Portrait Photography

Start with the subject's identity (age, hair, facial features, makeup) at the sentence beginning, then add clothing, action, lighting, and lens details in descending importance. Use specific nouns like focal lengths and film stocks instead of abstract adjectives, and include at least five realism details such as visible pores or clothing creases.

Z-Image responds well to Chinese natural language with a structure of subject description, art style, environment atmosphere, and composition. Krea 2 prefers long narrative sentences covering subject, scene, composition, lighting, mood, and technical detail rather than fragmented tags.

Use photo series mode: copy the identity lock section (hairstyle, face, makeup, core clothing) verbatim to the start of every prompt. Then vary only four variables per image—lighting, frame, action, and environment—across 3 to 12 images.

This happens when prompts lack specific facial features and skin texture descriptions. Add concrete details like oval face, small freckles, visible pores, and environmental oil sheen, plus natural shadows under eyes and nose to suppress the AI's default smoothing.

Yes, switch to English output using the bilingual terminology mapping reference, which translates photography terms for lighting, lens, skin texture, and style. A universal English negative prompt template is also provided for these models.