Key Findings
- Standing bent-over is less stable than all-fours —
bent overalone produces only a slight lean in most cases - Hand anchoring is the key to deeper bends —
hands on kneesandhands grabbing anklesstabilize the forward bend - Wall pose works with
wide stance+ upper body lean —hands on wallalone produces upright poses, but addingwide stanceandtorso tilted forward toward wallreliably creates the butt-out composition (3/3) lace lingerie settends to become a bodysuit in standing poses —black thong and matching bralettemaintains separationfrom behindandrear vieware equivalent (same conclusion as the all-fours test)- Recommended:
hands on kneesfor forward bend,wide stance+torso tilted forward toward wallfor wall pose
What You’ll Learn
- Pose descriptions for standing forward-bend + butt-out compositions
- How hand position specifications dramatically affect bend stability
- How to achieve the “hands on wall, butt pushed out” pose (
wide stanceis the key) - Lingerie type differences in standing poses
- Comparison with all-fours test results
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: Pose Description Comparison
Condition A: bent over (2 tokens)

Observations
All 3 produced a slight forward lean with looking back — more of a “showing butt while standing” than a deep bend. Butt is emphasized in all outputs. lace lingerie set became a bodysuit in 3/3 cases.
Condition B: bending forward (2 tokens)

Observations
Nearly identical to Condition A. No practical difference between bent over and bending forward.
Condition C: leaning forward, arched back (4 tokens)

Observations
arched back adds some back line variation, but bend depth is similar to A/B. For standing poses, hand anchoring is more effective than arched back for deepening the bend (verified in Experiment 4).
Condition D: bent at waist, straight legs (4 tokens)

Observations
Only seed 42 produced a clear waist bend with straight legs. Seeds 123 and 789 were nearly upright. straight legs has high variance and is unreliable.
Condition E: touching toes (2 tokens)

Observations
Actual toe-touching pose rarely appears. Results are mostly upright standing rear views. The semantic intent doesn’t translate well.
Condition F: ass out, arched back (4 tokens)

Observations
ass out emphasizes the butt composition but doesn’t produce a forward bend. The pose remains mostly upright.
Pose Comparison Table
| Seed | bent over (A) | bending fwd (B) | lean+arch (C) | bent at waist (D) | touch toes (E) | ass out (F) |
|---|---|---|---|---|---|---|
| 42 | ![]() | ![]() | ![]() | ![]() | ![]() | ![]() |
| 123 | ![]() | ![]() | ![]() | ![]() | ![]() | ![]() |
| 789 | ![]() | ![]() | ![]() | ![]() | ![]() | ![]() |
Experiment 2: Angle Description Comparison
Condition A: from behind
Same prompt as Experiment 1 Condition A — refer to those images.
Condition B: rear view (2 tokens)

Condition C: from behind, low angle (4 tokens)

Condition D: from behind, full body (4 tokens)

Observations
from behindandrear vieware equivalent (same as all-fours test)low angleemphasizes leg length in standing posesfull bodyincludes the entire figure but reduces butt close-up impact
Angle Comparison Table
| Seed | from behind (A) | rear view (B) | low angle (C) | full body (D) |
|---|---|---|---|---|
| 42 | ![]() | ![]() | ![]() | ![]() |
| 123 | ![]() | ![]() | ![]() | ![]() |
| 789 | ![]() | ![]() | ![]() | ![]() |
Experiment 3: Lingerie Type Comparison
Condition A: lace lingerie set
Same prompt as Experiment 1 Condition A — refer to those images.
Condition B: sheer babydoll nightgown (3 tokens)

Condition C: black thong and matching bralette (5 tokens)

Condition D: white lace bodysuit (3 tokens)

Condition E: garter belt and stockings with lingerie (6 tokens)

