OpenAI has added animated GIF generation to ChatGPT through Images 2.6, a new image model that turns a text prompt, a photo or a rough sketch into a looping animation.

Users can adjust a GIF’s animation speed, view and edit individual frames and share the result, the company said. Until this update, ChatGPT could not natively generate GIFs in a single step.

Images 2.6 is rolling out globally to all ChatGPT and Codex users on web, desktop and mobile. OpenAI said the model generates images 50 per cent faster than its predecessor, Images 2.5. It did not say which workloads account for the gain.

In one example the company shared, ChatGPT answered the prompt “can you make a gif of a cat dancing?” with an animation, plus options to change the playback speed and step through it frame by frame.

The GIF tool can be used to animate a brand mascot, turn a product shot into a loop, make reactions and memes for social media or build animated storyboards to “explore character movement before video production”. It added that “a simple sketch can become an expressive, animated character for videos, games, and more.”

The feature is not a video generator. OpenAI discontinued its AI video app Sora earlier this year. The Wall Street Journal reported that the app had few users and was losing the company about $1 million a day.

Images 2.6 also brings a reference-image feature, which saves likenesses of people, pets and characters so they can be reused without uploading them again.

Two new models are rolling out in OpenAI’s API. GPT-Image-2.6 Flare is the speed option: the company said it can generate images “in as little as six seconds” with quality set to low. GPT-Image-2.6 Sunburst offers higher-quality generation and more consistency across edits at medium or high quality settings, according to OpenAI.

Image markup, already available in ChatGPT, is coming to the API so teams can mark exactly where they want changes made. Developers can also export designs as SVG files for websites, apps and print.

A new complexity parameter sets how much planning the model does for a request. OpenAI recommends low complexity for most editing tasks, and medium or high for scenes with many elements or constraints.

Users now create four billion images a week across ChatGPT, Codex and the API, which averages out to more than half a billion a day. In September the company reported three billion a week.

Last month OpenAI launched Intelligent UI, a ChatGPT feature that generates interactive visual interfaces.