ChatGPT Images 2.6 brought animated GIFs into the chat
On November 4, 2026, OpenAI began rolling out GIF creation in ChatGPT. The release notes describe asking for a looping animation or starting with an uploaded sketch, then adjusting playback speed and sharing the GIF. The separate announcement says the Images 2.6 model was available globally across ChatGPT and Codex subscription tiers on desktop, mobile and web. The GIF bullet uses rollout language, so model availability alone establishes little about when the animation control reached an individual account. ChatGPT release notes · Images 2.6 announcement
From a still image to a shareable loop
- A pet-photo revision on November 5. TechRadar’s Graham Barlow attached a photo of his cat Chester and asked for a Halloween GIF to share with family. The first loop added a hat and cape, slow blinking and a flickering pumpkin. He followed up asking for more movement and gave ChatGPT freedom to improvise. The second GIF added a raised paw and a floating pumpkin, and Barlow found the movement more convincing. The article links both GIFs and reports share and download options. It also shows a later lunar-chase GIF made after saving the original photo as a reference. TechRadar’s first-person test
- A character nod on November 5. RumbleversePlayer supplied a ChatGPT-generated anime-style character and specified an upward nod, a quicker downward nod, and a return to the starting pose with the eyes closing. The published result came on the third attempt. The Reddit gallery carries a GIF and the source PNG. The author attributed the output to ChatGPT’s built-in image generator in a reply. The first two attempts are unavailable. Original post and prompt · Submission metadata
- A mobile sketch on November 11. Strange-Call5277 drew a mushroom in the app, asked for a finished still, then requested a loop in the same conversation. The author reports a blinking and smiling loop, a visible 1× speed control and a completed download. The same test appeared twice, once as a text walkthrough and once with the media, so it represents one operator. Original walkthrough · Same test with media
- Seven photo prompts on November 12. No-Intention3844 posted a gallery pairing source photos with loops. The author found that one small motion and an explicit return pose worked better for these short loops. The pet came back framed closer, and two spin attempts jumped until the prompt gave a more specific per-frame turn. The reported nine frames at 200 ms describe this test. Paired photo and GIF gallery
What counts as native GIF creation
A GIF displayed inside ChatGPT with playback controls and a download option is an observable product workflow. A still sprite sheet has to be split into frames and assembled before it plays. For example, a September Images 2.5 operator generated a 4×4 sheet through Atlas Cloud and then used Codex to assemble the GIF. OpenAI’s November launch also showcases an OpenArt sprite-sheet prompt and a Town API experiment. Those examples show adjacent animation routes. The same announcement covers SVG export and the addition of image markup to the API under separate API controls. September sprite-sheet workflow · OpenAI’s GIF and API sections
The press and Reddit examples give dated instances of ChatGPT delivering animated output and show where a retry helped. They provide no per-plan access schedule, serving-variant identification or repeatable measure of motion quality. In the October 28 through November 18 R1 window, launch summaries and the mobile cross-post add context without adding independent tests. ChatGPT release notes · Mobile operator’s availability note