AI image editing models and APIs.
Edit an existing photo with a plain-language instruction: add or remove objects, change the background, fix lighting or retouch a portrait.
How to pick an image editing model
These models change an existing photo from a plain instruction. They differ in how closely they follow it, whether you can mask an area, and how many reference images they accept.
Edits from one instruction
Nano Banana 2.1 handles most edits from a single sentence, including text on signs and packaging. FLUX Kontext Pro gives consistent results when you edit the same image several times, and FLUX Kontext Max is the stronger version for typography.
Change only a marked area
GPT Image 2.5 Sunburst and GPT Image 2.5 Flare accept an optional mask, so only the area you mark is changed. Sunburst can also return a transparent background.
Many references
Nano Banana 2.1 Advanced takes up to 14 reference images and can ground the edit with Web or Image Search, which helps with real products and places.
Retail and fashion
Virtual Try-On V2 puts your garments on a model, Product Photoshoot restages a product in a new scene, and Camera Angle Editor shows it from another angle.
Quick previews
Nano Banana 2 Lite edits at 1K and is the quickest of the Google image models here. Use it to try an idea before you run a bigger model.
Frequently asked questions
What is the difference between image editing and image to image?
Image editing is part of image to image. An editing model keeps the original photo and changes only what you ask. Image to image also covers restyling and generating new scenes from a reference. See all image to image models.
Can I edit only part of an image?
Yes. GPT Image 2.5 Sunburst and GPT Image 2.5 Flare accept a mask that marks the area to change. Models without a mask work from the instruction alone, so describe the area clearly.
Which model is best for changing text in an image?
Nano Banana 2.1, GPT Image 2 and FLUX Kontext Max all handle text well. Put the exact wording in quotes in your prompt.
How do I try an edit before writing code?
Open the model page, upload your photo in the playground and run it. When you like the result, the API tab gives you the same request as code.