SVG
Vector output
GPT Image 2.6 supports SVG output. Vector shapes can stay sharp at different sizes, making the format useful for icons, illustrations and design assets. Review the exported file before using it in a final design.
AI image generator
A new chapter for image creation
GPT Image 2.6
Bring your ideas to life with GPT Image 2.6. Create images from text, shape scalable SVG artwork, explore animated GIFs and refine your visuals with guided edits.
Three ways to shape an idea
Scalable shapesLooping motionFocused changes
What is GPT Image 2.6?
GPT Image 2.6 is OpenAI’s image generation and editing model family. It works with text and reference images, with SVG artwork and animated GIFs extending the creative possibilities beyond a single raster image. Start with a clear brief, then refine the composition, colors or specific details to suit your project.
The family includes Flare for fast creative iteration and Sunburst for detailed visual work. Both support generation and editing. The useful distinction is your creative goal: exploring a direction quickly or working through a composition where small details matter.
Beyond a single still image
Give each idea the right format: an image for a campaign, vector artwork for a design or a looping animation for a little extra expression. GPT Image 2.6 brings generation and guided editing into your creative toolkit.
GIF
Animated image output
The model supports animated GIF output for short, looping visuals. Think of motion accents, expressive reactions and small animated illustrations—not a replacement for a full video workflow.
EDIT
Markup-guided edits
Guide an edit with visual markup and a clear description of what should change. Specify what to preserve as well as what to adjust, then check the result for unintended differences.
Flare and Sunburst: two creative priorities
Explore and iterate
GPT Image 2.6 Flare
A starting point for trying visual directions and revising an idea quickly. Use a consistent brief when comparing variations so you can judge the style, subject and composition rather than changing everything at once.
Work through the details
GPT Image 2.6 Sunburst
A starting point for demanding compositions and precise editing tasks. Look closely at the details that determine whether a result works: text placement, object relationships and the parts of a reference that need to remain recognizable.
SVG, GIF or a still image: choose by purpose
The right format depends on how the artwork will be used. GPT Image 2.6 supports raster images as well as SVG and GIF output; a scalable mark, a looping reaction and a product photograph each call for a different approach.
| Format | Useful for | What to check |
|---|---|---|
| SVG vector artwork | Icons, simple brand marks and illustrations that need resizing or recoloring. | Inspect paths, lettering and small-size legibility in a vector editor. Keep shapes simple when the artwork must work at several sizes. |
| Animated GIF | Short repeating gestures, reaction images and motion accents. | Check the transition back to the starting pose. Keep one main action and a clear camera direction so the loop is easy to follow. |
| PNG, JPEG or WebP | Product scenes, editorial illustrations and other still images with texture or lighting. | Review sharpness at the intended display size. For transparent PNG or WebP assets, check the edges against both light and dark backgrounds. |
Use cases for AI-generated images and animations
An icon that scales
Describe the silhouette, colors and intended use. Keep the concept simple and plan to check small-size readability.
A little movement
Choose one subject and one repeating action. Explain what should stay still and how the loop should feel.
A considered revision
Identify the area to change and the details to keep, such as the subject, layout or surrounding colors.
A product scene with a clear focal point
Plan the surface, lighting and background around one product. Leave deliberate space for a headline rather than asking the image to carry every part of a campaign at once.
How to plan a GPT Image 2.6 creation
A useful image starts with a brief you can evaluate. Work from purpose to composition, then refine one decision at a time instead of stacking unrelated requests into a single prompt.
Define the deliverable
Decide where the result belongs: an app icon, an editorial image, a presentation or a looping animation. Note the layout, intended background and any space needed for text.
Describe the scene
Specify the main subject, its position, the viewpoint and the visual style. If a reference image matters, explain which shapes, colors or identity details should carry through.
Make the change explicit
For an edit, separate the requested change from the details to preserve. For a GIF, describe a starting pose, the action and the return to that pose; say whether the camera stays still.
Review for the final use
Check spelling, proportions, edges and the parts that must match the reference. Test vectors at small sizes and GIFs across several loops before including them in a finished design.
