Hi there,
I’m interested in comparing two images to analyze layout differences—specifically, comparing a Figma design with the corresponding webpage layout.
Is it possible to perform this kind of visual comparison using the OpenAI Vision model via the API? If so, how to do it?
Any Resources related to this topic will be helpful.
Following this post for any working solutions
See here: https://platform.openai.com/docs/guides/images-vision#analyze-images
Note: The API allows you to load multiple images.
Simple example using GPT-5.2:
Comparison Of Two Eye Illustrations
I. Shared Subject And Composition
Both images depict an extreme close-up of a single human eye centered in the frame, with a vivid green iris and a dark pupil. Each uses a high-contrast, hyper-detailed, digitally illustrated style that emphasizes eyelashes, eyelid contours, and the glossy highlight on the cornea, creating a dramatic, poster-like focal point.
II. Color Palette And Surrounding Elements
The first image surrounds the eye with intense orange and yellow flames, producing a warm, fiery palette that dominates the background. The second image replaces flames with branching white-blue lightning on a deep blue field, establishing a cool, electric palette that reads as stormy and energized.
III. Mood And Visual Impact
The flame-based image conveys heat, urgency, and combustion-like intensity, with the background appearing to engulf the eye from all sides. The lightning-based image conveys charge, tension, and kinetic power, with the bolts radiating outward and giving a sharper, more piercing sense of motion.
IV. Detail Differences Within The Eye Region
In both images, the sclera shows visible red veining and the inner corner of the eye appears reddened, but the surrounding lighting changes the perceived tone. The warm lighting in the flame image adds a more amber cast to the skin and eyelid edges, while the cool lighting in the lightning image shifts those same areas toward bluish and purplish tones, making the eye appear colder and more starkly illuminated.
