ChatGPT Images 2.6
ChatGPT Images 2.6 is OpenAI’s image generation and editing release, announced November 4, 2026. Its Flare and Sunburst API models offer two starting points for creative work: rapid iteration or detailed visual production. The release adds markup-guided edits, vector SVG export and animated GIFs alongside still-image creation.
Create stills for a launch, a social post or a visual concept. A short brief can define the people, setting and lettering, as in this streetwear example. Flare is positioned for quick iteration; Sunburst is the option to test for more demanding detail. Proofread every headline before delivery.
Markup adds a visual instruction to an image edit. The official example marks a vase, then changes its surface to blue. Use a circle or arrow to make the target clear, and describe the change in words. Inspect the rest of the room after the edit rather than assuming perfect preservation.
Give a character one readable movement and a starting pose it can return to. The official lantern example demonstrates a compact, repeating gesture. GIF output belongs to the Image API generation route, rather than the edit route. Plan a short loop; use a video workflow when you need a longer sequence or sound.
A transparent product image can sit over a campaign background or a presentation layout. Request the cutout and choose a format that retains transparency. This official shampoo example shows the intended isolated subject. Check fine edges and lettering against both light and dark surfaces before using the asset.
For a mark or illustration that needs to scale, the generation route can return SVG. This bakery example is an actual vector file, rather than a raster picture styled to look like one. SVG export has its own request restrictions. Inspect its shapes and lettering before adapting it for an application or print layout.
Complexity controls planning; quality controls rendering detail. The two football examples use the same prompt but arrange the scene differently. Low, medium and high complexity are documented, with high as the default. Start with a focused edit, compare settings on the same task and keep the version that meets your brief.
Source: OpenAI official GPT Image 2.6 Flare and Sunburst examples.
| DIMENSION | GPT IMAGE 2.6 FLARE | GPT IMAGE 2.6 SUNBURST |
|---|---|---|
| Creative starting point | Fast drafts and interactive iteration | Detailed compositions and careful edits |
| Inputs | Text and images | Text and images |
| Generation and editing | Both supported | Both supported |
| Planning control | Low, medium or high complexity | Low, medium or high complexity |
| Specialized output | GIF and SVG on the generation route | GIF and SVG on the generation route |
| How to choose | Test response time and acceptable detail | Test the details that decide usability |
OpenAI’s best-case six-second Flare example uses low quality. It is not a fixed time for every request. Compare both models with your own prompts and constraints.
Start with a clear product message, then explore composition and headlines. Separate a promising concept from a final asset by checking the wording, proportions and brand details.
Use the approved picture as the next input. Identify one change at a time, compare the entire output with the original and keep a record of the versions you accept.
Develop still frames or a short character loop before planning a larger sequence. A readable gesture or composition can help a team discuss what the finished production should communicate.
Explore Pexo’s video workflows for the next stage of a creative project.
OpenAI announced Images 2.6 on November 4, 2026 for ChatGPT, Codex and the API. The API family includes GPT Image 2.6 Flare and GPT Image 2.6 Sunburst. Both create and edit images from text and image inputs; GIF generation adds short animated loops.
Start with Flare for interactive drafts and frequent changes. Try Sunburst when a composition or edit needs more detail. Compare both on your actual inputs before choosing. OpenAI describes six-second generation for Flare at low quality as a best-case result, not a guaranteed response time.
The Image API accepts a transparent PNG annotation overlay aligned to the first input image. A circle or arrow can identify the intended change. Pair the annotation with a short instruction and inspect the entire result: guidance does not guarantee that everything outside the marked area stays identical.
Yes. The Image API generation endpoint supports GIF and SVG output. Those formats are not supported as output from its edit endpoint. Transparent raster assets use PNG or WebP. SVG is suited to editable graphic shapes; a looping GIF is a different deliverable from a long video with sound.
Review spelling, fine detail, product geometry and changes outside an edited area. Inspect transparent edges on different backgrounds. For animation, check the loop. Output quality and response time vary with the request; review each asset in the format and size you intend to deliver.
