Upload a photo and describe what you would like to change. Check the generation cost before you start.
Sign up to create your first image free with Nano Banana 2 Lite.
Image Preview
Your generated images will appear here.
Made with Froging AI
Original
ResultA controlled wardrobe change that preserves identity and the candid CCD look.
Change only the cobalt track jacket to a butter-yellow vintage racing jacket with narrow navy piping. Preserve the exact same person, face, hair, headphones, expression, pose, hands, convenience-store background, direct-flash lighting, camera angle, crop, compact-camera texture, and every other detail.
Original
ResultRedesign the backing card while keeping the figure and every accessory stable.
Redesign only the backing card and graphic packaging as an energetic 1980s arcade theme using purple, cyan, black, pixel motifs and geometric grids. Keep the exact same cyclist figure, face, body, pose, lime jacket, helmet, bicycle, camera, bottle, headphones, clear plastic blister shape, front-facing composition, and lighting unchanged. Use fictional decorative symbols only, no readable words.
Original
ResultChange weather and color temperature without rebuilding the miniature scene.
Transform only the weather and season into a cozy snowy winter night: delicate snow on the takeout-box roof and pavement, a few falling flakes, cooler blue ambient light outside, and warmer lantern glow inside. Preserve the exact red box, miniature noodle stall, chefs, diners, food, camera angle, scale, layout, and macro depth of field.
Original
ResultA full material and palette transformation that keeps the model and composition intact.
Restyle only the color palette and materials into a monochrome liquid-silver editorial: reflective chrome tubes, silver eye makeup, cool gray seamless background, and crisp white highlights. Preserve the exact same model identity, facial features, skin texture, gaze, pose, hair arrangement, roller geometry, crop, and high-fashion composition.
Upload a clear photo, illustration, or product picture you want to modify. The original gives the model a reference for the subject and composition.
Explain what to change and what to preserve. For example, change the background while keeping the person, clothes, and framing consistent.
Review the edited image against the original. Check faces, edges, and textures; refine the instruction if needed, then download the result.
Move a portrait or product into a different setting. Describe the new environment and how the lighting should fit it.
Explore a new clothing color, material, or artistic treatment. Say what should stay the same, such as the person’s face, pose, and surroundings.
Try a focused change to an existing visual instead of writing a new scene from scratch. Change one element at a time to compare the effect.
Creative decisions
Editing starts with an existing image, so your prompt does not need to describe a whole new scene. Identify the target of the change, describe its replacement, and name the visible details you want to retain. A focused request is easier to compare with the original.
Choose a source with the subject and framing you want to keep. If the target detail is tiny or hidden, the model has less visual information to work from. Preserve a copy of the original so you can compare independent revisions without losing your starting point.
“Change the chair upholstery” is more specific than “make the room luxurious.” Describe the new color, material, or environment, and distinguish it from nearby objects. For a background change, also consider how light and shadows should fit the new setting.
List the identity, pose, geometry, crop, and other details that matter. The model may still change some of these details. Check the whole output, including areas outside the requested edit, before deciding whether the change is acceptable.
Test the background before combining it with a wardrobe and style change. If a version moves too far from your source, start again from the original with a narrower request. This makes it easier to identify which instruction caused the unwanted difference.
Prompt workshop
In this example, we want to change an armchair’s fabric without changing the room. Name the object, describe the new fabric, and list the details to keep.
Starting idea
Make this room look different.
More specific direction
Change only the armchair upholstery to deep blue velvet with a subtle fabric nap. Preserve the chair shape, wooden legs, position, camera angle, surrounding furniture, wall color, and window lighting. Keep the same composition and realistic contact shadows.
Review the requested edit first, then compare the rest of the output with your source. Pay attention to faces, hands, outlines, repeated patterns, and shadows even when they were not the target.
A broad instruction may be interpreted as a scene-wide redesign.
Try next
Name the target more precisely and list the nearby elements that should remain.
The model may reinterpret identity or geometry during a larger edit.
Try next
Return to the original and request a smaller change with explicit face, pose, and crop constraints.
New materials or backgrounds may not match the existing light.
Try next
Describe the light direction and the shadows or reflections that should connect the edited area to the scene.
Each new result becomes a different reference if you keep reusing it.
Try next
Compare against the original and restart from it when the accumulated changes outweigh what you want to preserve.
Read about each model below, or try one in the generator to see which suits your project.
Plans and credits
Check the cost before each edit. Trying a different prompt creates a new result and uses more credits, so keep a copy of any version you like.
Lite
For trying out AI creation
Billed $149/year
400credits/month
≈ 10 videos
Starter
Perfect for getting started
Billed $299/year
1000credits/month
≈ 25 videos
Pro
For creators and professionals
Billed $499/year
1800credits/month
≈ 45 videos
Enterprise
For teams and businesses
Billed $999/year
4000credits/month
≈ 100 videos
Ultra
For studios and high-volume creators
Billed $1999/year
9000credits/month
≈ 225 videos
Included with every option
Video estimates assume 40 credits per 5-second 720p clip. Actual cost depends on model, length and resolution.
Image and video tools use the same account credit balance. View full pricing details.
Upload an image and describe your requested edit. The model uses your photo and instructions to make an edited version.
The editor’s model selector includes models that support image input. Text-only models remain available in AI Image Generator.
You can describe what to change and what to keep, but you cannot select or lock individual pixels here. Check the rest of the image for unintended changes.
Keep your original upload as a reference and inspect the generated preview. The examples on this page show the original image beside the edited result.
No. You upload an image and describe the edit in words. For precise changes to a selected area, use an editor with layers and masks.
You can request a background replacement while explicitly preserving the person, pose, clothes, and crop. Review identity, hair edges, lighting, and shadows in the output because generated edits may affect those details.
Use the original when you want independent alternatives or need to recover details that drifted. Use a reviewed edited version only when its changes are intentional and should form the starting point for the next request.
Qwen Image 2.1
GPT Image 2.5 Flare