Conclusion
- tall / very tall / tall model all show limited visual body type differences
- long legs / very long legs induces outfit changes that expose legs rather than actually lengthening legs
- small head / small face / 8-9 head proportions show no confirmed proportion changes — ineffective
- long legs + heels produces the largest visual change (heels worn + leg-exposing outfits)
- Direct control of body proportions is difficult — the primary effect is indirect outfit and styling changes
What This Article Covers
- Whether height keywords (tall / very tall / tall model) affect body type
- The effect of leg length specifications (long legs / very long legs / slender legs)
- The effect of small head and body ratio specifications (small head / small face / 8-9 head proportions)
- Effects of combined specifications and model-related keywords
Experimental Conditions
| Item | 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 (3 fixed seeds) |
Base Prompt
Experiment 1: Height Keywords
Condition A: Baseline (no modifier)
Observations
Standard body type and height. Approximately 6.5–7 head heights. Standard leg length.
Condition B: tall (1 token)
Observations
No clear difference from A. Limited effect.
Condition C: very tall (2 tokens)
Observations
Similar to B. Height difference from A is visually difficult to confirm.
Lab Director thought: Comparing against the bookshelf — is she actually taller?
Condition D: tall model (2 tokens)
Observations
Similar to A. No clear body type change.
Experiment 2: Leg Length
Condition E: long legs (2 tokens)
Observations
Compared to A, there’s a tendency for more leg exposure. The primary effect is an outfit change (inducing leg-exposing designs) rather than the legs themselves being longer.
Condition F: very long legs (3 tokens)
Observations
Similar to E. Outfits change to mini-length shirt dresses + heels, etc.
Condition G: slender legs (2 tokens)
Observations
Legs are covered by a long-length dress, making the effect difficult to confirm.
Condition H: long legs + heels (4 tokens)
Observations
All 3 images show heels being worn. Leg exposure increases. The largest visual change of all conditions.
Experiment 3: Head and Face Size
Condition I: small head (2 tokens)
Observations
No clear change in head size. No effect.
Lab Director thought: Head isn’t smaller, but does she look more baby-faced?
Condition J: small face (2 tokens)
Observations
Same as I. No effect.
Experiment 4: Head-to-Body Ratio
Condition K: 8 head proportions (3 tokens)
Observations
Limited effect. Head-to-body ratio change is difficult to confirm.
Lab Director thought: Writing “8 head proportions” produced someone with 8 heads (seed-dependent maybe). There’s a risk that number + “head” combinations get misinterpreted.
Condition L: 9 head proportions (3 tokens)
Observations
Same as K. Numerical specifications are hard to reflect.
Experiment 5: Model Keywords & Combined Specifications
Condition M: model body (2 tokens)
Observations
Tendency for more outfit design variation (halter necks, belted pieces, etc.). Body type change is limited.
Condition N: tall + long legs + small head (combined, 5 tokens)
Observations
No dramatic body proportion change even with combined specifications. Only outfit change effect.
Condition O: fashion model proportions (3 tokens)
Observations
Outfits shift toward leg-emphasizing styles (mini dresses, heels, etc.). Body proportions themselves unchanged.
Cross-Comparison
Height Keywords (baseline / tall / very tall / tall model)
| Seed | Baseline (A) | tall (B) | very tall (C) | tall model (D) |
|---|---|---|---|---|
| 42 | | | | |
| 123 | | | | |
| 789 | | | | |
Visual differences across all 4 conditions are limited.
Leg Length (baseline / long legs / very long legs / long legs + heels)
| Seed | Baseline (A) | long legs (E) | very long legs (F) | long legs + heels (H) |
|---|---|---|---|---|
| 42 | | | | |
| 123 | | | | |
| 789 | | | | |
long legs + heels produces the largest visual change. The primary effect is outfit change rather than actual leg length.
Head Size and Body Ratio (baseline / small head / small face / 8head / 9head)
| Seed | Baseline (A) | small head (I) | small face (J) | 8 head (K) | 9 head (L) |
|---|---|---|---|---|---|
| 42 | | | | | |
| 123 | | | | | |
| 789 | | | | | |
No proportion changes confirmed in any condition.
Key Findings
Ineffective / Limited Keywords
| Category | Keywords | Result |
|---|---|---|
| Height | tall / very tall / tall model | No visual difference |
| Head size | small head / small face | No effect |
| Body ratio | 8 head proportions / 9 head proportions | No effect |
| Combined | tall + long legs + small head | No body type change |
Keywords with Indirect Effects
| Keyword | Primary Effect |
|---|---|
| long legs / very long legs | Induces leg-exposing outfits (legs themselves don’t get longer) |
| model body | Refined styling |
| fashion model proportions | Model shoot-style styling change |
Most Effective Combination
long legs + high heels produces the largest visual change of all conditions. The heel specification is reliably reflected, and leg exposure also increases. Rather than directly changing body proportions, the practical approach is to create “conditions that make legs look longer.”
Summary
In z-image-turbo, directly controlling body proportions (body ratio, leg length, head size) through prompts is difficult. If you want legs to appear longer, the best approach is an indirect one through outfits and accessories using long legs, high heels.
![[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)
![[Experiment] MILF & Mature Prompt Comparison — Output Differences Between milf / mature / mother](/tips/milf-mature-descriptor-test/I_seed42_0_0000_42.webp)
![[Verified] Image Generation Prompt Best Practices](/tips/prompt-best-practices/cover_0_0000_4517457392071889496.webp)