GPT Image 2.6 prompt examples
Use these original creative briefs as starting points and adapt the subject, palette and layout to your project. They illustrate how to describe a result; they are not promises of an identical output.
SVG · Icon design
A simple compass mark
Design a clean vector icon for a weekend hiking journal: a rounded compass with one coral needle and a deep navy outline. Use flat shapes, consistent line weight and generous negative space. No text, gradients or shadows. The mark should be easy to recognize at a small size.
A limited palette and a clear silhouette give you concrete criteria for judging the result.
GIF · Looping animation
A sprout that sways and returns
A small clay-style sprout in a terracotta pot on a warm cream background. The leaves gently lean left, pass through the center, lean right and return to their starting pose. Keep the pot and camera still. Use soft light and make the end join naturally to the beginning.
Describe the return to the starting pose, not just the motion in the middle of the loop.
EDIT · Guided change
A focused color revision
Change only the scarf inside the marked area from red to muted sage green. Keep its woven texture, folds and existing shadows. Preserve the person’s face, hair, coat, pose and background. Do not add objects or lettering, and remove the guide marks from the finished image.
State what changes and what stays. Inspect the surrounding image rather than assuming every other pixel is preserved.
IMAGE · Product concept
A calm desk scene
Create an editorial product image of a matte ivory desk lamp on a pale oak table. Soft morning light comes from the left; a muted blue wall fills the background. Place the lamp on the right with open space on the left for a headline. Keep the scene minimal, with no logos or generated text.
Position the subject and reserve text space explicitly instead of relying on a generic request for a professional image.
Get more useful results from generation and edits
Separate layout from surface detail
GPT Image 2.6 distinguishes planning complexity from rendering quality. A dense diagram needs clear relationships and placement; a close-up material study needs convincing texture. When reviewing a result, identify which problem you are trying to solve before changing the brief.
Treat markup as guidance, not a lock
A marked region helps communicate the target of an edit. Also name the subject and explain what must remain unchanged. Compare the entire edited result with the reference, especially faces, product geometry and neighboring objects.
Keep critical text under review
For a poster or presentation graphic, provide the exact wording and describe its hierarchy. Check spelling, line breaks and contrast at the final size. If exact typesetting matters, add approved text in your design tool after selecting the image.
Questions about GPT Image 2.6
What can I create with GPT Image 2.6?
GPT Image 2.6 supports image generation, SVG vector output, animated GIFs and markup-guided image edits. Use it to explore visual concepts, illustrations, icons and short looping animations.
What is SVG output useful for?
SVG describes vector shapes rather than a fixed grid of pixels. It is useful for icons, illustrations and other artwork that needs to stay sharp at different sizes. Check the exported artwork in your design software before using it in a final project.
How do I write a useful image prompt?
Describe the subject, composition, style and colors, then explain the intended use. For an edit, identify the area to change and the details to preserve. For an animation, describe one clear action and how it should loop.
How is an animated GIF different from a video?
An animated GIF is suited to short, repeating visuals such as reactions, motion accents and small animated illustrations. For longer scenes, camera movement or a complete video sequence, explore an image-to-video model.
What is the difference between GPT Image 2.6 Flare and Sunburst?
Flare is oriented toward fast iteration; Sunburst is oriented toward detailed compositions and demanding edits. Both accept text and image inputs. Compare results on the same creative brief rather than assuming one option is best for every task.
Can GPT Image 2.6 create transparent backgrounds?
Yes. The model family supports transparent backgrounds, including transparent PNG or WebP images and SVG artwork. Check edges and any remaining background areas against the surface where the asset will be placed. JPEG is not a transparent format.
Will a marked edit leave everything else exactly unchanged?
Markup directs the model toward the intended change, but it is not a guarantee that every unmarked pixel remains identical. Specify the details to preserve and compare the result with the original. For pixel-identical areas, keep the approved original content in a design editor.
Can I use reference images with GPT Image 2.6?
Yes. The model family accepts text and image inputs for generation and editing. Explain which aspects of a reference matter, such as the silhouette, palette or subject identity, and distinguish those from the elements you want to change.
Continue your creative workflow
Choose the next tool by the result you need: another image model, a video from a still image or a broader view of the model library.