ChatGPT can now make GIFs and save your photos for later chats

OpenAI introduced Images 2.6 on November 4. The new image model is available in ChatGPT, in the Codex coding tool and through the API. OpenAI says it is available worldwide on all subscription plans across desktop, mobile and web. The company says images now generate 50% faster than before. Edits should also be more consistent, so clothing, faces and product details stay the same when the background or another part of the image changes. Two new features go beyond still images.

Prompts and photos can become looping GIFs

ChatGPT can now create animated GIFs. Users describe what should move or upload a still image and ask for an animation. The OpenAI help article on ChatGPT Images says the playback speed can be adjusted afterwards, and the GIF can be saved or shared. Changes are requested in the chat. ChatGPT cannot edit an animated GIF that a user uploads. As of November 7, the help article still describes GIF creation as rolling out, so it may not appear in every account yet. On November 8, the ChatGPT Images page in Germany already showed an anime reaction GIF template without signing in. OpenAI has not published GIF limits for any plan.

Saved photos can be reused by name

The second feature concerns personal photos. Users upload a picture, give it a name and call it up in later chats with an @ mention. OpenAI's example is a dog photo saved as Pepper. A later prompt can then ask for a watercolor portrait of @Pepper "without uploading the photo again." ChatGPT may also offer to save a photo on its own when a user describes it. The photo stays stored in the account after that. OpenAI's help page does not say where saved references can be managed or how to delete them. Generated images follow a clearer rule. They are automatically saved under Images, and they are only removed when the conversation that created them is deleted.

Saving photos of real people raises another issue. Trending Topics points out that a saved reference photo combined with animation makes it easier to place people in scenes that never happened. OpenAI has not said how it guards against misuse of GIFs showing real people. Its help article on provenance signals says images generated with ChatGPT include both Content Credentials and an invisible SynthID watermark. GIFs are not listed separately. Anthropic now marks AI output in a similar way, as our report on watermarks in Claude text explains.

Photos of children or other people should only be saved with their consent. Uploads are part of your chats, and whether OpenAI may use them for training is controlled in Data Controls, as our guide to opting out of AI training shows. Our AI subscription comparison lists what the ChatGPT, Gemini and Claude plans cost.