual difference when combined with metallic silver
What This Article Covers
- Accuracy when adding color specification (red, pink, gold, white) to bunny girl outfits
- Texture changes from material specification (leather, metallic, lace, velvet, latex)
- Reflection rate of design changes (strapless) and accessory additions (fishnet stockings)
- Simultaneous reflection rate when two modifiers are combined (color × material, color × accessory, material × accessory)
- Differences from the base test (see related article “Classic Bunny Girl Outfit Test” below)
Experiment Design
| Item | Value |
|---|---|
| Model | z-image-turbo (6B, photorealism-focused distilled) |
| Steps | 8 |
| Sampler | euler |
| Scheduler | ddim_uniform |
| CFG | 1.0 |
| Size | 1024x1024 |
| Seed | 3 fixed (all conditions common) |
| Per condition | 3 images (3 seeds) |
| Total | 12 conditions = 36 images (+ 4 deep-dive conditions = 12 images, 48 total) |
| Changed variable | Outfit modifier only (1 variable) |
Base Prompt
Stage lighting in a bar interior is fixed; only the bunny girl outfit modifier is swapped.
Evaluation Criteria
- Color/material reproduction accuracy (does it look as specified?)
- Retention rate of basic bunny girl parts (bunny ears, bow tie, cuffs, leotard)
- Visual difference from control (no modifier)
Results by Condition
B00: Control (bunny girl outfit)
| seed 1 | seed 2 | seed 3 |
|---|---|---|
 |  |  |
Observation: 3 out of 3 produced a typical bunny girl outfit: black leotard, bunny ears, bow tie, white cuffs. Ear color: seeds 1 and 2 are black, seed 3 is white. Leg wear varies (seed 1: bare legs, seed 2: black stockings, seed 3: skin-tone fishnet). Basic 4-part retention: 3/3.
B01: Red Color (red bunny girl outfit)
| seed 1 | seed 2 | seed 3 |
|---|---|---|
 |  |  |
Observation: 3 out of 3 changed the leotard to red. Bunny ears (3/3 red), bow tie (red), and shoes (red) all spread the red color throughout. White cuffs maintained 3/3. Clear difference from control (B00) — high color specification accuracy. Basic 4-part retention: 3/3.
B02: Leather Material (leather bunny girl outfit)
| seed 1 | seed 2 | seed 3 |
|---|---|---|
 |  |  |
Observation: 3 out of 3 have a black leotard, but compared to the control, leather-characteristic gloss and shine increased. Especially seed 2 — leather knee-high boots appeared, showing leather specification spreading to the legs. seed 3 shows little difference from control. Color stays black. Basic 4-part retention: 3/3.
B03: Metallic Material (metallic silver bunny girl outfit)
| seed 1 | seed 2 | seed 3 |
|---|---|---|
 |  |  |
Observation: 3 out of 3 changed the leotard to silver. Bunny ears also turned silver, with clear difference from control. seed 1: white stockings, seed 2: no stockings, seed 3: white stockings with bow tie changing to blue. Color and material both show high metallic silver specification accuracy. Basic 4-part retention: 3/3.
B04: Lace Material (lace bunny girl outfit)
| seed 1 | seed 2 | seed 3 |
|---|---|---|
 |  |  |
Observation: 3 out of 3 changed the leotard to lace material with floral lace pattern and transparency clearly confirmed. Clearly different from the control’s opaque black leotard. Shape also shifts toward more lingerie-style design; seed 2 shows a garter belt and stockings. Bunny ears stayed black fabric/satin (not lace). Basic 4-part retention: 3/3.
B05: Fishnet Stockings Addition (bunny girl outfit, fishnet stockings)
| seed 1 | seed 2 | seed 3 |
|---|---|---|
 |  |  |
Observation: 3 out of 3 showed fishnet stockings, with larger mesh clearly visible compared to the control. While the control had bare legs (seed 1), black stockings with no mesh (seed 2), and fine fishnet (seed 3), the fishnet specification produced consistent large-mesh fishnet across all 3. Basic 4-part retention: 3/3.
B06: Pink (pink bunny girl outfit)
| seed 1 | seed 2 | seed 3 |
|---|---|---|
 |  |  |
Observation: 3 out of 3 changed the leotard to pink satin. Bunny ears (3/3 pink), bow tie (pink), and shoes (pink) all spread pink throughout. White cuffs maintained 3/3. Leg wear: seed 1 white stockings, seed 2 black stockings, seed 3 white knee-high stockings. Color specification accuracy is as high as red. Basic 4-part retention: 3/3.
