Your real asset stays in the brief
A reference preserves the product, character, layout cue or art direction you need the model to consider, instead of asking text alone to recreate it.
Image to Image
Edit or transform a reference image with AI
Model
GPT Image 2NEWImage editUp to 4KReference images
Upload up to 4 reference images to edit or transform.
Click to upload images or drag and drop
JPEG, PNG, or WEBP, up to 10MB each.
0/4 images uploaded
Image prompt
0/2000
Resolution
Outputs
Credits required:3
Your generated images appear here. Describe what you want to see and click Generate.
Generation tips
- - Describe the subject first, then the environment and the exact action or pose.
- - Name the medium, lighting, lens, color palette, and composition instead of asking for a vague style.
- - Put required wording in quotation marks and keep it short for more reliable typography.
- - Use 1K to explore a prompt, then switch to 4K for the final asset.
AI image to image — edit a real reference with a written instruction
Image to Image starts with the asset you already have. Upload one to four JPEG, PNG or WebP reference images, describe the change, and GPT Image 2 creates edited variations while working from your source rather than an imagined substitute.
That makes the page useful for focused transformations: replace a background, shift the style, alter color, add or remove an element, or turn several references into one new direction. Select 1K, 2K or 4K, an aspect ratio and one to four outputs before generation; the credit quote is shown up front.
- Upload up to four reference images
- Describe a precise visual change
- Edit at 1K, 2K or 4K
What makes image-to-image editing useful
A good edit starts from the right source, a clear instruction and an output format chosen for where the result needs to go.
One instruction can target one change
Ask for a background swap, a different material, added packaging, a new season or a cleaner composition. Focused instructions make the edit easier to evaluate.
More than one reference when context matters
Upload up to four source images when the result needs to draw from several angles, materials, visual references or product details.
How to direct an image-to-image edit
First choose the image or images that contain the information you cannot afford to lose: a product silhouette, a person’s wardrobe, a logo placement, a visual mood or a specific composition. Then write the transformation in plain language. State what should remain, what should change, and the result you want to see.
Keep the instruction narrow on the first pass. “Keep the bottle shape and label, replace the studio background with a sunlit kitchen, natural food photography” is easier to judge than a paragraph requesting unrelated changes. If the first output is near the target, adjust one detail at a time and use a larger resolution only after the direction is approved.
Separate what stays from what changes
Name the elements that must survive the edit before describing the transformation. This anchors feedback to the source image.
Use references with intention
Add a second, third or fourth image only when it contributes a specific detail or visual direction; fewer, clearer sources are easier to evaluate.
Image-editing controls
What you can set for an image-to-image task
The edit workspace keeps source images, the change request and output decisions together before you send the task.
- Up to 4K
- Choose 1K, 2K or 4K output
- 8 ratios
- Choose the target canvas before editing
- 1–4 images
- Edited variations from one request
- 1–4 sources
- JPEG, PNG or WebP reference upload
- PNG export
- Downloadable output from every completed edit
- No watermark
- A clean result for your workflow
Why use FrameAI for image to image
It combines reference upload, prompt-led editing, resolution choices and output history in the same image workspace.
Begin with the actual reference
Use your source material as context for the edit, which is more practical than attempting to reproduce a product or design direction from words alone.
Choose an output size deliberately
Try an edit at 1K first, then choose 2K or 4K for a selected direction that needs more detail or a larger placement.
Adapt the canvas to the destination
Set a square, portrait, landscape or ultrawide ratio based on where the revised image will be used.
Know the cost before the edit runs
The workspace shows the credit total from the requested resolution and output count before you submit the generation.
Compare edited alternatives
Generate up to four outputs to compare how the same reference and instruction resolve into different compositions or treatments.
Return to completed work
Edited images are available in My Creations, alongside your other FrameAI outputs, for later review and download.
Edit an image with AI in three steps
The image-to-image page opens in edit mode, with reference upload first and text guidance immediately below it.
- 1
Upload your references
Add one to four JPEG, PNG or WebP images. Use the images that best communicate the product, composition or visual direction you need to preserve.
- 2
Describe the transformation
Explain what must remain and what should change, then select an aspect ratio, 1K, 2K or 4K resolution, and one to four output variations.
- 3
Generate and download the selected edit
Review the returned images in the workspace, keep the strongest variation and download its PNG from the result or My Creations.
What to do with AI image-to-image editing
Reference-led editing is useful when a generic prompt is not enough and a real image needs to guide the result.
Product scene variations
Keep a product reference while changing the setting, color treatment, season, props or photographic mood for different campaign placements.
Creative restyling and cleanup
Use a reference to explore a new visual treatment, remove distracting elements, add a requested detail or create a cleaner composition direction.
Reference-led concept development
Combine up to four source images to develop a concept around a material, character cue, product detail or established art direction.
AI image to image — frequently asked questions
What is AI image to image?
AI image to image starts from one or more uploaded reference images and a written instruction. GPT Image 2 uses that context to create an edited or transformed PNG output.
How many reference images can I upload?
You can upload one to four reference images for one task. Each source can be a JPEG, PNG or WebP file up to 10MB.
What kinds of changes can I ask for?
You can describe changes such as a new background, colors, lighting, style, composition, props or elements to add or remove. State clearly what should remain from the source and what should change.
Which output sizes and ratios are available?
Choose 1K, 2K or 4K and one of eight aspect ratios: 1:1, 3:2, 2:3, 16:9, 9:16, 4:3, 3:4 or 21:9.
Can I generate multiple edited versions?
Yes. Select one to four output images before generating to compare variations from the same references and instruction.
How is an edit charged?
Credits are calculated from the selected output resolution and number of images. The total appears in the generator before you submit the task.
Use the result in the next workflow
Create a new visual from text, animate an edited still, or use a reference-led image as the first step toward a video project.
Start with the image you already have
Upload the reference, state the change and generate a focused set of alternatives. Keep the source details that matter while moving the visual direction forward.
- Upload up to four references
- 1K, 2K and 4K choices
- PNG results in My Creations