AI Image & DesignChatGPTOpenAIGIFimage generationFlareSunburst

ChatGPT Turns Any Photo Into a GIF Now: Images 2.6 Is 50% Faster, With a $8 vs $30 Model Choice

OpenAI released Images 2.6 on 4 November 2026, adding animated GIF creation to ChatGPT, Codex and the API, with image generation OpenAI says is 50% faster than before. GIFs work on every ChatGPT plan, and developers get two new API models, Flare for speed and Sunburst for quality.

Published November 6, 2026

Muhammad Usman

By Muhammad Usman · Founder & Lead Reviewer

ChatGPT Turns Any Photo Into a GIF Now: Images 2.6 Is 50% Faster, With a $8 vs $30 Model Choice

Quick Answer

OpenAI released Images 2.6 on 4 November 2026, and it adds animated GIF creation to ChatGPT, Codex and the API, with image generation OpenAI says is 50% faster than before. Anyone on any ChatGPT plan can turn a prompt or a photo into a GIF, and developers get two new API models, Flare for speed and Sunburst for quality, priced at $8 per million image input tokens and $30 per million image output tokens.

Key takeaways

  • Images 2.6 launched on 4 November 2026 in ChatGPT, Codex and the API, on every subscription tier, on web, desktop and mobile.
  • ChatGPT can now turn a text prompt or an uploaded photo into an animated GIF, change its playback speed, and edit individual frames.
  • Reference images let you save yourself, a pet or a character once, then reuse that look without uploading the photo again.
  • OpenAI says generation is 50% faster than before. Its own Flare benchmark shows generation time falling from 7.08 seconds to 4.84 seconds.
  • Developers get two API models: GPT-Image-2.6 Flare for speed, and GPT-Image-2.6 Sunburst for quality and consistency across edits.
  • API pricing is $8 per million image input tokens and $30 per million image output tokens, with cached image input at $2 per million tokens.
  • GIF creation has no separate paywall in ChatGPT, but existing account level usage caps still apply.

What is Images 2.6, in plain terms?

Images 2.6 is OpenAI's new image generation release, and it is the first version of ChatGPT's image tool that makes motion, not just still pictures. Until now, if you wanted a GIF from ChatGPT, you made a still image, then took it somewhere else to animate it. Now the animation happens inside ChatGPT itself. You describe a scene or upload a photo, ask for a GIF, and you get a short looping animation you can download and share.

The release has three parts that matter to different readers. For everyday ChatGPT users, the headline is GIFs plus reference images. For developers, the headline is two new API models, Flare and Sunburst, with a speed claim and a published price. For everyone, the headline is that image generation is now faster, which matters most when you are iterating on a design and waiting on every generation.

Scale is worth noting, because it explains why OpenAI is investing here. According to the company's release announcement, ChatGPT and the API now generate more than 4 billion images each week, up from 3 billion in September. That is a lot of images, and a lot of them are social posts, product shots, memes and reaction images, exactly the formats where a short GIF beats a still picture.

How do you turn a photo into a GIF in ChatGPT?

The flow is short, and it works from a prompt or from an image you already have. In ChatGPT, start a new image request and describe the animation you want, or upload a photo and tell ChatGPT what should move. For example, upload a product photo and ask for a slow spin, or upload a pet photo and ask for a head tilt and a blink. ChatGPT returns an animated GIF, and you can ask it to change the playback speed or edit individual frames before you share it.

The GIF controls are practical, not decorative. A GIF is 2 to 64 frames, with each frame lasting 20 to 2,000 milliseconds. The default is 9 frames at 200 milliseconds each, which makes a 1.8 second loop. Short loops work best for reaction images and social replies. Longer loops, with more frames, work better for showing a product feature or a short storyboard beat. If your GIF looks choppy, ask for more frames. If it feels slow, ask for shorter frame durations.

Two practical tips make the results much better. First, describe the motion, not just the scene. "A coffee cup steaming on a desk" gives you a still image with a request attached. "A coffee cup on a desk, steam rising and curling upward, looped smoothly" gives the model a motion to render. Second, keep the moving part small. A GIF where one element moves and the rest stays still looks clean and professional. A GIF where everything moves at once usually looks messy, because the model has to keep every part consistent across frames.

What are reference images, and why do they matter?

Reference images are the quieter feature in this release, and for many people they are the more useful one. You can save a reference image of yourself, a pet, a character or a product, and ChatGPT keeps it ready to reuse. The next time you generate an image, you do not upload the photo again. You just refer to it, and the model keeps the same face, the same pet, or the same product look across new images.

This solves a real annoyance. Before reference images, every new generation started from scratch, and keeping a character consistent across ten images meant re-uploading the same photo ten times and hoping the model remembered the details. With a saved reference, consistency becomes the default instead of a struggle. That matters for a small business with a mascot, a creator with a recurring character, or anyone making a series of images about the same person or pet.

There is a limit worth knowing. Reference images help with consistency, they do not guarantee it. Faces, hands and small product details can still drift across generations, especially at low quality settings. If a detail matters, check it in the output before you publish, and use Sunburst, the quality focused model, when consistency is the priority.

Flare vs Sunburst: which model should you pick?

