Lock identity, composition, lighting, and wardrobe before style words so a portrait series stays consistent instead of drifting frame to frame.
The prompt
Create an editorial portrait of [SUBJECT] for [USE]. Preserve facial identity and natural skin texture. Composition: [FRAMING]. Lighting: [LIGHTING]. Wardrobe: [WARDROBE]. Background: [BACKGROUND]. Color grade: [PALETTE]. Avoid plastic skin, distorted hands, text, logos, and over-sharpening.
Replace the variables
[SUBJECT]The person or reference image.
[FRAMING]e.g. head-and-shoulders, 3/4.
[LIGHTING]e.g. soft key, rim light.
Which model to use
- Nano Banana Pro — Identity-preserving image edits are handled by the image model, not the text model.
Example input
- [SUBJECT] = uploaded reference photo. [USE] = "LinkedIn headshot".
Example output
We only publish example outputs from real model runs. This template has not been formally tested yet, so no output is shown. Run it yourself with the input above.
Why this structure works
- Structure-before-style stops adjectives from fighting each other.
- An explicit negative list removes common artefacts.
Common failures & fixes
Face drifts between generations.
Fix: Reuse the same reference and keep identity terms first in the prompt.
Known limitations
- Image models vary; results are not guaranteed and need review.
FAQ
Where are the tested portrait prompts?
The image gallery has portrait prompts with real example outputs. This page is a reusable template for building your own.