Krea 3 Portrait Prompt Dual-Style Demo: Realistic vs. Artistic

This entry demonstrates how krea-3 performs differently on portraits via the pure text-to-image path: the same shooting intent is submitted with two prompt directions, "documentary realism" and "artistic texture," showing the model's controllable range in skin texture, lighting layering, and color tendency. The author pipeline only declares prompt as required, with the rest as optional generation parameters; this run fixes variant=berry (quality-leaning), image_size=portrait_4_3, reasoning_level=medium, and uses negative_prompt to suppress common plastic skin and over-smoothing. It serves as a prompt engineering reference: developers can compare how the two prompts differ in subject description, lighting description, and material description, understand the model's degree of reliance on "detailed prompts," and then transfer this to draft generation for e-commerce portraits, character design, and editorial visuals. The work is genuinely generated by the author pipeline, and the input snapshot is the base and top-level prompt below.