OpenAI adds editable GIF creation to ChatGPT with its Images 2.6 release

A photo or text request can now become an animated GIF inside ChatGPT, with controls for playback speed and individual frames. OpenAI released Images 2.6 on November 4, 2026, bringing animation alongside new image-editing tools and two developer models. The release targets both creating a visual and keeping its defining details intact through revisions.

A flipbook inside the conversation

The animation mechanism is straightforward: generate related still images, then string them together like a flipbook. In an OpenAI demonstration described by The Decoder, a request for a dancing cat produces slight variations of the same character, which become the frames of a GIF.

That example depends on the cat remaining recognizable from frame to frame. The Decoder’s assessment is that the model preserves enough visual consistency for the sequence to work. That is an observation about the demonstration, not a measured result showing how reliably it handles other characters or scenes.

The result is not just a finished file to accept or reject. Users can inspect the sequence frame by frame, change playback speed and edit a single frame before sharing it. Those controls put both the pace of the animation and its component images within the editing workflow.

OpenAI pitches the feature for animated mascots, social reactions, product loops and storyboards used to explore movement before video production. Its announcement also shows that a polished source image is not required: a simple sketch can be the starting point for an animated character.

Two models for different image jobs

For developers using the API—the interface through which software calls OpenAI’s models—the release offers GPT-Image-2.6 Flare and GPT-Image-2.6 Sunburst. OpenAI positions Flare around speed and Sunburst around greater detail and consistency across edits. Both are available through the API as part of the same release.

OpenAI says Images 2.6 generates images 50% faster than before. At the low quality setting, it says Flare can produce an image in as little as six seconds. Flare also improves layout quality for posters and product imagery, according to the company; its pitch is not limited to shorter waits.

Sunburst’s stated advantage appears at medium or high quality, where OpenAI says it delivers higher-quality images and greater consistency through revisions. Those are company performance claims. OrcaRouter reported that neither new model appeared on Artificial Analysis’s image-generation or image-editing leaderboards when it checked on November 5.

More control over the next revision

The API changes go beyond choosing a model. They add ways to identify where an edit belongs, decide how much planning a request needs and export a design for use elsewhere. OpenAI distinguishes planning effort from image quality rather than treating both as one setting.

• Image markup brings a tool already available in ChatGPT to the API. Developers can mark the part of an image that should change, giving the model a visual target rather than relying only on a written description.
• SVG export adds a vector output format for designs used across websites, apps and print. Vector graphics can scale without the pixelation associated with enlarging a raster image, making the format relevant to logos and icons.
• A complexity parameter controls how much planning the model does to satisfy a request. Quality controls visual detail and fidelity; complexity addresses the scene and its requirements. Raising one is not the same as raising the other.

OpenAI recommends low complexity for most editing tasks, paired with whatever image quality the work requires. It suggests medium or high complexity for scenes with many elements or constraints. The recommendation gives developers separate choices about the effort spent interpreting a composition and the detail of the rendered result.

OrcaRouter lists the same token rates for both API models: $5 per million text-input tokens, $8 per million image-input tokens and $30 per million image-output tokens, with lower rates for cached inputs. Its account says those rates are unchanged from Images 2.5. These are token prices, not fixed charges for each finished image.

OpenAI says Images 2.6 is available globally in ChatGPT and Codex across all subscription tiers, on desktop, mobile and web. The announcement therefore places the consumer tools and developer models in one release, rather than reserving the new image capabilities for API customers.