Lock identity, composition, lighting, and wardrobe before style words so a portrait series stays consistent instead of drifting frame to frame.
Le 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.
Remplacez les variables
[SUBJECT]The person or reference image.
[FRAMING]e.g. head-and-shoulders, 3/4.
[LIGHTING]e.g. soft key, rim light.
Quel modèle utiliser
- Nano Banana Pro — Identity-preserving image edits are handled by the image model, not the text model.
Exemple d'entrée
- [SUBJECT] = uploaded reference photo. [USE] = "LinkedIn headshot".
Exemple de sortie
Nous ne publions que des exemples de sorties issus d'exécutions réelles sur un modèle. Ce modèle n'a pas encore été formellement testé, aucune sortie n'est donc affichée. Exécutez-le vous-même avec l'entrée ci-dessus.
Pourquoi cette structure fonctionne
- Structure-before-style stops adjectives from fighting each other.
- An explicit negative list removes common artefacts.
Échecs courants et correctifs
Face drifts between generations.
Correctif : Reuse the same reference and keep identity terms first in the prompt.
Limites connues
- 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.