Even OpenAI's Latest Image Model Struggles with Composition

1a3orn · x · 2026-07-04

Beth Barnes notes that even OpenAI's latest image model, which can produce comic-style compositions on a single page, appears erratic, blurry, and rough compared to hand-drawn art. While hand-drawing clearly conveys "what is happening," current AI images still fall short in narrative clarity.

Related event: AI Images Tend to Overcrowd Details, but Negative Space Makes Great Composition(2 posts)→

Original post →

More from Multimodal

Multimodal channel →