The reaction GIF library is a quiet, decades-old monopoly on shared jokes. OpenAI just pointed a image model at it. Images 2.6 lets ChatGPT users generate a custom GIF from a text prompt, a photo, or even a rough sketch, then edit it before sharing.
OpenAI rolled out Images 2.6 on Nov. 4, 2026, across ChatGPT, Codex and the API. The workflow is simple. Type a prompt, such as a tortoiseshell cat yawning, and ChatGPT produces a looping animation. From there you can change playback speed, describe edits in plain language, or open a “Sheet” view that shows individual frames so you can comment on and mark up specific ones. Then you download or share it.
The model is also faster. OpenAI says generation runs about 50 percent quicker than the previous version, which matters more for GIFs than for stills, since people iterate on a joke until it lands. Reference images are the other useful addition. You can save and tag a person, pet or character, then reuse it in later generations. Consistency has been the weak spot in AI image tools for a while, and Midjourney and Google’s Gemini tools have both been chasing it.
Developers get a choice too. Through the API, you can pick between two image models, Flare and Sunburst, depending on how much you care about speed, detail or consistency. That’s a practical tradeoff, and it suggests OpenAI is treating image generation as a product line with tiers rather than a single model.
Access is broad. Images 2.6 is available to everyone in ChatGPT and Codex, free users included. Your plan only determines how many images you can generate. OpenAI claims users already make more than 500 million images a day, though it didn’t say how many will be GIFs.
So does this matter? For casual users, it’s a fun feature with a short shelf life. Most people pick GIFs from a keyboard search in two seconds, and generating one takes longer. Giphy and Tenor win on speed and shared cultural memory. Nobody’s group chat laughs at a generated cat the way it laughs at Michael Scott.
But the direction is clear. Chatbots are absorbing every small creative task that used to need a separate app, and animation is the next one. Expect Google and others to match it quickly. For marketers and social teams, the frame-level editing and reusable characters are the parts worth testing. Everyone else can treat this as a nice toy, for now.