Developers get two models, and the choice is a real trade off, not a good versus bad ranking. Flare is built for speed. At low quality it can generate an image in as little as 6 seconds, and OpenAI's benchmark shows Flare generation time falling from 7.08 seconds on Images 2.5 to 4.84 seconds on Images 2.6. Sunburst is built for quality and consistency. It produces more detailed images and holds a character or product steadier across several rounds of edits, at medium or high quality settings.

| Model | Built for | Best use | Speed claim | | GPT-Image-2.6 Flare | Generation speed | Interactive apps, high volume requests, latency sensitive interfaces, quick drafts | As little as 6 seconds at low quality; 7.08s to 4.84s in OpenAI's benchmark | | GPT-Image-2.6 Sunburst | Quality and edit consistency | Product imagery, character studies, posters, assets that need several revisions | Slower than Flare, more detail and steadier edits at medium or high quality |

The practical rule is simple. If you are generating many images and speed matters more than perfection, use Flare. If you are generating one image that has to look right, or you are editing the same image for the third time and the character keeps drifting, use Sunburst. Most apps will end up using both: Flare for the first draft and the fast preview, Sunburst for the final asset.

The release also adds a complexity setting, SVG output, and visual markup tools for editing specific areas of an image while keeping the rest unchanged. SVG output matters for logos and icons, because a vector file scales cleanly to any size, unlike a raster image that blurs when enlarged.

What does Images 2.6 cost on the API?

API pricing is token based, not per image, so the cost of one image depends on its size and quality setting. OpenAI's published pricing lists both new models at $8 per million image input tokens and $30 per million image output tokens, with cached image input at $2 per million tokens. Text input is $5 per million tokens, or $1.25 for cached text input.

That structure has two consequences worth understanding. First, a low quality, small image costs much less than a high quality, large one, because it uses fewer tokens. Second, editing an existing image costs less than generating from scratch when the input image is cached, because cached image input is $2 per million tokens instead of $8. If you are building an app that edits the same product photo repeatedly, caching is where the savings live.

Do not trust any article that quotes a fixed price per image for Images 2.6 without showing the size and quality setting, because the token count changes with both. The honest way to estimate cost is to run your own typical request, read the token usage in the API response, and multiply by the rates above. For a quick sense of scale, a simple low quality image at the cheapest settings costs a fraction of a cent in tokens, while a large high quality image with many edits costs several cents. Your own usage will tell you the real number within a day.

What are the limits and catches?

Images 2.6 is a real upgrade, but it has limits, and knowing them saves frustration. First, GIFs in ChatGPT have no separate paywall, but existing account level usage caps still apply. Free plan users will hit image limits sooner than Plus or Pro users, and heavy GIF use counts against the same allowance. If you plan a big batch of GIFs, check your remaining allowance first.

Second, the 50% faster claim is OpenAI's, not an independent measurement. OpenAI's own Flare benchmark shows a 32% reduction in generation time, from 7.08 to 4.84 seconds, which is a real improvement but not the same as the headline 50% figure. One independent timing reported by a user showed 6.7 to 8.9 seconds for a 1024 by 1024 image on 2.6, against 12.0 seconds on 2.5 at medium quality. Your speed will vary, and the biggest gains show at low quality settings.

Third, GIFs are short loops, not videos. With a maximum of 64 frames and a default of 9, you get seconds of motion, not a scene. If you need a 30 second video with a story arc, you need a video model, not an image model with GIF output. GIFs are for reactions, product spins, memes, storyboards and short loops, and they are very good at those jobs.

Fourth, animated GIFs are only available through the generations endpoint on the API, not the edits endpoint, according to the API documentation as reported on 5 November. In ChatGPT itself, you can edit GIF frames in chat, but developers building with the API should plan around the generations endpoint for GIF output.

Who gets the most value from this?

Three groups should try Images 2.6 today. Social media managers and creators get a faster way to make reaction GIFs, memes and short product loops without leaving ChatGPT or opening a separate animation tool. Small businesses get a cheap way to animate a product photo or a mascot for a launch post, a story, or an ad creative. Developers get a faster image model with a clear price and a quality focused sibling for the final asset, plus GIF and SVG output that were missing before.

The test to run is small. Take one product photo or one character image, make three GIFs with different motions, and post the best one where your audience already is. If it gets more replies, saves or shares than your still images, GIFs belong in your regular workflow. If it does not, you have spent ten minutes learning something real about your audience, and you have lost nothing.

Internal links: Category hub: https://www.topuseai.com/blog | Related live posts: https://www.topuseai.com/blog/best-free-ai-image-generators-open-source | https://www.topuseai.com/blog/free-ai-image-generation-vs-paid

Sources: OpenAI Developer Community announcement, "Images, upgraded, introducing ChatGPT Images 2.6", 4 November 2026; OpenAI API pricing as reported by Okay News, 5 November 2026; AlphaSignal report on the Images 2.6 announcement, 5 November 2026; Omniakey, "GPT Image 2.6 vs GPT Image 2.5", updated 5 November 2026.

Frequently Asked Questions

Can ChatGPT really turn a photo into a GIF now?

Yes. Images 2.6, released on 4 November 2026, lets ChatGPT turn a text prompt or an uploaded photo into an animated GIF, with playback speed controls and frame editing in chat.

Is GIF creation free in ChatGPT?

There is no separate paywall for GIFs, but existing account level image usage caps still apply, so free plan users will hit limits sooner than paid plan users.

What is the difference between Flare and Sunburst?

Flare is the fast API model, as little as 6 seconds per image at low quality. Sunburst is the quality model, with more detail and better consistency across edits at medium or high quality.

How much does Images 2.6 cost on the API?

$8 per million image input tokens and $30 per million image output tokens, with cached image input at $2 per million. The cost per image varies with size and quality setting.

How long can a ChatGPT GIF be?

GIFs run from 2 to 64 frames, with each frame lasting 20 to 2,000 milliseconds. The default is 9 frames at 200 milliseconds, a 1.8 second loop.

What are reference images in Images 2.6?

Saved images of a person, pet, character or product that ChatGPT reuses across new generations without re-uploading, so the same look stays consistent across many images.