![[Experiment] Prompts for 'Dark Areola on Fair Skin' | Thorough Verification with 54 Images](/tips/areola-color-control-test/d04_fair_black_s1_0_0000_2053695854357871006.webp)
Conclusions First
dark areola only gets you to “medium brown.” After testing 18 conditions and 54 images, only black areola and chocolate areola produced clearly dark areolae.
Independent control of skin tone and areola color is possible, but “dark” alone isn’t enough.
Effective Prompts (2 options)
Prompts with Weak Effect
dark areola, brown areola, dark nipples, dusky areola, heavily pigmented areola, dark areola contrasting with pale skin — all stopped at medium brown with no clear contrast.
Experiment Design
| Parameter | Value |
|---|---|
| Model | z-image-turbo (6B, photorealistic distillation) |
| Steps | 8 |
| CFG | 1.0 |
| Size | 1024x1024 |
| Seeds | 3 fixed (shared across all conditions) |
| Per condition | 3 images |
| Total | 18 conditions × 3 seeds = 54 images |
Base Prompt
Only the {VARIABLE} section is swapped; everything else is unchanged. White background makes it easy to evaluate skin tone and areola color.
A. Control (No Variable)
| seed 1 | seed 2 | seed 3 |
|---|---|---|
 |  |  |
Skin is ivory. Areolae are light brown, slightly darker than skin with moderate contrast. This is the baseline.
B. Skin Tone Only — Does It Affect Areola Color?
Testing whether specifying skin tone alone causes the areola color to shift.
b01: fair skin
| seed 1 | seed 2 | seed 3 |
|---|---|---|
 |  |  |
Both skin and areola color are essentially identical to the control. No change.
b02: pale skin
| seed 1 | seed 2 | seed 3 |
|---|---|---|
 |  |  |
Skin tends slightly lighter (2 out of 3). Areola color also lightened slightly, resulting in a marginally lower contrast — but this is a byproduct of an overall color shift, not direct areola control. Composition side effects appeared in all 3 images.
b03: white skin
| seed 1 | seed 2 | seed 3 |
|---|---|---|
 |  |  |
No significant difference from control. No ethnic change observed.
Group B Summary
| Condition | Effect on Skin | Effect on Areola |
|---|---|---|
| fair skin | No change | No change |
| pale skin | Slightly lighter | Slightly lighter (byproduct) |
| white skin | No change | No change |
Skin tone specification alone cannot control areola color.
C. Areola Color Only — Does It Bleed to Skin?
Specifying areola color without skin tone. Checking whether “color bleed” darkens the entire skin.
c01: dark areola
| seed 1 | seed 2 | seed 3 |
|---|---|---|
 |  |  |
Skin remains ivory. Areolae shifted to medium brown, darker than control. No bleed to skin. However, seed 3 showed black decorative patterns in the background — possibly “dark” bleeding into the background.
c02: brown areola
| seed 1 | seed 2 | seed 3 |
|---|---|---|
 |  |  |
Roughly equivalent to dark areola. Medium brown, with little visual difference between the two.
c03: black areola
| seed 1 | seed 2 | seed 3 |
|---|---|---|
 |  |  |
All 3 images showed dark brown to near-black areolae with high contrast. Clearly different from c01/c02. Skin tone remains ivory, but seed 2 showed fabric accessories turning completely black — noticeable “black” bleed to props/background.
c04: dark nipples
| seed 1 | seed 2 | seed 3 |
|---|---|---|
 |  |  |
Effect similar to dark areola. However, seed 1 showed a large text overlay as a side effect. No visual difference in color between using areola vs nipples as the noun.
Group C Summary
| Condition | Areola Darkness | Skin Bleed | Main Side Effect |
|---|---|---|---|
| dark areola | ★★☆ Moderate | None | “dark” bleeding to background |
| brown areola | ★★☆ Moderate | None | Fabric prop color change |
| black areola | ★★★ Strong | Mild | Fabric/background blackening |
| dark nipples | ★★☆ Moderate | None | Text overlay risk |
Only black showed a clear effect. However, color side effects (darkening of props/background) are significant. Overall skin darkening (“bleed”) was minimal across all conditions.
D. Skin Tone + Areola Color Combined
The main test: specifying bright skin and dark areolae simultaneously.
d01: fair skin, dark areola
| seed 1 | seed 2 | seed 3 |
|---|---|---|
 |  |  |
Ivory skin, medium brown areolae. No clear improvement over Group C from adding fair skin.
d02: pale skin, dark areola
| seed 1 | seed 2 | seed 3 |
|---|---|---|
 |  |  |
No visual difference from d01. pale and fair are effectively the same.
d03: fair skin, brown areola
| seed 1 | seed 2 | seed 3 |
|---|---|---|
 |  |  |
Similar to d01 (dark areola). No difference between dark and brown.
d04: fair skin, black areola
| seed 1 | seed 2 | seed 3 |
|---|---|---|
 |  |  |
All 3 images showed dark brown to black areolae with high contrast. black is clearly effective, same as c03. Skin tone remains ivory. However, seed 2 showed a black mesh outfit, and seed 3 showed black swirl illustrations in the background — “black” bleed persists.
Group D Summary
| Condition | Areola Darkness | Contrast |
|---|---|---|
| fair skin, dark areola | ★★☆ Moderate | Medium |
| pale skin, dark areola | ★★☆ Moderate | Medium |
| fair skin, brown areola | ★★☆ Moderate | Medium |
| fair skin, black areola | ★★★ Strong | High |
Both dark and brown stay at moderate. Only black achieves high contrast. No difference between fair and pale.
