GPT Image 2.6 creators demonstrate animated GIF output

• Animated GIF tests now include a personalized cat reaction from ozansihay and a photograph-style café loop from ai_artworkgen.
• Still-image preferences remain mixed: ai_artworkgen said they preferred GPT 2.5 over 2.6 in the accompanying comparison thread.
• GIF generation tops out at 64 frames, a limit aakashgupta reported and OpenAI’s API guide confirms.

OpenAI’s launch examples include an OpenArt prompt that packs six animation poses into a 3-by-2 sprite sheet. The API documentation also exposes frame timing controls and SVG output with editable paths, shapes, and colors.

ozansihay supplied a photo of the family cat, Robin, with this request, translated from Turkish:

Create a GIF of our cat Robin. Have the cat look at the camera and wink, and have a hat reading “Robin” fall onto its head.

ai_artworkgen’s café prompt takes a pleasingly restrained approach to animation. Its instructions separate the photographic setup from the few things allowed to move:

• Scene: A man in his late fifties, wearing a worn navy jumper, sits by a café window on an overcast morning.
• Photography: Candid 35mm framing, shallow depth of field, Kodak Portra 400 grain, soft grey daylight, and a shadowy room behind him.
• Motion: Only the hand stirring coffee, a blink, and rising steam.
• Loop: The last frame should match the first.

The still-image tests also used a reference image created with Runway Frames. In the accompanying discussion, ai_artworkgen preferred GPT 2.5 over 2.6, naming Nano Banana Pro and 2.1 as personal favorites and pointing to the lighting comparisons.

OpenAI’s November 4 announcement separates the two API models by speed and detailed editing work:

• Flare: Still-image generation in as little as six seconds at low quality, according to OpenAI.
• Sunburst: Higher detail and greater consistency across edits at medium or high quality, according to OpenAI.
• Availability: Images 2.6 is available globally in ChatGPT and Codex across all subscription tiers, with both models also available through the API.

Both models have a separate planning control. The [generation guide](https://developers.openai.com/api/docs/guides/image-generation# set-image-complexity) lists complexity levels of low, medium, and high, defaulting to high; quality separately controls rendering quality. OpenAI cautions that higher complexity can increase latency and introduce unwanted changes.

ai_artworkgen also shared the generation settings alongside an output example:

The GIF documentation specifies these request controls:

• Format: output_format: "gif" selects animated GIF output.
• Frame count: frame_count accepts 2 to 64 frames, defaulting to 9.
• Frame timing: frame_duration_ms accepts 20 to 2,000 milliseconds per frame, rounded to 10-millisecond increments. The default is 200 milliseconds.

GIF output is supported by POST /v1/images/generations. The image-edit endpoint does not support GIF output.