GPT Image 2.6, called Images 2.6 in ChatGPT, is OpenAI's model for creating and editing images. Use it right in your browser on Oeneo AI with no API key: draft fast with Flare, then switch to Sunburst for detail that holds across edits.
Flare and Sunburst take the same prompts, reference images and settings, so you can draft with one and finish with the other. The GIF and SVG versions turn a text prompt into a short looping animation or a vector file.
Start here. Switch to Sunburst if edits keep changing parts you meant to keep.
Use it for final assets and multi-step edits.
One GIF per run, made from your description alone.
One SVG per run, sharp at any size on the web or in print.
Start from a prompt, or add reference images and describe the change. Generating and editing happen in the same workbench, on either model.
Text to image
Describe the layout along with the subject: where the headline sits, how big it is and what goes next to it. Put every word you want on the image in quotes and keep headlines short, then zoom in to check the spelling before you publish. Flare is the quicker way to try several layouts.
Product edit
Upload a clean product photo and describe only the new setting. Name what has to stay fixed, such as the shape, label and cap, so the edit changes the scene and leaves the product alone. For a set of campaign shots, run them on Sunburst and compare each one with the original.
Character consistency
Use a finished render or a character sheet as the reference, then describe the new pose and setting. Naming what defines the character, like its colors, eyes and outfit, helps it stay recognizable. Add another reference when one angle doesn't show enough.
Sketch to image
Photograph a pencil sketch and describe the finished style. The sketch sets the layout and poses, and your prompt sets the medium, colors and light. Ask for the paper and ruled lines to be removed so they don't carry into the result.
Choose the GIF or SVG version, then describe what should move or what the mark should look like.
GIF version
Keep it to one simple motion that can repeat, like a lantern being lifted or steam curling up. Short loops like this work for chat reactions, social posts and website banners.
SVG version
Name the shapes, the colors and the exact wording in quotes. The result stays sharp at any size and can be recolored or reshaped in design tools. To trace a logo or photo you already have, use the Image Vectorizer.
Open the model menu in the workbench above and choose Flare, Sunburst, GIF or SVG.
Describe the subject and the layout. With Flare or Sunburst you can add reference images to edit; GIF and SVG start from text.
The Generate button shows the credits for your settings. Generate, then download the result you want to keep.
Both generations come as Flare and Sunburst and run in the same workbench here. Credits for each are listed in the specs below.
| Compare | GPT Image 2.6 Flare | GPT Image 2.6 Sunburst | GPT Image 2.5 Flare | GPT Image 2.5 Sunburst |
|---|---|---|---|---|
| Released | November 4, 2026 | November 4, 2026 | September 8, 2026 | September 8, 2026 |
| Quality | low · medium · high | low · medium · high | low · medium · high · xhigh · max | low · medium · high · xhigh · max |
| Output formats | PNG, JPEG, WEBPPlus GIF and SVG versions | PNG, JPEG, WEBP | PNG, JPEG, WEBP | PNG, JPEG, WEBP |
| OpenAI's 2.6 focus | Poster and product layouts | Steadier detail across edits | — | — |
| Where to use it | Workbench on this page | Workbench on this page | Open model page | Open model page |
Paste a prompt into the workbench and swap in your own subject. With Flare and Sunburst, test at low quality and switch to high for the final image.
GPT Image 2.6 Flare
Prompt length 389
Magazine ad layout for a sparkling grapefruit soda. A pink can bursts out of a splash of soda with grapefruit slices flying around it, filling the right half of the frame. On the left, a large white headline reads "PINK FIZZ" with one small line under it that reads "ZERO SUGAR". Flat grapefruit-pink background, bright studio light, crisp condensation on the can. No other text, no logos.The job and the layout
Say what the image is for, such as a poster, a product shot or a story page, and where the main elements go.
Exact words in quotes
Quote every word that should appear and say where it goes. Short headlines come out cleaner than long lines of small text.
What changes and what stays
For edits, name the one change and list what must not move: the face, the product shape, the label, the background.
Several changes in one edit
Make one change per run and build on the result. Stacked requests are harder to check and more likely to drift.
Vague style words
Words like "modern" or "nice" leave the model guessing. Name the medium, the main colors and the light instead.
Unlabeled reference images
With several references, say what each one is for, such as "the bottle from the first image" or "the colors from the second".
Vague
A cool summer movie poster.
Specific
Summer poster for a rooftop movie night. Large headline "MOVIES UNDER THE SKY" at the top, a couple on a picnic blanket watching a glowing screen below, coral and teal palette, Swiss-style grid, no other text.
The second version names the job, the layout, the exact headline and what to leave out.
Each model has its own page and runs in the same workbench on one credit balance.
OpenAIFaster everyday generation, natural lighting and rich textures
View model
OpenAIPrecision-first detail, premium visual fidelity
View model
OpenAIReadable text, complex briefs, precise edits
View model
New
GoogleConsistent characters, product placement, exact short text
View modelRemove the background, upscale, vectorize or denoise any result, all on the same credit balance.
Transparent PNG or WebP with clean hair and product edges, ready for a listing or a new composition.
2× / 4× / 8× with recovered texture instead of smeared pixels, so a 1K draft can become a print.
Logos, icons and flat illustrations become editable SVG that scales to any size.
Remove grain and compression artifacts from low-light photos and scans while keeping real detail.
GPT Image 2.6 is OpenAI's image generation and editing model, released on November 4, 2026. ChatGPT calls it Images 2.6; for developers it comes as two models, Flare for speed and Sunburst for detail. You can use both on this page.
It runs on credits, and new accounts get 10 credits when they sign up. The price depends on the quality, resolution and number of reference images you pick, and the Generate button shows it before you start.
At low and medium quality, yes: the same settings use more credits on 2.6 than on 2.5. At high quality the two cost the same. The specs above list the exact credits for both.
No. Sign in to Oeneo AI and use GPT Image 2.6 in the workbench at the top of this page. There is no API key, code or OpenAI account to set up.
Yes. Choose the GPT Image 2.6 GIF or SVG version from the model menu in the workbench and describe what you want. Both work from text alone: you can't upload a photo, so turning a photo into a GIF is not supported, and each run makes one file, either a short looping GIF or a vector SVG. To animate a photo, use the AI Video Generator; to convert an existing image into a vector, use the Image Vectorizer.
No. Generate the image first, then remove the background with the Background Remover.
Yes, by describing it; there is no mask or markup canvas here. Upload the image, name the one change and list what must stay the same, then check the whole image after each round.
Yes. You can generate at 1K, 2K, 4K. OpenAI marks sizes above 2560×1440 as experimental, so zoom in on faces, text and edges before you use a 4K result.
Results from a free account are public. After you buy any plan or credit pack, new results are private by default.
Yes. Oeneo AI's generated content license lets you use your results in personal and commercial projects without attribution. OpenAI's usage rules for the model still apply.
Turn a prompt or a reference image into a finished poster, product shot or illustration with Flare or Sunburst.