I’m sending product photos to an OpenAI vision model and trying to decide whether to request high detail every time. Some images need close inspection of labels or small components; others only need a broad description.

How much does the image detail setting change the results in practice, and is there a noticeable difference in usage for high-resolution image input? I’d be interested in examples where the extra detail made a real difference.