ChatGPT can now produce animated GIFs, adding motion to an image generator that until now only made still pictures. The feature is rolling out with OpenAI’s new Images 2.6 model and works from either a text prompt or an uploaded photo. Users can then adjust playback speed, step through the animation frame by frame, edit individual frames and share the finished loop.
That editing layer is the more interesting part of the update. Generating a short animation from a prompt is impressive, but most AI-generated motion fails on the details: a paw that melts for a moment, a face that changes between frames, a loop that doesn’t quite connect. Being able to correct specific frames without regenerating the whole clip gives users some control over those problems. OpenAI’s demo uses a simple request for a dancing cat, which shows the target audience clearly. This is aimed at reaction GIFs, memes and quick social content rather than professional animation.
Before this update, getting a GIF out of ChatGPT meant creating a series of stills and assembling them in another app. Moving the whole process into one conversation puts OpenAI closer to the kind of everyday, disposable content currently handled by GIF libraries like Giphy and Tenor, which have long been built into messaging apps and keyboards. If people can create a custom reaction GIF as quickly as they can search for one, the habit could change.
OpenAI also says Images 2.6 generates images faster, though it hasn’t published detailed benchmarks. The company’s usage figures help explain why it keeps investing in this area. It reports more than 4 billion images generated each week across ChatGPT and its API, up from the 3 billion it cited in September when Images 2.5 launched. Those numbers come from OpenAI and can’t be independently verified, but a billion-image weekly jump in roughly two months suggests image creation has become one of ChatGPT’s main uses rather than a side feature.
The update is rolling out to all ChatGPT and Codex users on web, desktop and mobile. Developers also get two new API models, GPT-Image-2.6 Flare and Sunburst, which OpenAI positions as options for faster and higher-quality output, respectively.
Several practical questions remain unanswered. OpenAI hasn’t said how long these GIFs can be, what resolutions or frame counts are supported, or whether usage limits differ between free and paid accounts. The photo-to-GIF option also raises a familiar concern. Animating a real person’s photo is much more powerful than animating a cartoon cat, and how OpenAI handles that will matter as much as how smooth the motion looks.