Key Findings
- Hand position has minimal impact in rear compositions — hands are usually out of frame
- Back arch (arched back / lordosis) is the most important variable for butt emphasis
- Face direction:
looking back over shoulderis the most natural and stable - Leg spread can be controlled with
spread legsbut affects overall composition balance - Outfit choice has more impact on output than any single pose modifier
What You’ll Learn
- How pose micro-adjustments affect lingerie visibility from behind
- The most effective pose modifiers for rear compositions
- Optimal pose + lingerie combinations by outfit type
Test Conditions
| Parameter | Value |
|---|---|
| Model | z-image-turbo (6B, photorealistic distilled model) |
| Steps | 8 |
| Sampler | euler |
| Scheduler | ddim_uniform |
| CFG | 1.0 |
| Image Size | 1024×1024 |
| Seeds | 42, 123, 789 (fixed) |
Experiment 1: Hand Position Variations
Outfit: lace lingerie set (different outfit per experiment to observe cross-outfit interactions)
Condition A: No specification (default)

Condition B: arms stretched forward (+3 tokens)

Condition C: elbows on bed, face down (+5 tokens)

Condition D: one hand reaching back (+4 tokens)

Observations
In rear compositions, hand position is mostly out of frame, limiting its output impact. elbows on bed, face down could raise the butt higher but risks inducing a prone position. one hand reaching back can target a hand-on-butt composition but stability is low.
Hand Position Comparison Table
| Seed | Default (A) | Arms forward (B) | Elbows down (C) | Hand back (D) |
|---|---|---|---|---|
| 42 | ![]() | ![]() | ![]() | ![]() |
| 123 | ![]() | ![]() | ![]() | ![]() |
| 789 | ![]() | ![]() | ![]() | ![]() |
Experiment 2: Back Arch Variations
Outfit: sheer babydoll nightgown
Condition A: No specification (default)

Condition B: arched back (+2 tokens)

Condition C: deeply arched back, lordosis (+4 tokens)

Observations
arched back presence directly affects the butt-out feel. Adding lordosis (lumbar lordosis) further emphasizes the arch but trades off natural pose appearance. The babydoll hem was observed to ride up with the arch — a notable interaction between pose and outfit.
Lab Director: “Lordosis” is medical terminology, but drop it in a prompt and the back arch goes dramatic. The babydoll riding up at the waist? Chef’s kiss. Though
arched backalone hits the sweet spot between impact and natural look.
Back Arch Comparison Table
| Seed | Default (A) | arched back (B) | lordosis (C) |
|---|---|---|---|
| 42 | ![]() | ![]() | ![]() |
| 123 | ![]() | ![]() | ![]() |
| 789 | ![]() | ![]() | ![]() |
Experiment 3: Face Direction Variations
Outfit: black thong and matching bralette
Condition A: No specification (default)

Condition B: looking back over shoulder (+4 tokens)

Condition C: face buried in pillow (+4 tokens)

Condition D: looking back at viewer with smile (+6 tokens)

Observations
looking back over shoulder is the most stable and natural looking-back composition. face buried in pillow concentrates attention entirely on the butt, but lacks facial expression. looking back at viewer with smile adds intimacy but costs 6 tokens.
Lab Director: The “face buried in pillow” shot where there’s literally nothing to look at except the butt? Peak fetish. But for gravure purposes, having facial expression adds more value.
looking backwins on versatility.
Face Direction Comparison Table
| Seed | Default (A) | Over shoulder (B) | Face in pillow (C) | Smile back (D) |
|---|---|---|---|---|
| 42 | ![]() | ![]() | ![]() | ![]() |
| 123 | ![]() | ![]() | ![]() | ![]() |
| 789 | ![]() | ![]() | ![]() | ![]() |
Experiment 4: Leg Spread Variations
Outfit: white lace bodysuit
Condition A: No specification (default)

Condition B: spread legs (+2 tokens)

Condition C: knees together (+2 tokens)

Observations
spread legs widens the leg spread and emphasizes the butt area. With bodysuits, wider legs change the crotch area visibility — use with caution. knees together creates a more modest impression but reduces the butt-out feel.
Leg Spread Comparison Table
| Seed | Default (A) | spread legs (B) | knees together (C) |
|---|---|---|---|
| 42 | ![]() | ![]() | ![]() |
| 123 | ![]() | ![]() | ![]() |
| 789 | ![]() | ![]() | ![]() |
Best Combinations
Combining the best from each experiment:
- Back arch:
arched back(balance of natural look and butt emphasis) - Face:
looking back over shoulder(expression visible) - Outfit:
black thong(maximum butt line emphasis)
- Back arch:
deeply arched back, lordosis(maximum emphasis) - Face:
face buried in pillow(butt-focused) - Outfit:
garter belt and stockings(eye-guiding effect)
Summary
Impact Ranking
- Outfit choice — highest impact on output
- Back arch — directly affects butt emphasis
- Face direction — significantly changes composition mood
- Leg spread — effective but limited impact
- Hand position — nearly no effect in rear compositions
Recommended Prompt Parts
| Part | Recommendation | Tokens | Reason |
|---|---|---|---|
| Back arch | arched back | 2 | Best balance of impact and naturalness |
| Face direction | looking back over shoulder | 4 | Most stable. Expression adds value |
| Leg spread | No specification | 0 | Default is fine. Save tokens |
| Hand position | No specification | 0 | Not visible from behind. Save tokens |

![[Experiment] All Fours Pose Test — Effects of Pose Keywords and Angle Combinations](/tips/all-fours-pose-test/A_seed42_0_0000_42.webp)

![[Verified] Image Generation Prompt Best Practices](/tips/prompt-best-practices/cover_0_0000_4517457392071889496.webp)


