Key Takeaways
- All 3 conditions successfully produced mirror reflections. “Mirror”-based prompts reliably generate mirror + reflection compositions.
- Condition B (vanity mirror) was the most stable. 3/3 produced vanity mirror + lights + lipstick application with zero artifacts.
- Condition A (full-length mirror) produced reflections in 3/3, but 1 image had a figure duplication artifact. The boundary between reflection and real figure became ambiguous, resulting in 3 visible people.
- Condition C (back view reflection) produced results very similar to Condition A. “Seen through mirror from behind” and “standing in front of full-length mirror” converged to nearly the same composition.
- Mirror compositions default to a back-facing subject with front-facing reflection. The camera position tends to lock behind the subject.
Experiment Design
| Parameter | Value |
|---|
| Model | z-image-turbo |
| Seeds | 3 fixed |
| Outfit | black lace lingerie |
| Background | bedroom (A, C) / bathroom (B) |
| Total | 3 conditions x 3 = 9 images |
Conditions
| Condition | Mirror Prompt |
|---|
| A | reflected in large mirror, standing in front of full-length mirror, looking at own reflection |
| B | reflected in bathroom mirror, applying lipstick, vanity mirror with lights, close-up upper body |
| C | seen through mirror from behind, back view in mirror reflection, looking at mirror |
Results by Condition
Condition A: Full-Length Mirror
1girl, 32yo japanese actress, wearing black lace lingerie, reflected in large mirror, standing in front of full-length mirror, looking at own reflection, bedroom
Observations: 3/3 produced mirror reflections. The common pattern is the subject’s back facing the camera with her front reflected in the mirror. However, seed 1 produced a figure duplication artifact — 3 visible people instead of the expected 2 (1 real + 1 reflection). The boundary between reflection and real figure became ambiguous. Seeds 2 and 3 showed clean 1+1 mirror compositions. The specified full-length mirror frame was present in all outputs.
Condition B: Vanity Mirror
1girl, 32yo japanese actress, wearing black lace lingerie, reflected in bathroom mirror, applying lipstick, vanity mirror with lights, close-up upper body
Observations: 3/3 produced vanity mirror + round bulb lights + lipstick application. The most stable results of all conditions. vanity mirror with lights generated Hollywood-style mirrors with surrounding bulbs, and applying lipstick added a natural action. close-up upper body produced tighter framing than Conditions A and C. No figure duplication occurred in any output. The reflection-to-real relationship was clean and unambiguous.
Lab Director’s Comment
Really into the vanity mirror vibe. Those round bulbs give it this whole mood, and having her actually doing something — applying lipstick — adds so much more depth than just standing in front of a mirror. If you want a mirror composition that won’t break on you, this is the one to go with.
Condition C: Back View in Mirror
1girl, 32yo japanese actress, wearing black lace lingerie, seen through mirror from behind, back view in mirror reflection, looking at mirror, bedroom
Observations: 3/3 produced back-facing subject + front-facing reflection compositions. The results were very similar to Condition A. “Seen through mirror from behind” and “standing in front of full-length mirror” effectively converged to the same “back of subject + front in reflection” layout. Seed 2 featured a decorative gold-framed mirror, and seed 3 had floral wallpaper, showing some background variation. No figure duplication occurred.
Comparison Summary
| Condition | Mirror Present | Reflection Present | Figure Duplication | Framing | Stability |
|---|
| A: Full-length | 3/3 | 3/3 | 1/3 occurred | Full to upper body | Slightly unstable |
| B: Vanity | 3/3 | 3/3 | 0/3 | Upper body close-up | Most stable |
| C: Back view | 3/3 | 3/3 | 0/3 | Full to upper body | Stable |
Practical Tips
- Mirror basics: “Mirror”-based keywords reliably produce mirror + reflection compositions. All 3 conditions generated reflections across all seeds
- Figure duplication risk: Full-length mirror compositions can blur the boundary between reflection and subject, resulting in extra figures. Use the vanity mirror approach for more stability
- Vanity mirror recommended: Combining
vanity mirror with lights with an action (applying lipstick, etc.) gives you integrated control over composition, framing, and situation with minimal artifacts - Conditions A and C converge: “Standing in front of a full-length mirror” and “seen through mirror from behind” produce nearly identical results. To differentiate mirror compositions, specify the mirror type and add an action
Related Articles
![[Verified] Image Generation Prompt Best Practices](/tips/prompt-best-practices/cover_0_0000_4517457392071889496.webp)