Observations
lace lingerie setbecomes bodysuit in standing poses (3/3) — didn’t happen as much in all-foursblack thong and matching bralettemaintains separation. Best for butt line emphasis in standing toogarter belt and stockingspairs especially well with standing — full leg visibility makes the garter-stocking intersection point a strong eye-guiding element
Note: Garter may be the standout choice for standing poses. With the full leg visible, the stocking-top × garter-line crossing point on the thigh becomes a natural focal point. In all-fours poses, the knee-down area is hidden, cutting this effect roughly in half.
Lingerie Comparison Table
| Seed | lace set (A) | babydoll (B) | thong (C) | bodysuit (D) | garter (E) |
|---|---|---|---|---|---|
| 42 | ![]() | ![]() | ![]() | ![]() | ![]() |
| 123 | ![]() | ![]() | ![]() | ![]() | ![]() |
| 789 | ![]() | ![]() | ![]() | ![]() | ![]() |
Experiment 4: Auxiliary Elements (Hand Position & Props)
This experiment contains the most important discovery: hand placement determines bend depth.
Condition A: looking back over shoulder (+4 tokens)

Condition B: hands on knees (+3 tokens)

Observations
hands on knees most reliably induces the forward bend. Anchoring the hands at knee level naturally pulls the upper body forward. 1/3 (seed 42) produced the ideal deep bend + butt-out composition.
Condition C: hands on wall (+3 tokens)

Observations
hands on wall generates a wall-leaning pose rather than a forward bend — nearly upright. Notably, lingerie tends to maintain separate pieces (bra + panties) in this context.
Condition D: hands grabbing ankles (+3 tokens)

Observations
hands grabbing ankles produces the deepest forward bend — deeper than hands on knees. However, the pose can look unnatural. Overall balance favors hands on knees.
Condition E: bent over desk (+3 tokens)

Observations
Desk acts as contextual anchor, reinforcing the bent-over context. Bend depth is shallow-to-moderate. Adds narrative quality (office vibe).
Note:
hands on kneesproduces the result closest to a classic gravure bent-over shot.grabbing anklesgoes too far and can look like a gymnast stretch, makinghands on kneesthe more practical choice. The desk version adds an office-scene quality.
Auxiliary Comparison Table
| Seed | Look back (A) | Knees (B) | Wall (C) | Ankles (D) | Desk (E) |
|---|---|---|---|---|---|
| 42 | ![]() | ![]() | ![]() | ![]() | ![]() |
| 123 | ![]() | ![]() | ![]() | ![]() | ![]() |
| 789 | ![]() | ![]() | ![]() | ![]() | ![]() |
Experiment 5: Deep Dive into Wall Poses — Spread Stance + Forward Lean
Experiment 4 Condition C (hands on wall) collapsed into upright wall-leaning. Here we test 8 wall pose variations to find a working combination.
Phase 1: Adding bent-over descriptors (Conditions A–E) — All Failed
All 5 conditions below produced upright rear shots despite adding bent over, ass out, hips pushed back, etc. The wall environment strongly pulls the pose toward upright.
| Condition | Pose Description | Result |
|---|---|---|
| A | bent over, hands on wall, ass out | Upright rear (3/3) |
| B | bent over, hands pressed against wall, hips pushed back | Upright rear (3/3) |
| C | leaning against wall, bent forward, ass out, arched back | Upright rear (3/3) |
| D | hands flat on wall, bending forward at waist, ass out | Upright rear (3/3) |
| E | pushing against wall with both hands, bent over, ass out | Upright rear (3/3) |
Wall Pose Phase 1 Comparison
| Seed | A | B | C | D | E |
|---|---|---|---|---|---|
| 42 | ![]() | ![]() | ![]() | ![]() | ![]() |
| 123 | ![]() | ![]() | ![]() | ![]() | ![]() |
| 789 | ![]() | ![]() | ![]() | ![]() | ![]() |
Phase 2: Adding Spread Legs + Forward Lean (Conditions F–H)
The failure pattern was the model pressing the body against the wall. The fix: spread the feet away from the wall and tilt only the upper body forward.
Condition F: legs apart (Failed)

Observations
legs apart had no effect on leg spread — same upright rear as Phase 1 (3/3). This keyword doesn’t translate to open-leg standing poses.
Condition G: wide stance + torso tilted forward (Success)

