When prompting AI image generation for canned or bottled products, explicitly request condensation on the container. This one small detail makes the render look substantially more realistic and is a fast fix for the flat, "too plastic" look that bare product renders otherwise have.
Condensation reads as a visual cue of a cold, physical object under real lighting and moisture — a texture cue image models rarely infer unprompted from a plain "product render" request, but that's cheap to specify explicitly in the prompt.
Condensation isn't the only realism lever: output resolution is treated as a hard requirement too — 4K is specified whenever a generated product image needs to read as photorealistic, on the same logic as the condensation trick (bare/low-res renders read as 'plastic').
Из тем: Prompting Foundations: Briefing the Model Like a Director