B07: Gold (gold bunny girl outfit)
| seed 1 | seed 2 | seed 3 |
|---|---|---|
 |  |  |
Observation: 3 out of 3 changed the leotard to gold metallic. Bunny ears also gold, bow tie also gold. seeds 2 and 3 show gold curtains and mirror balls appearing in the background — gold specification spread to the environment. Shoes are also gold heels. All 3 are bare-legged. White cuffs maintained in seed 1; seeds 2 and 3 saw the bow tie change to a choker-like form. Basic 4-part retention 2–3/3.
B08: Velvet (velvet bunny girl outfit)
| seed 1 | seed 2 | seed 3 |
|---|---|---|
 |  |  |
Observation: 3 out of 3 confirmed velvet’s characteristic nap texture and depth of color on the leotard. Color: seeds 1 and 2 are navy (dark blue), seed 3 is dark purple — different color tones from the control black. Bunny ears also same-colored velvet. Bow tie maintained 3/3. Leg wear: seed 1 bare, seed 2 black stockings, seed 3 bare. White cuffs not confirmed in seed 1, maintained in seeds 2 and 3. Basic 4-part retention 2–3/3.
B09: Latex (latex bunny girl outfit)
| seed 1 | seed 2 | seed 3 |
|---|---|---|
 |  |  |
Observation: 3 out of 3 showed strong gloss on the leotard, clearly reproducing the latex material texture. Color stays black. Bunny ears also black with latex-like sheen. seed 1: garter belt + black stockings, seed 2: knee-high latex boots appeared — latex specification spread to the legs. seed 3: stockings + garter belt. Compared to B02 (leather), the gloss is stronger and the difference is more apparent. Basic 4-part retention: 3/3.
B10: White (white bunny girl outfit)
| seed 1 | seed 2 | seed 3 |
|---|---|---|
 |  |  |
Observation: 3 out of 3 changed the leotard to white satin. Bunny ears also white, stockings also white — the entire outfit unified in white. Bow tie: seed 1 pink, seed 2 pink, seed 3 white — only the bow tie tends to pick up an accent color. White cuffs maintained 3/3 (blending into the outfit). All 3 wear white stockings. High color accuracy. Basic 4-part retention: 3/3.
B11: Strapless (strapless bunny girl outfit)
| seed 1 | seed 2 | seed 3 |
|---|---|---|
 |  |  |
Observation: 3 out of 3 have black leotards, but the strap area changed compared to the control. seeds 1 and 3 became tube-top style leotards without straps, reflecting the strapless specification. seed 2: black stockings with a glossy material feel. Leg wear: seed 1 bare, seed 2 black stockings, seed 3 white stockings. Overall, there’s a leather/latex-like sheen. Basic 4-part retention: 3/3.
Full Condition Comparison Table
| Condition | seed 1 | seed 2 | seed 3 |
|---|---|---|---|
| B00: control | | | |
| B01: red | | | |
| B02: leather | | | |
| B03: metallic | | | |
| B04: lace | | | |
| B05: fishnet | | | |
| Condition | seed 1 | seed 2 | seed 3 |
|---|---|---|---|
| B06: pink | | | |
| B07: gold | | | |
| B08: velvet | | | |
| B09: latex | | | |
| B10: white | | | |
| B11: strapless | | | |
Summary
| Condition | Color change | Material change | Basic 4-part retention | Diff from control |
|---|---|---|---|---|
| B00: control | – | – | 3/3 | – |
| B01: red | 3/3 | – | 3/3 | Large (entire outfit turns red) |
| B02: leather | – | 1–2/3 | 3/3 | Small to medium (increased gloss, boots may appear) |
| B03: metallic silver | 3/3 | 3/3 | 3/3 | Large (changes to silver) |
| B04: lace | – | 3/3 | 3/3 | Large (lace transparency, shape changes) |
| B05: fishnet stockings | – | – | 3/3 | Medium (mesh consistently large) |
| B06: pink | 3/3 | – | 3/3 | Large (entire outfit turns pink) |
| B07: gold | 3/3 | 3/3 | 2–3/3 | Large (gold metallic, background also affected) |
| B08: velvet | 3/3 | 3/3 | 2–3/3 | Large (navy to purple, nap texture) |
| B09: latex | – | 3/3 | 3/3 | Large (strong gloss, boots and garter appear) |
| B10: white | 3/3 | – | 3/3 | Large (entire outfit unified in white) |
| B11: strapless | – | – | 3/3 | Medium (straps disappear, increased sheen) |
- Color specification (red / pink / gold / white) achieved 3-out-of-3 reproduction in all conditions, spreading color throughout the outfit
