GPT Image 2.6: Stills, GIFs and Edits That Hold

OpenAI's newest image model, on VidCella without a ChatGPT plan. It turns a sentence or a photo into a looping GIF, lets you decide how much detail to pay for, and keeps a face or a product recognizable through repeated edits. A 1K square starts at 7 credits.

Highlights

What GPT Image 2.6 Adds

Two things GPT Image 2.5 could not do, and one it now does better. All three examples were generated with the model on this page.

A GIF From One Sentence

Set Output to Animated GIF and describe the motion. This prompt asked for a red paper boat rocking on a calm pond. The model returned a 16-frame loop where the water and reflections move, and the boat, rocks and lily look the same in every frame.

Animate a Picture You Already Have

Upload a still and say what should move. Here the steam curls upward and the light shifts. The cup, its handwritten label and the framing are the ones from the original image.

Small Print That Comes Back Intact

Every line on this poster was typed into the prompt in quotes: the headline, the date line and three lines of small print at the bottom. All five came back spelled as written.

Three Edits Deep, Still the Same Person

Each image was made by editing the one before it, on Sunburst. The coat changes color, a dog shows up, a lamp post goes in. The face, the pier and the fog carry through all three.

Start: yellow raincoat on a foggy pier

Edit 1: the coat is now red

Edit 2: a black dog sits beside her

Edit 3: a lit lamp post on the right

What Is GPT Image 2.6?

GPT Image 2.6 is OpenAI's follow-up to GPT Image 2.5. One still on its own looks only a little better than before. What changed is how much you can ask for: motion, a quality setting you control, and more shapes to work in.

Two Versions, One Price
Flare is the fast default. Sunburst takes longer per image and drifts less when you keep editing the same subject. They cost the same, so choose by the job.
Five Quality Levels
Low, Medium, High, X-High and Max. Draft on Low and move up only for the versions you keep.
Price Follows Quality, Size and Shape
Wide and tall images cost less than a square at the same settings. The Generate button shows the exact credits before you click.

The Same Prompt at Four Quality Levels

A macro shot of a vintage wristwatch at 1K square, run once per setting. They are separate generations, so the framing differs. The dial lettering is readable even at Low. The higher settings pay off when leather grain and scratches have to hold up at full size.

Low: 7 credits

Medium: 12 credits

High: 15 credits

Max: 57 credits

GPT Image 2.6 Pricing in Credits

Each cell is square / 16:9. Flare and Sunburst cost the same, Auto ratio is charged as a square, and each reference image adds 4 credits.

Quality	1K	2K	4K	GIF
Low	7 / 5	14 / 6	23 / 13	14
Medium	12 / 8	24 / 10	40 / 24	24
High	15 / 9	29 / 12	49 / 28	29
X-High	26 / 16	52 / 21	86 / 49	n/a
Max	57 / 35	116 / 46	193 / 109	n/a

How to Use GPT Image 2.6 on VidCella

One form covers text to image, editing and GIFs:

GPT Image 2.6 Features on VidCella

What the generator on this page supports:

Animated GIF Output

512×512 loops of 8 or 16 frames, with frame timing adjustable from 20 to 2000 ms.

Up to 16 Reference Images

Edit, combine or animate your own images. Each reference adds 4 credits.

15 Aspect Ratios

From 1:1 to 21:9 and 3:1 banners, in portrait or landscape, plus Auto.

1K, 2K and 4K Stills

Three output sizes, each available at all five quality levels.

Transparent Backgrounds

Set Background to Transparent at 1K and the image comes back as a cutout.

Pay Per Image

Credits only. No ChatGPT Plus, no subscription, no waitlist.

FAQ

FAQs about GPT Image 2.6

Common questions about GPT Image 2.6 on VidCella

What is new in GPT Image 2.6 compared with GPT Image 2.5?

GIF output, a quality setting with five levels, 15 aspect ratios instead of 12, and steadier results when you edit the same subject repeatedly. A single still is only a modest step up, so the reason to switch is one of those features.

How do I make a GIF with GPT Image 2.6?

Set Output to Animated GIF, describe the motion, and choose 8 or 16 frames and how long each frame stays on screen. GIFs come out at 512×512 and loop. You can also upload an image and animate it. A GIF costs 14, 24 or 29 credits at Low, Medium or High quality.

How much does GPT Image 2.6 cost?

It depends on quality, size and shape. At Low quality a 1K square is 7 credits and a 1K 16:9 image is 5. At Medium those are 12 and 8. Each reference image adds 4 credits. The table on this page lists every quality level, and the Generate button shows the exact price before you submit.

Why does the aspect ratio change the price?

The model is billed by how much it has to draw, and a wide or tall image is less work than a square at the same quality and resolution. Auto can come back in any shape, so it is charged as a square.

Flare or Sunburst?

Flare for everyday generation and quick drafts. Sunburst when you are editing the same face or product over several rounds and need it to stay put. Same price either way.

Should I use GPT Image 2.6 or GPT Image 2.5?

GPT Image 2.5 is still on VidCella at a flat 5, 15 or 24 credits for 1K, 2K or 4K. Use it when you want a still at a fixed price. Use 2.6 when you need a GIF, a wide banner, or control over the quality level.

Can it generate a transparent background?

Yes, at 1K. Set Background to Transparent and you get a cutout. At 2K and 4K the model decides the background itself.

Why was my prompt declined?

OpenAI filters this model strictly. Prompts that name copyrighted characters, public figures or brands are refused. Original characters and designs go through, and a declined request does not use credits.

Do I need ChatGPT Plus?

No. VidCella runs on pay-as-you-go credits, so there is no OpenAI subscription and no VidCella subscription either.

Can I use the images commercially?

Yes. Handle real-person likeness, trademarks and third-party IP responsibly under applicable law and our terms.