GPT Image 2.6: images, animated GIFs and SVG vectors

Describe an image, or upload up to 16 references and mark exactly what to change. GPT Image 2.6 adds animated GIF and editable SVG output, mark-up and select-area editing, and a planning control, on top of the quality tiers and formats you know from 2.5.

GPT Image 2.6 image generator

AI Image Generator
0 / 32,000 characters (Han characters count as 2)
Aspect ratio5 common
Image ratio 1:1, best for avatars, product hero shots, and square social posts.
Output format
Background
Resolution
Output: 1024×1024
Quality
Fastest and cheapest — good for drafts and quick tries.
Planning
Plans layout and text most carefully. Same price, a little slower.
What do you want to create?
Describe it in a sentence, or pick an example below to fill it in

What GPT Image 2.6 adds

Everything from GPT Image 2.5 is still here: text-to-image, up to 16 references, 13 aspect ratios, up to 4K and transparent PNG/WebP. 2.6 adds four new ways to shape the result.

Animated GIF

Pick GIF as the output format to get a looping animation of 4 to 16 frames at 512×512, each shown for 100 to 500 ms. GIF is text-to-image only and has no transparency.

Editable SVG vectors

Pick SVG to get a vector file you can open and edit in Figma, Illustrator or a code editor. The image is drawn at 1K, then rewritten as vector shapes; scripts and external links are stripped before you download it.

Mark up a reference

In image-to-image, draw arrows, circles or notes on the first reference image. The model reads the marks as instructions, so you can circle a cup and write "replace the circled cup with a glass" in the prompt.

Select an area to redraw

Paint over the part of the first reference image you want changed. The model redraws that area and keeps the rest of the picture as it is.

Planning: Quick, Balanced or Thorough

Set how much the model plans the layout and text before it draws. The price is the same at every level: Quick is fastest, Thorough is best for posters and text-heavy layouts.

5 quality tiers, up to 4K

Standard, Enhanced, Fine, Ultra and Extreme at 1K, 2K or 4K across 13 aspect ratios, saved as PNG, JPEG or WebP, with transparent backgrounds for PNG and WebP.

Where the new features help

Each 2.6 feature solves a specific job. Pick the one that matches what you need to deliver.

Animated stickers and reactions

Make short looping GIFs for chats, social posts and slides: a waving mascot, a blinking logo, steam rising from a cup. Simple motion on a plain background loops most smoothly.

Logos, icons and flat illustrations

Generate SVG icons, badges and flat illustrations that stay sharp at any size and can be recolored or tweaked in your design tool instead of redrawn.

Precise photo and product edits

Select the area to change, such as a product on a table or a sign in the background, and describe the replacement. Everything outside the selection keeps its original look.

Posters and layouts with text

Set Planning to Thorough for posters, menus and covers where headline placement and text accuracy matter most. Use Quick for simple scenes and fast drafts.

Flare, Sunburst, or GPT Image 2.5?

01

GPT Image 2.6 Flare

Flare is selected by default on this page. Start with it and check the first result against your composition and detail requirements.
02

GPT Image 2.6 Sunburst

Switch to Sunburst in the model menu; it costs the same as Flare. To compare, keep the prompt, references and settings the same and see which suits your task better.
03

When to use 2.5 instead

2.6 Standard and Enhanced cost more than on 2.5 (4 credits vs 2 at Standard 1K); Fine, Ultra and Extreme cost the same. Use 2.5 for low-cost still-image drafts and 2.6 when you need GIF, SVG, area edits or planning control.

Create with GPT Image 2.6 in three steps

01

Describe or mark up

Write what you want. For edits, upload references, then use Mark up / select area on the first image and explain in the prompt what the marks or the selected area mean.
02

Choose the output

Pick PNG, JPEG or WebP for a still image, GIF for an animation or SVG for a vector, then the quality tier and planning level. The generator shows the credit cost as you go.
03

Review and download

Check the result, then download it or keep refining. Still images can go back in as a reference; history keeps every result with the model and settings used.

Prompt examples for the new features

01

Animated GIF

Try with GIF, 8 frames, 150 ms: A small paper boat bobbing on gentle blue waves, flat illustration. The boat rocks left and right in a seamless loop on a plain light background.
02

SVG icon

Try with SVG: A minimal line icon of a coffee cup with rising steam, thick rounded strokes in dark navy and orange, centered, no background, no text.
03

Select-area edit

Try in image-to-image after selecting the mug on the table: Replace the selected mug with a clear glass of iced tea with a lemon slice. Match the existing light and shadows.

GPT Image 2.6 questions

Four things on top of GPT Image 2.5: animated GIF output, editable SVG output, mark-up and select-area editing on the first reference image, and a planning control (Quick, Balanced or Thorough). Text-to-image, up to 16 references, 5 quality tiers, 13 aspect ratios, up to 4K and transparent PNG/WebP all carry over.
Text-to-image is priced by quality and resolution at 1K/2K/4K: Standard 4/9/15, Enhanced 8/15/25, Fine 10/18/30, Ultra 16/32/52 and Extreme 35/70/115. Flare and Sunburst cost the same. Image-to-image starts at the selected tier's cost and settles based on the actual generation cost. The generator shows the credits before you submit.
Choose GIF in Output format, then the number of frames (4 to 16) and how long each frame shows (100 to 500 ms). Frames are 512×512 and the animation loops. Up to 8 frames costs 4/6/7 credits at Standard/Enhanced/Fine, 12 frames costs 6/10/12 and 16 frames costs 9/15/18. GIF is text-to-image only.
The model draws the image at 1K, then rewrites it as vector shapes. That second step is billed at its actual cost, so the price shown (10/14/16 credits at Standard/Enhanced/Fine) is the minimum, and you need a balance of at least twice that to start. SVG is text-to-image only. For safety, scripts, links and embedded images are removed from the file.
Mark up lets you draw arrows, circles or notes that the model reads as instructions. Select area lets you paint the part to redraw while the rest stays as it is. Both work in image-to-image on the first reference image, one at a time. Describe in the prompt what the marks or the selected area should become.
It sets how much the model plans composition and text before drawing. Quick is fastest and suits simple edits and GIFs, Balanced is a middle ground, and Thorough plans layout and text most carefully. All three cost the same.
Yes. The Standard tier is open to every account, including trial credits. Tiers above Standard unlock once your account has bought credits. SVG needs a balance of at least twice its minimum price.
Yes, for still images: set Background to Transparent and save as PNG or WebP. The Background setting does not apply to GIF and SVG; GIF has no transparency.
Download from the results area after generation, or reopen any result from history, which also records the model and settings used. Still images can be reused as references for further edits.

Still have questions? Email support@image-2.net.

Try GPT Image 2.6

Start with a still image at Standard, then try a GIF, an SVG or a select-area edit when you need it.

Start creating