- Material specification (metallic silver / lace / velvet / latex) brought clear texture and shape changes. Latex in particular has stronger gloss than leather — more apparent difference
- Leather specification only increased gloss and had boot appearance (1/3) — relatively small difference from control
- Velvet specification brought color changes to navy and dark purple — not just material, but color shifts too
- Gold specification spread gold decorations to the background — the largest non-outfit effect
- Strapless specification confirmed strap disappearance — also effective for design change control
- Fishnet stockings addition stabilized the leg output that showed variance even in the control
- Basic 4 parts (bunny ears, bow tie, cuffs, leotard) were largely maintained across all 12 conditions and 36 images. Cuffs dropped in some gold and velvet conditions, but otherwise retention was 100%. The
bunny girl outfitconcept doesn’t collapse easily when modifiers are added
Deep Dive: Modifier Combinations
In the first phase, single color, material, and design specifications all worked effectively. Here we test whether combining two modifiers — “color × material”, “color × accessory”, “material × accessory” — produces simultaneous reflection of both specifications.
Experiment Design (Deep Dive)
| Item | Value |
|---|---|
| Added conditions | 4 |
| Seed | 42, 123, 789 (3 fixed) |
| Per condition | 3 images (3 seeds) |
| Total | 12 images |
| Changed variable | Outfit modifier combination (1 variable) |
Base prompt is identical to Phase 1.
D00: Color × Material (red lace bunny girl outfit)
| seed 1 | seed 2 | seed 3 |
|---|---|---|
 |  |  |
Observation: 3 out of 3 changed the leotard to red lace material. Floral lace pattern and transparency confirmed — different shape and texture from B01 (red alone)’s opaque red leotard. Compared to B04 (lace alone)’s black lace, the color has been replaced with red. Bunny ears are red in 3/3. Bow tie red in 3/3. White cuffs maintained in seed 3; seeds 1 and 2 changed to lace wristbands. Leotard shape tends toward lingerie-style, similar to B04 (lace alone). Color and material specifications are simultaneously reflected.
Comparison with Phase 1:
- B01 (red): opaque red leotard → D00: red lace leotard (material changed to lace)
- B04 (lace): black lace leotard → D00: red lace leotard (color changed to red)
- Both features reflected simultaneously
D01: Color × Material (metallic silver leather bunny girl outfit)
| seed 1 | seed 2 | seed 3 |
|---|---|---|
 |  |  |
Observation: 3 out of 3 changed the leotard to a silver metallic gloss material. Compared to B03 (metallic silver alone), the surface gloss is stronger with added leather-characteristic rigid texture. While B02 (leather alone) stayed black with only increased gloss, D01 shows both silver color change and leather texture confirmed. Bunny ears also silver in 3/3. Bow tie black (3/3). White cuffs maintained in 3/3. seed 2 shows silver knee-high boots — same leather-to-legs spread as B02. Basic 4-part retention: 3/3.
Comparison with Phase 1:
- B02 (leather): black leather → D01: silver leather (metallic silver color specification reflected)
- B03 (metallic silver): silver metallic → D01: silver leather (leather texture added)
- Leather alone showed small difference, but combining with metallic silver dramatically increased visual distinction
D02: Color × Accessory (red bunny girl outfit, fishnet stockings)
| seed 1 | seed 2 | seed 3 |
|---|---|---|
 |  |  |
Observation: 3 out of 3 produced red leotard + fishnet stockings. The leotard is glossy red (enamel/latex-like texture), slightly different material quality from B01 (red alone). Fishnet stockings appear in 3/3 with large mesh — same stability as B05 (fishnet alone). Bunny ears red in 3/3. Bow tie red in 3/3. White cuffs maintained in 3/3. Shoes are also red heels. Both color and accessory specifications reflected stably.
