Background swap
Move a portrait from a bedroom to a studio, a street, or a seasonal set without changing the face or the pose.
Image editing · Workflow guide
AI Image Editor: Change Background, Outfit, and Light
The fastest way to ruin an edit is to ask for everything at once. A reliable AI photo edit reads like a change order: these details are protected, this one layer changes, this is the new scene, and these are the failure cases to avoid.
- Independent guide
- No sign-in here
- Prompt runs locally
- Generator opens flashomni.ai
Quick answer
An AI image editor uses a source photo plus a text instruction. List what to preserve first, then what to replace, then the new scene and light. Name the negatives explicitly: no identity change, no extra people, no distorted hands. Change one layer per pass.
At a glance
Best input
One sharp, front-facing photo with even light
Best output
The same person, with one layer changed
Image models
Nano Banana 2.1, GPT Image 2, Seedream 5.0, FLUX Kontext
Next step
Animate the edited still on Flash Omni
Build your prompt
Pick a preset, adjust the subject and scene, then copy the prompt. The optional photo preview stays in your browser: this page does not upload it. The generator is a separate site with its own terms and pricing.
1. Choose a starting point
No file selected
2. Copy the prompt
Copying the prompt does not create an account or charge credits. The hosted generator opens a separate site with its own terms and pricing.
What you can make
Outfit restyle
Change wardrobe and accessories while keeping body proportions and skin tone. Useful for concept previews and lookbooks.
Relight a portrait
Replace flat phone light with window light, golden hour, or a controlled studio key. Keep the catchlights plausible.
Seasonal portrait
Add a Halloween, holiday, or campaign context to an existing photo without rebuilding the person from scratch.
How it works
Choose a source that gives the model something to keep
Use a sharp photo with a clear face, visible shoulders, and no extreme shadows. A good source is half the edit.
Write the protect list first
Name the face, skin tone, age, body proportions, pose, and number of people. The model needs to know what is not allowed to change.
Change one layer
Pick background, outfit, or light for this pass. Stacking three changes makes it impossible to tell which instruction caused the failure.
Inspect edges and hands
Check hair edges, glasses, fingers, jewelry, and where the subject meets the new background. Fix one defect per follow-up prompt.
Image edit formula
Use this structure as a checklist. Replace the bracketed fields, keep the protection lines, and remove anything that does not apply to your shot.
Use the uploaded photo as the primary reference.
Preserve: [face, skin tone, age, body proportions, pose, number of people].
Edit request: [one layer to change].
Place the subject in: [new scene].
Camera and framing: [shot size, lens, angle].
Light and finish: [key light, texture, realism].
Avoid: altered identity, extra people, extra limbs, duplicated fingers, distorted anatomy, text, logos, watermarks.- Lead with preservation. If the first sentence asks for a change, the model may redesign the face.
- One edit per pass. Generate, inspect, then continue from the best result.
- Say what the light does, not how the image feels. “Warm window light from the left” is actionable.
- Keep the same number of people. Group photos fail when the model adds a background character.
Model notes
These notes describe where each model tends to fit. They are not a ranking. Model availability changes by provider: WaveSpeed, SeeAPI, FAL, Replicate, and Kie all serve image and video models in this class. The hosted generator shows the available models and credit cost before you run a job.
- WaveSpeed
- SeeAPI
- FAL
- Replicate
- Kie
| Model | Best fit |
|---|---|
| Nano Banana 2.1 | Reference-heavy edits where the face and identity must survive a wardrobe or scene change. |
| GPT Image 2 | Detailed natural-language instructions and clean studio compositions. |
| Seedream 5.0 | Photoreal skin, fabric, and product textures during an edit. |
| FLUX Kontext | Targeted local edits when only one area of the frame should change. |
Three prompts you can adapt
Background swap without a face change
This is the safest first edit. Protect the person, change the room, and keep the original crop.
Use the uploaded photo as the primary reference. Preserve the subject's face, hair, skin tone, body proportions, and pose. Replace the background with a deep charcoal studio with a soft gradient and a clean floor line. Keep the same framing and 85mm portrait look. Add soft key light from camera left. Avoid altered identity, extra people, distorted hands, text, and watermarks.Wardrobe restyle for a lookbook
Change the garment, not the model. Keep the pose and the original light direction so the result still reads as the same photo.
Use the uploaded photo as the primary reference. Preserve the model's face, hair, skin tone, body proportions, and pose. Replace the outfit with an elegant black tailored jacket and subtle silver earrings. Keep the pale grey studio backdrop. Use soft editorial lighting and realistic fabric texture. Keep the same number of people. Avoid changing the face, adding accessories that were not requested, or distorting the hands.Relight a flat phone photo
A relight is a lighting edit, not a style filter. Name the key direction and the fill so the face keeps its shape.
Use the uploaded photo as the primary reference. Preserve the person's face, expression, and pose. Relight the portrait with warm low sun from camera left and a cool ambient fill from the right. Add a soft golden rim on the hair. Keep the background simple and slightly out of focus. Do not change the outfit or identity. Avoid plastic skin, clipped highlights, text, and watermarks.Common mistakes
“Make it better”
A vague instruction gives the model permission to redesign everything. Name the one thing that should change.
Changing face and scene together
When identity and environment change in the same pass, you cannot tell which one broke the likeness.
No protect list
If you do not explicitly protect age, skin tone, and body proportions, the model may treat them as style choices.
Ignoring hands and glasses
Inspect fingers, frames, earrings, and hair edges at full size before accepting the edit.
FAQ
Can an AI image editor keep my face?
It can preserve a recognizable likeness when you use a clear source photo and put preservation instructions first. It is not a biometric identity guarantee; inspect the eyes, teeth, and hairline before publishing.
What photo works best?
A sharp, front-facing photo with even light, visible shoulders, and no heavy filters. Avoid sunglasses, extreme angles, and photos where the face is partly hidden.
Can I edit a product photo?
Yes, but protect the product shape, label, and logo placement explicitly. For packaging, use a reference-first model and check the text after every pass.
Does this page upload my photo?
No. The optional preview stays in your browser. This page only builds the prompt; the actual edit happens in your image model.
How many edits should I stack?
One layer per pass. Background, outfit, and light can be changed in sequence, but each change should be reviewed before the next one.
Can I animate the edited photo?
Yes. Export the edited still, then use the image-to-video workflow to add a slow push-in, a head turn, or atmospheric motion.
Related workflows
Turn the prompt into a clip
Open the hosted generator, paste the prompt, and run one short test. Keep the first generation small; a clean five-second clip is easier to judge than a long one.
Flash Omni is an independent guide and is not affiliated with Google. The hosted generator is a separate product with its own terms and pricing.