Krea 2 Medium Turbo Product Photography Texture: E-commerce-grade Image Generation Prompt Approach

Taking the fast image generation capability of krea-2-medium-turbo as a premise, this demonstrates an approach to breaking down prompts for e-commerce-grade product images: splitting the scene into five sections—"subject material—lighting—background—lens and aspect ratio—post-processing texture"—using natural language descriptions instead of stacking keywords, and controlling fidelity and delivery channel ratios through creativity and aspect_ratio. The example subject is a frosted glass perfume bottle, because it is most sensitive to lighting descriptions in terms of highlight shape, edge refraction, and shadow softness/hardness, making it easy to observe the impact of each sentence in the prompt on the final image. Applicable scenarios: rapid solution validation for e-commerce main images and detail page styles, batch draft testing across multiple ratios for delivery channels (1:1 main image, 4:5 feed, 16:9 horizontal banner), and prompt template debugging before integrating the image generation step into design tools. This model is positioned as a speed-first distilled variant, suitable for high-frequency iteration and concept exploration; high-fidelity final artwork is still recommended to be finalized manually or through a higher-fidelity workflow. This work only uses the schema-declared prompt, creativity, aspect_ratio, seed, and output capabilities, and does not involve undeclared editing or upscaling parameters.