ChatGPT Images 2.6 adds GIF generation and saved reference photos for people and pets

OpenAI suggests use cases like animating brand mascots, looping product shots, building reaction memes, and sketching out character movement for storyboards, and says a basic doodle is enough to get a GIF started. The other major addition is saved reference images. Users can upload a set of photos of a specific person, pet, or character once, give that set a name, and then pull it back into any future prompt just by mentioning that name, without re-uploading the same photos every time. It’s a shortcut aimed squarely at a common frustration with AI image tools: having to attach the same reference photos over and over just to keep a face, a pet, or a recurring character looking consistent across different generations.

Underneath both features, Images 2.6 also claims a 50% speed improvement over its predecessor, the Images 2.5 model, and better consistency for characters, products, and brand elements across repeated edits. On the API side, OpenAI introduced two new model variants: GPT-Image-2.6 Flare, tuned for speed and able to produce a low-quality image in as little as six seconds, and GPT-Image-2.6 Sunburst, tuned for sharper detail and steadier consistency across edits at medium or high quality. New API tools include image markup for highlighting exactly what to change, SVG export, and a complexity setting that controls how much the model plans before generating, useful for keeping small edits fast rather than triggering a full re-render.

The update lands across ChatGPT, Codex, and the API. OpenAI says people already generate about 4 billion images a week across its platforms, a volume that both of these additions, cheaper small edits and reusable references, seem built to keep growing rather than slow down.