Observations
3/3 produced wide-spread legs with upper body leaning toward the wall and butt pushed back. wide stance reliably opens the legs, and torso tilted forward toward wall tilts the upper body toward the wall. Feet positioned away from the wall naturally push the butt out. lace lingerie set maintained separates (bra + panties) in 2/3 — a side benefit of the wall pose context.
Condition H: spread legs + bending upper body toward wall (Success)

Observations
Same effect as Condition G. spread legs also reliably opens the legs (3/3). bending upper body toward wall tilts the torso, pushing the butt backward. palms on wall is equivalent to hands on wall but gives a slightly stronger “flat hands” impression.
Note: The key turned out to be the leg spread.
wide stancereads like an actual photography posing direction and functions accordingly. It’s notable thatlegs apartdoes nothing whilewide stanceworks reliably.
Wall Pose Phase 2 Comparison
| Seed | legs apart (F) | wide stance (G) | spread legs (H) |
|---|---|---|---|
| 42 | ![]() | ![]() | ![]() |
| 123 | ![]() | ![]() | ![]() |
| 789 | ![]() | ![]() | ![]() |
Comparison with All-Fours Test
| Aspect | All-Fours | Standing Bent Over | Wall Pose |
|---|---|---|---|
| Best pose | crawling on bed | standing, bent over, hands on knees | hands on wall, wide stance, torso tilted forward toward wall |
| Butt-out stability | Med-High (2/3) | Low-Med (1/3) | High (3/3) |
| Best angle | from behind | from behind | from behind |
| Outfit stability | Separates maintained | Tends to become bodysuit | Separates maintained |
| Natural look-back | Occurs spontaneously | Requires explicit specification | Occurs spontaneously |
| Garter impact | Knee-down hidden | Full leg visible — best for garter | Full leg visible — best for garter |
| Leg spread | — | — | Stable with wide stance |
Summary
Pose Descriptions
| Description | Tokens | Bend Depth | Stability | Notes |
|---|---|---|---|---|
bent over | 2 | Shallow | Med | Slight lean only |
bending forward | 2 | Shallow | Med | Same as bent over |
leaning forward, arched back | 4 | Shallow | Med | arched back has limited effect |
bent at waist, straight legs | 4 | Med (high variance) | Low | Ideal when it works |
touching toes | 2 | Very shallow | Low | Becomes upright |
ass out, arched back | 4 | Shallow | Med | Emphasizes butt but no bend |
Auxiliary Elements (Hand Position Is Key)
| Description | Extra Tokens | Bend Depth | Recommendation |
|---|---|---|---|
hands on knees | 3 | Med-Deep | Best pick |
hands grabbing ankles | 3 | Deep | Risk of unnatural pose |
bent over desk | 3 | Shallow-Med | Adds narrative |
hands on wall | 3 | Very shallow | Becomes upright (see Exp. 5 for fix) |
Wall Pose (Experiment 5 Results)
| Description | Butt-out | Leg Spread | Stability | Notes |
|---|---|---|---|---|
hands on wall + bent over variants (A–E) | None | None | — | All become upright |
hands on wall, legs apart (F) | None | None | — | legs apart ineffective |
hands on wall, wide stance, torso tilted forward toward wall (G) | Med-High | 3/3 | High | Best pick |
palms on wall, spread legs, bending upper body toward wall (H) | Med-High | 3/3 | High | Equivalent to G |
Recommended Prompts
- Pose:
standing, bent over, hands on knees(hand anchor stabilizes bend) - Angle:
from behind(2 tokens is enough) - Outfit:
black thong(maintains separation, best butt emphasis) - For garter showcase:
garter belt and stockingspairs especially well with standing
- Pose:
hands on wall, wide stance, torso tilted forward toward wall, hips back(spread legs + lean upper body toward wall) wide stanceis the key for leg spread —legs apartdoesn’t work- Feet positioned away from wall naturally push the butt out
lace lingerie setalso maintains separates in this context (wall pose side benefit)
Note: Two usable patterns emerged —
hands on kneesfor the classic bend, and the wall pose for the “push your butt out” look. The wall pose requireswide stance, or it collapses to upright every time. Being able to reproduce the classic “hands against the wall” gravure pose is useful for variety.


















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

