Portrait Style Dual Comparison: Photorealistic and Artistic Image-to-Image Reference

Starting from the same portrait base image, gpt-image-2.6-flare/image-to-image is used to generate two directions—"photorealistic" and "artistic"—as reference samples for prompts. The base image is a medium shot, natural-light half-body portrait generated by the pipeline; both edits use the same base image as the sole source image, and the differences come only from the prompt and the render quality tier: the realistic version emphasizes skin texture, catchlights, and natural skin-tone transitions, while the artistic version emphasizes painterly brushstrokes, color-block simplification, and restrained saturation. It is suitable for developers who need to set a baseline prompt for "image-to-image portrait rewriting," and also works as a comparison sample when tuning style strength. Because the model is weak in scenarios that require precise restoration (proportions, text, logos, materials), this work is positioned as an exploration of style and atmosphere, not a production-grade retouching deliverable.