Image to image
Start generation from an uploaded image instead of relying only on a text prompt.
Preview
More image tools
After image-to-image generation, continue with Nano Banana 2, GPT image 2, Seedream 5 Lite, AI Image Editor, background removal, and image upscaling tools.
AI image to image
Upload a reference image and generate a controlled new image
The Image to Image page starts from an existing visual: upload a product image, portrait, sketch, screenshot, or layout reference, then describe what should stay and what should change. The default model is Nano Banana 2, useful for subject preservation, background replacement, style redesign, product image revisions, and social asset expansion.
Reference image generationSubject preservationBackground and style editsSwitchable models
Keep + change
Separate source-image details that must stay consistent from the background, style, object, or lighting changes you want.
Image model
Nano Banana 2 is the default. Switch to GPT image 2 or Seedream 5 Lite when a different image-to-image behavior fits better.
Which image editing tasks fit AI image to image
When a reference image defines the subject, composition, or brand direction, image-to-image is more useful than pure text-to-image: it anchors the result to the source while generating a new scene, style, or marketing version.
Reference-driven generation
Upload one or more images as visual context so the model can generate around a subject, composition, color, material, or layout.
Preserve source information
Use the prompt to call out product shape, person identity, pose, logo, texture, edges, color, or spatial relationships that must stay intact.
Replace background and style
Keep the subject while changing background, surface, scene, lighting, material, color tone, seasonality, or advertising mood.
Make targeted image revisions
Use it to add details, remove distractions, change small objects, adjust composition, or create a version better suited for publishing.
Choose a model by goal
Nano Banana 2 is a strong default for image-to-image drafts; GPT image 2 fits complex prompts and reference editing; Seedream 5 Lite is useful for quick variation tests.
Pair with other image tools
Use text-to-image when there is no source image, background removal when you only need a cutout, and image upscaling when you need a sharper final output.
Common AI image to image use cases
Image-to-image turns existing visuals into new usable assets: preserve product identity, replace shooting environments, redesign visual style, or expand sketches and screenshots into complete images.
Revise product images while preserving identity
Use the original product photo as a reference and generate a new image for product pages, ads, or social posts while keeping the core appearance.
Upload a clear product image
Prefer a product photo where the subject is complete, edges are clear, and materials or key details are visible.
Specify what must stay
State which shape, proportion, color, texture, logo, port, button, or package text cannot change.
Compare against the source
Before using it on a product page or ad, compare output and source for product identity, edges, and detail drift.
Keep the subject and replace the scene
Place the subject from the uploaded image into a new indoor, outdoor, tabletop, studio, seasonal, or advertising scene without reshooting.
Lock the subject with the source
Let the reference image define the subject outline, pose, scale, angle, and main edges.
Describe the new scene
Specify background environment, light direction, surface material, depth of field, color tone, and image use case.
Check the blend
Review shadows, reflections, contact edges, perspective, and the overall fit between subject and new background.
Restyle an existing image
Keep the composition, person, object, or layout direction while testing a new color system, material treatment, illustration style, lighting mood, or brand look.
Upload an existing visual
Start from a photo, mockup, sketch, screenshot, cover, or older content image.
Set style boundaries
Describe the style change you want, then list what cannot drift, such as identity, product appearance, composition, or text areas.
Compare different results
If preservation or change is off, adjust the prompt, switch models, or use the generated image as the next reference.
How to use AI image to image online
01
Upload reference images
Add images that should guide the subject, product identity, pose, layout, style, composition, or brand direction.
02
Write keep and change instructions
Separate source details that must stay from the background, style, lighting, object, material, or scene changes you want.
03
Choose a model and generate
Generate first with the default Nano Banana 2 model, or switch to another image model that supports image-to-image, then compare and refine.
AI image to image FAQ
What is AI image to image?
AI image to image is an image tool where uploaded images guide the generated result. Your prompt explains which source details to preserve and how the background, style, lighting, objects, or details should change.
Can image-to-image preserve the original subject?
Yes. Upload a clear reference and explicitly preserve subject identity, product appearance, pose, composition, color, material, and key details in the prompt.
Can I replace a background with image-to-image?
Yes. Describe the new background, surface, room, light, and image purpose while asking the model to keep the subject outline, proportions, and important source details.
Which models work well for image-to-image editing?
This page starts with Nano Banana 2 by default. You can also switch to GPT image 2 or Seedream 5 Lite depending on reference count, ratio, resolution, and style needs.
When should I not use image-to-image?
If you do not have a reference image, use text-to-image. If you only need to remove a background, use Background Remover. If you need higher clarity or size, use Image Upscaler.
Start image-to-image
Upload a reference, adjust the prompt, and generate an edit.
Reference image InputKeep + change Prompt