Conclusions First
- control generates a normal-sized woman standing on a city street — no giant elements
giantputs a woman above the urban skyline, but scale varies. Comparable to or slightly taller than buildingsgiantessproduces similar results togiant— no clear difference in scalecolossalshows a woman standing in an overhead city view — sometimes depicted at twin-tower-height scalekaiju scaleis the most distinctive — all 3 images feature a Godzilla-like creature behind the woman, with the city rendered as a miniature model, giving the largest sense of scaletoweringplaces the woman among city buildings at roughly building height — no clear difference fromgiantlooking downproduces a giant woman looking down over a city but only adds a viewing angle change, no significant scale increaseholding buildingproduces the woman reaching toward a building top, but actually gripping a building was only confirmed in 1 imagetiny peopleplaces miniature people around a normal-sized woman to create a relative giant effect — no urban background, props like pedestals appear- [Experiment 2] Miniature photography technique application is the most effective. Incorporating low angle, macro depth of field, backlighting, and sky background dramatically improved giant stability (all conditions 3/3) and miniature city scale
- In particular,
macro photography, shallow depth of field, tilt-shiftmost strongly evokes a diorama-style miniature feel backlit silhouette, rim lightcreates a majestic giant feeling with the strongest visual impact
Experiment Design
| Item | Value |
|---|---|
| Model | z-image-turbo (6B, realism-focused distilled) |
| Steps | 8 |
| Sampler | euler |
| Scheduler | ddim_uniform |
| CFG | 1.0 |
| Size | 1024x1024 |
| Seeds | 3 fixed (shared across all conditions) |
| Per condition | 3 images (3 seeds) |
| Total | 9 conditions = 27 images |
Base Prompt
An urban environment was set as the base condition to judge scale by comparison with buildings.
A00: control
| seed 1 | seed 2 | seed 3 |
|---|---|---|
 |  |  |
Observation: All 3 show a normal-sized woman on a city sidewalk. Cars, pedestrians, and shops are in the background at realistic scale. White dress with small accessories. No giant elements.
A01: giant
| seed 1 | seed 2 | seed 3 |
|---|---|---|
 |  |  |
Observation: S1 shows a normal-sized woman on a city street — no difference from control. S2 shows a woman standing above a city at roughly building height. S3 is also an overhead city composition with the head reaching the tops of buildings. Giant appearance in 2/3, but scale is inconsistent.
A02: giantess
| seed 1 | seed 2 | seed 3 |
|---|---|---|
 |  |  |
Observation: S1 shows a normal-sized woman on a city street — no giant effect. S2 and s3 show overhead city compositions at building-top height. No visual difference from giant. Using the feminine form of the word doesn’t change the scale.
A03: colossal
| seed 1 | seed 2 | seed 3 |
|---|---|---|
 |  |  |
Observation: All 3 show a woman standing in an overhead city view. S1 stands on a skyscraper roof at roughly the same height as surrounding buildings. S2 stands between twin towers at equal building height. S3 is surrounded by high-rises at roughly the same height. More stable giant appearance than giant or giantess (3/3), but no notable difference in scale magnitude.
A04: kaiju scale
| seed 1 | seed 2 | seed 3 |
|---|---|---|
 |  |  |
Observation: All 3 images show a Godzilla-like creature appearing behind the woman. The woman and creature stand on a miniature city model with clearly small buildings. Cars and roads are visible as miniature elements — the largest scale differential of all conditions. However, the intent was to evoke kaiju-scale size, but an actual kaiju appeared instead. The woman herself is at roughly 10-story building height, larger than other conditions.
Lab Director comment: Wait, a kaiju actually showed up. But the scale is the biggest of all conditions, and a giant woman standing next to Godzilla is honestly the best visual of the whole experiment.
A05: towering
| seed 1 | seed 2 | seed 3 |
|---|---|---|
 |  |  |
Observation: All 3 show a woman standing among city buildings at roughly building height. No notable difference from giant. towering alone did not amplify the scale.
A06: looking down
| seed 1 | seed 2 | seed 3 |
|---|---|---|
 |  |  |
Observation: All 3 show a woman standing in an overhead city composition with gaze slightly downward. S1 stands among mid-rise buildings looking down; s2 looks down over the city with buildings at knee-to-waist level; s3 stands among low-rises looking down. The “looking down” pose is reflected, but scale is comparable to giant — no scale amplification from adding this keyword.
A07: holding building
| seed 1 | seed 2 | seed 3 |
|---|---|---|
 |  |  |
Observation: S1 stands among buildings placing one hand on top of a building to convey scale. S2 is an overhead city composition with one hand extended but not actually gripping a building. S3 appears to be standing on a rooftop — more like “on top of a building” than giant. Gripping a building is not consistently reproduced; hand-extended poses are more common.
A08: tiny people
| seed 1 | seed 2 | seed 3 |
|---|---|---|
 |  |  |
Observation: All 3 show about 10 miniature figures placed around the normal-sized woman, making her appear relatively giant. No urban background — s1 shows her standing on a pedestal/measurement device in an open space; s2 shows her on a platform surrounded by tiny people; s3 shows a large measurement scale with tiny people nearby. This is a relative scale expression of “large person vs small people” rather than a giant-city comparison.
Experiment 2: Miniature Photography Technique Application
Testing whether incorporating miniature photography techniques (low angle, macro, lighting, sky background) into the prompt can enhance the sense of giant scale. All conditions include giant and use Experiment 1 results as baseline.
A09: Low angle shot
Looking up from ground level makes buildings appear smaller and the woman appear larger.
| seed 1 | seed 2 | seed 3 |
|---|---|---|
 |  |  |
Observation: All 3 show a giant woman standing on a miniature city road with buildings at waist-to-shoulder height. S1 shows cars and miniature street trees at her feet with clear scale contrast. low angle + miniature buildings at her feet combination is stable. Compared to Experiment 1’s giant alone (A01), scale is dramatically improved.
