OpenAI has started rolling out Images 2.6, a new image model that reaches ChatGPT, Codex and the company's API. According to The Verge, the update makes image generation 50 percent faster and, for the first time, lets ChatGPT build animated GIFs directly from a text prompt or an uploaded photo.
The feature is available to all ChatGPT and Codex users on web, desktop and mobile, and OpenAI says more than four billion images are now generated every week across ChatGPT and the API, up from the three billion reported in September. The story was first reported by Emma Roth at The Verge and picked up by Techmeme and The Decoder.
How ChatGPT GIF creation works
Users can ask ChatGPT for a simple animation, for example "can you make a gif of a cat dancing?", and the model returns a sequence of slightly different frames stitched together flipbook-style. OpenAI showed an example where the same prompt produced several variations of the character that stay visually consistent from one frame to the next.
Once a GIF exists, ChatGPT can change its playback speed and open it frame by frame for editing. Before Images 2.6, the chatbot could not create a GIF natively in one step. The Decoder notes that earlier models would likely have changed the character's appearance too much between frames for this trick to work, so image consistency is the technical advance behind the feature.
- Create a GIF from a text prompt or an uploaded photo
- Adjust animation speed and edit individual frames
- Share the finished GIF directly from ChatGPT
