OpenAI's previous image generation model that accepts both text and image inputs and produces image outputs
Settled by community votes across 1 shared challenge, with an AI judge weighing in on each.
GPT Image 1
13.3
arena score
#19 of 23 in Image Editing
Skill signature
Not enough comparable category data
The chart appears once both models have ratings across at least three shared arena categories.
Seedream 4.5
26.1
arena score
#10 of 44 in Text-to-Image
Vote tally
Where the votes landed
GPT Image 1
0%
win rate
Ties
0%
Seedream 4.5
0%
win rate
Shared challenges
1
Challenge by challenge
The strongest take from each model on every shared challenge, with the AI judge's read.
Outfit Transfer Challenge
EditingEdit instruction
“Use Image 1 as the base person. Dress them in the exact elaborate outfit from Image 2 (including all layers, accessories, jewelry, and shoes). Carefully adapt the clothing to the body shape and pose in Image 1 while maintaining realistic fabric behavior, correct proportions, and perfect lighting/shadow matching. Keep the person’s exact face, hair, and background completely unchanged.”
Source
GPT Image 1
Seedream 4.5
AI Judge Analysis
GPT Image 1
- + Excellent fabric textures and realistic lighting integration
- + High identity preservation of the face
- + Captures the intricate pattern of the scarf accurately
- − Significantly alters the vitiligo pattern on the forehead
- − Crop is much tighter than the original source image
- − Background is slightly simplified compared to the source
Seedream 4.5
- + Perfectly preserves the person's face, hair, and vitiligo patterns
- + Maintains the original image composition and wide background
- + Captures all accessories including the gold watch and ring from Image 2
- − Logic error in clothing layering where the skin is visible behind the scarf despite wearing a black shirt
- − Lower image resolution/clarity compared to Model A
- − The lighting on the coat feels slightly flat compared to the scene
Verdict: Seedream 4.5 is the winner for its superior ability to maintain the source person's unique features and the original image's composition, although it makes a mistake by showing skin through the shirt. GPT Image 1 produces a more polished and higher-quality render, but it fails the 'source preservation' requirement by changing the subject's vitiligo pattern and cropping the image.
Next steps
Explore each model
ByteDance's latest image generation model unifying text-to-image and image editing in a single architecture, with improved text rendering and 30-40% faster generation than v4.0