Comparison with Phase 1:
- B01 (red): red leotard + leg wear varies → D02: red leotard + fishnet stable
- B05 (fishnet): black leotard + fishnet → D02: red leotard + fishnet
- Color and accessory don’t interfere — both reflected 3/3
D03: Material × Accessory (lace bunny girl outfit, fishnet stockings)
| seed 1 | seed 2 | seed 3 |
|---|---|---|
 |  |  |
Observation: 3 out of 3 produced black lace leotard + fishnet stockings. The leotard has the same lace pattern and transparency as B04 (lace alone), with fishnet stockings stably added. Garter belt appeared in 3/3, higher than the 1/3 rate in B04 (lace alone). The lace and fishnet combination may have promoted garter belt appearance. Bunny ears black (3/3). Bow tie maintained 3/3. White cuffs not confirmed in seed 1, maintained in seeds 2 and 3.
Comparison with Phase 1:
- B04 (lace): black lace leotard + garter belt 1/3 → D03: black lace leotard + fishnet 3/3 + garter belt 3/3
- B05 (fishnet): black leotard + fishnet → D03: black lace leotard + fishnet
- Material and accessory coexist, with garter belt appearance rate rising
Deep Dive Summary
| Condition | Combination | Color reflected | Material reflected | Accessory reflected | Basic 4-part retention |
|---|---|---|---|---|---|
| D00: red lace | color × material | 3/3 | 3/3 | – | 2–3/3 |
| D01: metallic silver leather | color × material | 3/3 | 3/3 | – | 3/3 |
| D02: red + fishnet | color × accessory | 3/3 | – | 3/3 | 3/3 |
| D03: lace + fishnet | material × accessory | – | 3/3 | 3/3 | 2–3/3 |
- In all 4 conditions, both combined modifiers were reflected. No interference between modifiers was observed
- D01 (metallic silver leather): leather specification that showed little difference alone was dramatically improved when combined with metallic silver
- D03 (lace + fishnet): garter belt appearance rate rose from 1/3 in B04 (lace alone) to 3/3. The combination of material and accessory may promote appearance of related parts
- In D00 (red lace), cuffs sometimes changed to lace wristbands — lace specification also affected cuff shape
- The basic
bunny girl outfitconcept was maintained even with two modifiers combined
Applied: Changing the Mood with Lighting × Pose
Now that modifier control for bunny girl outfit is confirmed stable, here’s an applied example with different lighting and posing.
Key differences from the test prompt:
standing→kneeling on velvet sofa, hands on thighsfor specific pose with propsstage lighting, bar interior→dramatic side lighting, dark studio backgroundfor moody lightinglooking at viewer→looking up at viewerfor upward gaze
| seed=7710 | seed=7711 | seed=7712 |
|---|---|---|
 |  |  |
Observations: All 3 images maintained the core 4 parts (bunny ears, bow tie, cuffs, leotard) with black satin and fishnet stockings. dramatic side lighting produced strong directional shadows in all 3, creating a dramatically different atmosphere from the flat stage lighting of the test. Velvet sofa texture was stable in 3/3. kneeling reflected in 2/3, with 1 defaulting to sitting.
For more elaborate lighting and composition bunny girl examples, see the dedicated showcase article.
Lab Director Comment
Oh wow, this is way too much fun. The fact that all the color specs land every time is great, but the material specs are on another level. Latex’s intense wet gloss bunny, velvet’s navy-to-purple refined luxury feel — same bunny girl, completely different impression. Leather alone was kind of muted, but metallic silver leather turned into this silver chrome glare, and I’m obsessed. Lace bunny with that transparency, and then lace + fishnet making the garter belt appear too — I love that combining specs adds associated parts.
And on the technical side, bunny girl outfit as a concept is really strong — even with two modifiers stacked, the basic 4 parts (bunny ears, bow tie, cuffs, leotard) didn’t collapse. Knowing two modifiers is safe is a big takeaway.
![[Experiment] Classic Bunny Girl Outfit Testing — Can Materials, Colors, and Parts Be Differentiated?](/tips/bunny-girl-outfit-test/A_seed123_0_0000_123.webp)
![[Experiment] Fashion-Style Bunny Girl Variations — How Far Can Goth, Knit, Cyber, and Shrine Maiden Share Space with Bunny Ears?](/tips/bunny-girl-casual-test/X_seed42_0_0000_42.webp)
![[Test] Comparing 3 Cosplay Outfits | Can Maid and Sailor Uniform Appear with Just One Keyword?](/tips/cosplay-outfit-comparison/cover.webp)
