AI Portrait Photography Guide

Keep the same person while editing a portrait

Creating an original portrait and editing an existing person are different tasks. If identity matters, attach the original photo and choose a model that supports reference images. Describe what must stay fixed before describing the new background or light.

A prompt is an instruction, not an identity lock. The standard Generate Image interface does not offer a mask brush to protect selected pixels; inspect the whole result, including areas you asked to preserve.

Set up one controlled edit

1. Open Generate Image, choose a model that accepts image references, and attach a clear original portrait. Confirm that the intended reference is included in the request.
2. If you began with a gallery Use prompt action, attach your source yourself: it prefills the text and, when available and recognized, the model and ratio, but does not bring the example’s original reference photo.
3. List the invariants: face shape, eye spacing, expression, gaze, hairstyle, clothing, body pose, crop, and camera perspective.
4. Choose one change for the first pass: background OR lighting. For a background-only edit, explicitly keep the existing light direction.
5. Keep the ratio aligned with the source, generate, and compare with the original before making another edit.

A tested English background-replacement prompt

This is the exact prompt from our September 3, 2026 background test. It describes a fictional adult in a charcoal knit top. Replace those source-specific details with what is actually visible in your own photo; your adapted result has not been tested here.

Edit the supplied portrait reference. Preserve the adult subject's exact identity, face shape, skin texture, expression, gaze, hair, charcoal knit top, body proportions, crop, and camera perspective. Replace only the plain background with a softly out-of-focus pale sage plaster wall and a single diffused window-shadow shape. Keep the existing neutral daylight direction and realistic edge detail around the hair. Do not beautify, change age, alter facial features, add jewelry, add text, add logos, or introduce another person.

What the reference-image experiment supports

For that one background edit, Nano Banana Pro and GPT Image 2 both kept the face, hair, clothing, and crop close to the reference. The displayed Midjourney V8.2 result changed facial proportions, hair, and framing.

In the separate relighting task, GPT Image followed the requested warm key and cool fill most closely; Nano Banana Pro preserved the person but made the key light harder. These are dated observations of one accepted output per model and task, not guaranteed identity preservation or current-model success rates.

• Compare the source, exact prompts, and all editing outputs

Accept the edit only if the original subject survives

• Identity: compare face outline, eye spacing, nose, mouth, hairline, and expression with the original.
• Pose and crop: check head angle, shoulders, visible hands, body scale, and framing.
• Preserved details: inspect clothing, hair edges, glasses, jewelry, and the light you intended to keep.
• For a background-only edit: verify that the background changed without moving the camera or relighting the face.
• If the person drifts, return to the original reference, remove conflicting restyling requests, and retry one change. Do not build the next edit on an already altered identity.

Choose a different workflow when the goal changes

• Create an original fictional portrait from text
• Text-only portrait comparison: what it can and cannot tell you
• Turn a photo into illustration while checking face and pose