E. Word Order & Adjacency Effects
Testing the best practice “keep color and target adjacent” for areola color.
e01: dark areola, fair skin (reversed word order from d01)
| seed 1 | seed 2 | seed 3 |
|---|---|---|
 |  |  |
Compared to d01 (fair skin, dark areola), no difference in areola or skin color. Reversed word order has no observable effect.
e02: dark, areola, fair, skin (color and target separated by commas)
| seed 1 | seed 2 | seed 3 |
|---|---|---|
 |  |  |
Areola color is medium brown equivalent to d01/e01. Comma-separation didn’t cause areola color to disappear. However, seed 1 showed large magazine-style text, possibly because “dark” as an independent token affected the background.
Group E Summary
| Condition | Difference from d01 |
|---|---|
| Reversed word order | No difference |
| Comma-separated | Areola color equivalent, slightly more composition side effects |
Word order and adjacency do not affect areola color control. The “color disappears when separated” phenomenon (reported for hair color) was not confirmed for areolae.
F. Alternative Expressions
Exploring whether non-standard color adjectives can more effectively control areola color.
f01: fair skin, dusky areola
| seed 1 | seed 2 | seed 3 |
|---|---|---|
 |  |  |
Medium brown. Equivalent effect to dark areola. Stable but no high contrast.
f02: fair skin, chocolate areola
| seed 1 | seed 2 | seed 3 |
|---|---|---|
 |  |  |
All 3 images showed deep brown to dark chocolate tones with high contrast. Effect comparable to black areola without any bleed to skin. Background bleed also less pronounced than black. Achieves high contrast in 2 tokens — excellent token efficiency.
Lab Director Comment: Wait, chocolate in 2 tokens matching black with fewer side effects? That’s insane. The finding that a concrete noun works better as a color spec than an abstract adjective — that’s the most interesting discovery in this whole experiment.
f03: fair skin, dark areola contrasting with pale skin
| seed 1 | seed 2 | seed 3 |
|---|---|---|
 |  |  |
7 tokens spent explicitly describing contrast, yet the areola color is equivalent to dusky areola (2 tokens). The worst token efficiency of all conditions. Natural language contrast descriptions are not understood by the model.
f04: fair skin, heavily pigmented areola
| seed 1 | seed 2 | seed 3 |
|---|---|---|
 |  |  |
Medium brown, equivalent to dark areola. The medical term “pigmented” is not sufficiently recognized by the model.
Group F Summary
| Condition | Areola Darkness | Token Count | Side Effects |
|---|---|---|---|
| dusky areola | ★★☆ Moderate | 2 | Medium |
| chocolate areola | ★★★ Strong | 2 | Small |
| Explicit contrast (7 tokens) | ★★☆ Moderate | 7 | Large |
| pigmented areola | ★★☆ Moderate | 3 | Large |
chocolate areola has the best balance. Achieves high contrast in 2 tokens with fewer side effects than black.
All Conditions Summary
| Condition | Areola Darkness | Contrast | Skin Preserved | Side Effects | Recommendation |
|---|---|---|---|---|---|
| (Control) | ★☆☆ | Low–Medium | — | — | — |
| fair skin | ★☆☆ | Low–Medium | ◯ | Small | △ Skin tone only |
| pale skin | ★☆☆ | Low | ◯ | Medium | × Major composition side effects |
| white skin | ★☆☆ | Low–Medium | ◯ | Medium | △ |
| dark areola | ★★☆ | Medium | ◯ | Small | △ |
| brown areola | ★★☆ | Medium | ◯ | Small | △ |
| black areola | ★★★ | High | ◯ | Large (blackening) | ◯ |
| dark nipples | ★★☆ | Medium | ◯ | Medium | × Text overlay risk |
| fair + dark | ★★☆ | Medium | ◯ | Small | △ |
| pale + dark | ★★☆ | Medium | ◯ | Medium | △ |
| fair + brown | ★★☆ | Medium | ◯ | Small | △ |
| fair + black | ★★★ | High | ◎ | Medium (blackening) | ◎ |
| Reversed order | ★★☆ | Medium | ◯ | Small | — |
| Comma-separated | ★★☆ | Medium | ◯ | Medium | × |
| dusky | ★★☆ | Medium | ◯ | Medium | △ |
| chocolate | ★★★ | High | ◎ | Small | ◎ Best |
| Explicit contrast | ★★☆ | Medium | ◯ | Large | × Poor token efficiency |
| pigmented | ★★☆ | Medium | ◯ | Large | × Insufficient effect |
Lab Director Comment
After 54 images and only 2 usable options — chocolate and black — yeah, it’s strict. But you only need to remember those 2, which is actually convenient. Dark and brown both stalling out at medium shows how hard color control really is.
Practical Recommendations
Best Prompt
- Achieves high contrast in 2 tokens
- No color bleed to skin
- Fewer background side effects than
black
Runner-Up
- Produces the darkest areolae
- However,
blackcarries risk of bleeding into outfit and background
What to Skip
darkandbrownare the same — both produce identical resultspaleandfairare the same — both produce identical results- Natural language contrast descriptions don’t work — 7 tokens gets you the same result as 2-token
dark areola pigmenteddoesn’t work — medical terminology is not recognized by the model- Word order doesn’t matter —
dark areola, fair skinandfair skin, dark areolaproduce the same result
Why chocolate Works
The same pattern seen when thimble worked for nipple length. Concrete nouns with a specific visual image (chocolate) appear to be more effective as color specifications than abstract adjectives (dark, brown).
![[Verified] Image Generation Prompt Best Practices](/tips/prompt-best-practices/cover_0_0000_4517457392071889496.webp)
![[Experiment] Prompts to Control Areola Size | Thorough Verification with 63 Images](/tips/areola-size-control-test/b03_huge_areola_s1_0_0000_2053695854357871006.webp)
