Pose & Character Mashup

Image Editing

4 models were given the same image and edit instruction, and the community voted blind on which outputs looked best. How it works

This prompt is challenging because it forces models to do two hard things at once: accurately transfer complex body poses and joint angles from Image 1 while strictly preserving the exact face, identity, and clothing style from Image 2. Most models struggle with unnatural limb distortions, identity blending, incorrect fabric behavior on the new pose, and mismatched lighting, making it a strong test of advanced spatial reasoning and character consistency.

Blind Vote This Challenge

#1 — FLUX.2 [klein] 9B

Edit instruction

“Use Image 1 as the exact pose reference and Image 2 as the character reference. Recreate the person/character from Image 2 in the exact dynamic pose and body position from Image 1. Keep the exact face, hair, clothing style/details, and expression from Image 2. Match the lighting and environment of Image 1. The final image must show the character from Image 2 performing the precise action/pose from Image 1 with perfect anatomy and natural integration.”

Voters were asked to judge by Pose Accuracy Character Consistency Correct Anatomy Lighting Match

Challenge Rankings

4 models

#	Model	Price	¢/img	Elo
1	FLUX.2 [klein] 9B Black Forest Labs	$	0.6¢	1134
2	GPT Image 1 Mini OpenAI	$	0.5¢	1126
3	Seedream 5.0 Lite ByteDance	$$	3.5¢	1100
4	FLUX.2 [dev] Turbo fal	$	0.8¢	997

FLUX.2 [klein] 9B leads the category with 1134 Elo, though GPT Image 1 Mini remains highly competitive with a 60.0% win rate and similar $0.005–$0.006 budget pricing. A significant 129-point Elo gap separates the top performers from the mid-tier FLUX.2 [dev] Turbo, illustrating a sharp drop-off in the ability to maintain character consistency across complex pose transfers.