ChatGPT can now generate animated GIFs with Images 2.6, adding short loops to its image tools. OpenAI's November 4 release also brings new controls for targeted edits and animation to developers using its API.
OpenAI says the update is available globally in ChatGPT and Codex across all subscription tiers on the web, desktop and mobile. The company pitches GIFs for reactions, animated characters and product loops.
The API has two versions: Flare, intended for faster iteration, and Sunburst, intended for more detailed work. OpenAI says Images 2.6 generates images 50% faster than before, with low-quality Flare requests taking as little as six seconds.
Frames, timing and the finished loop
A GIF plays a sequence of still frames. OpenAI's image-generation documentation gives API users control over both the number of frames and how long each stays on screen. The documented range is two to 64 frames, with durations from 20 to 2,000 milliseconds per frame.
The defaults are nine frames at 200 milliseconds each. That makes a 1.8-second sequence before it repeats. Changing the frame duration changes playback speed, while adding frames gives the animation more separate poses or states.
These documented API settings do not establish which controls appear in ChatGPT. The documentation also distinguishes generation from editing: the image-edit endpoint cannot return a GIF.
A small independent test by Kingy AI returned a two-second, 512-by-512 GIF from 2.6 Sunburst. Its review found uneven movement, an added mark and a noticeable jump when the loop restarted. The test used one attempt per task and AI visual assessment. It establishes neither average reliability nor a general quality ranking.
More planning can change the image
OpenAI's prompting guide separates rendering quality from complexity, the setting that controls planning before the image is drawn. Flare and Sunburst support low, medium and high complexity, with high as the default when it is omitted.
Lower complexity can suit a focused edit, such as replacing a background or changing a sign. Higher complexity can help with dense text or relationships between objects, but OpenAI warns that it can increase latency and introduce unwanted changes. Asking the model to plan more does not guarantee that untouched details will stay fixed.
The API also adds markup-guided edits and static SVG output. SVGs contain editable vector paths and shapes, useful for graphics that need resizing. The documented output excludes scripts, external resources and embedded raster images. As with GIFs, SVG output is supported for generation, not through the image-edit endpoint.
