Use one image as the reference
Upload a JPEG, PNG, or WebP image to establish the starting subject, composition, and visual context.
IMAGE TO IMAGE
Edit an image with AI
Upload one reference image and describe the change you want. Ikat uses GPT Image 2.5 to generate a new version while you review what changed and what stayed recognizable.
Reference image Required
Add one reference image before generating on this Image-to-Image page.
REFERENCE-BASED IMAGE EDITING
Start from the image you already have
Describe the change in plain language
Tell the model what you want to change and which important details you want it to keep recognizable.
Keep the original and the new result
The generated edit becomes a new Ikat creation. Your original reference is not overwritten.
COMMON EDITING DIRECTIONS
Use a reference to explore a specific visual change
Change the setting
Keep the main subject as the reference and describe a different background, environment, or lighting direction.
Explore a visual style
Use the reference for composition and subject context, then describe the rendering or visual treatment you want to test.
Refine one visible detail
Ask for one focused change at a time, then compare the new result with the original before continuing.
These are examples, not guarantees of pixel-perfect preservation.
HOW IMAGE TO IMAGE WORKS
Upload, describe, generate
Upload one reference image
Choose the image that should guide the new result. Image V1 accepts one reference per generation.
Describe one clear change
Explain what should change and what should remain recognizable. Focused instructions make the result easier to compare with your source.
Generate and review the new version
Review the Credit quote, generate the new image, and check the result for unintended changes before you download, edit again, or animate it.
RELATED
Continue from your edited image
AI Image Generator
Create a new image from text or use an optional reference in the broader Ikat Image workspace.
Open AI Image GeneratorGPT Image 2.5
Compare Flare and Sunburst and see the GPT Image 2.5 support currently available in Ikat.
Explore GPT Image 2.5Image to Video
Use the edited result as a starting frame for video generation.
Animate the resultPricing
See how Image and Video use the same Ikat Credit balance.
See Credit optionsFAQ
Image to Image in Ikat
What does Image to Image mean?
Instead of generating only from text, Image to Image uses an existing image together with your prompt to create a new result.
How many reference images can I upload?
Image V1 accepts one reference image per generation.
Which image formats can I use?
The current Ikat Image input supports JPEG, PNG, and WebP.
Will the edit change only the detail I mention?
Not necessarily. Generative editing can also change nearby or related details. Describe what should change and what should stay recognizable, then review the result before using it.
Which GPT Image 2.5 model should I use for editing?
Flare remains the default for faster iteration. Sunburst is available when detailed editing precision is more important than speed.
How much does Image to Image cost?
A successful 1K GPT Image 2.5 generation costs 5 Ikat Credits in Image V1, whether you start from text or one reference image.
Can I animate the edited image?
Yes. After the image is stored in Ikat, you can hand it to Image-to-Video as the starting frame. Video generation has a separate Credit quote.
Edit the image you already have
Upload one reference, describe the change, and generate a new version with GPT Image 2.5.
Edit an image