ChatGPT's new Images 2.6 model can make GIFs, the short looping animations most people know from message threads and reaction posts.
You can generate one from a text prompt alone, or from a text prompt combined with reference images. After ChatGPT produces the GIF, you can modify its frames and adjust its speed before sharing it with others.
The notable shift is the format. Image generators inside chatbots have mostly returned stills. A GIF is a modest step toward moving images that stays clear of full video generation, which is a separate and more demanding product category.
The controls may matter more than the animation itself. A generated clip is only useful if it ends up the way you wanted, and putting frame changes and speed adjustment inside ChatGPT keeps the whole sequence of prompt, tweak and share in one place. The reference-image option points the same way: you can start from something you already have instead of describing it from scratch.
GIFs are built to be passed around, which makes them a natural test of how people use AI imagery. The open question is whether they become a quick-reply habit or a one-off novelty.