A10: Macro photography + shallow depth of field
Macro lens with tilt-shift effect focuses on the foreground and blurs the background to emphasize a diorama-like perspective.
| seed 1 | seed 2 | seed 3 |
|---|---|---|
 |  |  |
Observation: All 3 show the entire city depicted as a miniature model, with large bokeh circles in the background. Rooftop details and car scale look diorama-like — the strongest miniature feeling of any condition. tilt-shift + shallow depth of field is the most effective for producing miniature effects.
A11: Side lighting
Directional lighting emphasizes contrast in details, giving miniature models realistic dimensionality.
| seed 1 | seed 2 | seed 3 |
|---|---|---|
 |  |  |
Observation: All 3 show an overhead composition with the city depicted as a miniature model. A dramatic fluttering dress hem (s1) appeared. dramatic side lighting contributed more to the woman’s dimensionality than to miniature feel.
A12: Backlit silhouette
Backlighting silhouettes the woman to convey an impression of great size.
| seed 1 | seed 2 | seed 3 |
|---|---|---|
 |  |  |
Observation: All 3 show sunlight rim lighting behind the head, creating a majestic giant feeling. The city skyline spreads at her feet with buildings at knee-to-waist level. backlit + rim light emphasizes her silhouette and visually highlights her scale. The most impressive composition of all conditions.
A13: Sky background + miniature city
Real sky with clouds as background, with a miniature city model below.
| seed 1 | seed 2 | seed 3 |
|---|---|---|
 |  |  |
Observation: All 3 show a realistic cloudy sky in the background with a scale model city at her feet. S1 has green hills in the background, resembling an outdoor diorama. real sky background with clouds + scale model cityscape reproduces outdoor miniature photography.
A14: All techniques combined
Combining low angle + macro + side lighting + sky background.
| seed 1 | seed 2 | seed 3 |
|---|---|---|
 |  |  |
Observation: S1 shows a leaning-forward composition looking down with buildings between them against an overcast sky. S2 and s3 show overhead miniature city compositions. Combining all techniques broadened composition variety while contributing to overall quality improvement rather than one single technique standing out.
Cross Comparison
Experiment 1 vs Experiment 2: Scale comparison (giant / low angle / macro DoF / backlit)
| Seed | giant (A01) | Low angle (A09) | Macro+DoF (A10) | Backlit (A12) |
|---|---|---|---|---|
| s1 | | | | |
| s2 | | | | |
| s3 | | | | |
giant alone mixes in normal-sized images (s1). Adding photography techniques greatly stabilizes miniature city feel and giant scale.
Miniature photography technique comparison (low angle / side light / sky background / combined)
| Seed | Low angle (A09) | Side light (A11) | Sky background (A13) | Combined (A14) |
|---|---|---|---|---|
| s1 | | | | |
| s2 | | | | |
| s3 | | | | |
All consistently generate a miniature city. Low angle is dramatic looking upward; sky background has outdoor diorama realism.
Summary
Experiment 1: Giant Keywords
| Condition | Giant | Urban background | Characteristics |
|---|---|---|---|
| control | None | Street | Normal-sized woman |
| giant | 2/3 | Overhead | Building-height scale, inconsistent |
| giantess | 2/3 | Overhead | No difference from giant |
| colossal | 3/3 | Overhead | Stably at building height |
| kaiju scale | 3/3 | Miniature | Kaiju appears, largest scale |
| towering | 3/3 | Overhead | No difference from giant |
| looking down | 3/3 | Overhead | Looking down pose added |
| holding building | 2/3 | Overhead | Gripping is unstable |
| tiny people | 3/3 | None | Relative expression via miniature humans |
Experiment 2: Miniature Photography Techniques
| Condition | Giant | Miniature feel | Characteristics |
|---|---|---|---|
| Low angle (A09) | 3/3 | High | Upward-looking for impact, cars/trees at feet |
| Macro+DoF (A10) | 3/3 | Highest | Diorama bokeh, strongest miniature feel |
| Side light (A11) | 3/3 | High | Improved dress dimensionality, overhead miniature |
| Backlit (A12) | 3/3 | High | Rim light creates majestic giant feel, most impressive |
| Sky background (A13) | 3/3 | High | Outdoor diorama feel, realistic sky composite |
| Combined (A14) | 3/3 | High | All techniques, varied compositions |
Overall Insights
Giant keywords alone (giant, colossal, etc.) have limits in both stability and scale. Applying miniature photography techniques to the prompt dramatically improved both stability (all conditions 3/3) and scale perception.
Recommended approaches:
- For miniature diorama feel:
macro photography, shallow depth of field, tilt-shift lens effect(A10) - For dramatic, impactful giant feeling:
backlit silhouette, rim light, miniature city skyline below(A12) - For realistic outdoor scale:
real sky background with clouds, scale model cityscape(A13) - For stable basic composition:
low angle shot from ground level, miniature buildings at her feet(A09)
Consistently effective across all: miniature series words (miniature buildings / miniature city / scale model) — depicting the city as a miniature model is the key to conveying scale.
Lab Director comment: Instead of “make it bigger,” the “make the surroundings smaller” approach worked best — that’s an interesting prompt design insight. Miniature photography technique application looks like it has pretty broad versatility. The backlighting one is majestic and I love it.
![[Experiment] Complete Body Type Prompt Guide | 24 Conditions, 72 Images from slim to bbw](/tips/body-type-control-test/d01_curvy_s1_0_0000_2053695854357871006.webp)