Nano Banana 2.1Generate and edit images from text or references, with natural light and stable subjects.
Describe an idea above to get started.
GPT Image 2.5 is an AI image generation and editing model built for sharper visual quality, stronger reference fidelity and more precise refinement. It renders natural lighting and textures, keeps recognizable subjects consistent across repeated edits and follows complex visual instructions more accurately.
It comes in two versions, Flare and Sunburst, and both handle text-to-image and image editing: Flare is tuned for speed, Sunburst for quality. In one workspace you can switch versions, pick auto or 12 aspect ratios and choose 1K, 2K or 4K output.
From lighting and texture to targeted edits, take more creative control.
GPT Image 2.5 renders light direction, tonal transitions and material detail more naturally, so wood grain, glazed ceramics, fabric and skin feel real. Describe the light source, time of day and materials to get product shots and scenes that look photographed. It suits e-commerce hero images, food photography, interiors and brand mood images.
Upload a person, character or product and the model keeps facial features, outfit colors and product shapes more consistent while placing the same subject in a new scene. Image editing accepts up to 16 references, and mentions such as @image_1 tell the model what each image is for. Use it for character series, model reshoots and one product shown across many settings.
The model is better at targeted edits: swap a background, relight a scene, change one line of text or replace a single object while the rest of the composition, colors and details stay put. Write what should change and what should stay as separate instructions for more predictable results. Bring the edited image back as a reference to keep refining.
Set the resolution to 1K and the workspace shows a background option with transparent, opaque or auto. Choose transparent to create stickers, icons, product cut-outs and logo assets without manual masking. Describe an isolated subject with no backdrop for cleaner edges; the background option is not offered at 2K or 4K.
Flare is the smaller model tuned for speed, with image quality comparable to GPT Image 2. Sunburst is the base model tuned for quality, with finer detail and texture. Both share the same settings, features and credits, and you switch with one click below the prompt. Pick Flare for everyday images and quick drafts, Sunburst for final visuals and detailed edits.
Create stunning AI images in GPT Image 2.5 in three simple steps:
To start from scratch, choose text to image and describe the subject, setting, lighting and style. To change an existing picture, switch to image editing, add up to 16 references and say what to change and what to keep.
Pick Flare or Sunburst below the prompt, then choose an aspect ratio and 1K, 2K or 4K. For a transparent asset, select 1K and set the background to transparent.
Select generate and your image appears in your history, ready to download or reuse as a reference for the next edit. The generate button shows the credit cost before you submit.
GPT Image 2.5 upgrades image quality, reference fidelity and targeted editing. Pick the version that fits the job.
Flare and Sunburst cost the same, so you can run one prompt and set of references on both and keep the better result.
| Version | Resolution | Credits / image |
|---|---|---|
| Flare, Sunburst | 1K | 15 |
| Flare, Sunburst | 2K | 25 |
| Flare, Sunburst | 4K | 40 |
Answers about versions, transparent backgrounds, references and credits.
GPT Image 2.5 is an AI image model that generates and edits pictures. It creates new images from text or references and makes precise changes to existing ones, with more natural lighting and steadier reference subjects. You can use it online in this workspace with nothing to install.
Flare is tuned for speed and Sunburst for quality. Flare is the smaller model with quality comparable to GPT Image 2, while Sunburst is the base model with higher image quality. Both share the same settings, features and credits, so you can compare them on the same prompt.
Yes, at 1K resolution. Set the resolution to 1K and choose transparent in the background option to get an asset you can layer directly; the background option is not available at 2K or 4K.
Image editing accepts up to 16 reference images in JPG, PNG or WEBP, each up to 30 MB. Mention them in your prompt as @image_1, @image_2 and so on to say whether each one supplies the subject, the scene or the style.
You can use auto plus 12 ratios: 1:1, 3:2, 2:3, 4:3, 3:4, 16:9, 9:16, 21:9, 27:16, 16:27, 9:8 and 8:9, at 1K, 2K or 4K. The 27:16, 16:27, 9:8 and 8:9 ratios are 1K only, and the resolution switches to 1K automatically when you pick one.
Each image costs 15 credits at 1K, 25 credits at 2K and 40 credits at 4K, the same for Flare and Sunburst. You can submit up to four tasks at once, and the generate button shows the total.
A prompt can be up to 20,000 characters, enough for a full creative brief. Start with the subject and purpose, then add composition, lighting, materials and style, and put any text that should appear in the image in quotation marks.
Yes. Send the result from your history back to the workspace as a reference, switch to image editing and describe the next change, such as relighting only the background. State what to keep on every round to hold the subject steady.