ChatGPT is getting animated GIFs. OpenAI says Images 2.6, out now in ChatGPT, Codex and its API, can turn a prompt or a photo into a GIF, and can even generate images 50% faster than before.
Once a GIF is made, you can change its speed, edit individual frames and share it on repeat. OpenAI pitches this new feature for reactions, brand mascots, and product visuals.
These are short loops, not actual video. OpenAI’s API guide allows 2 to 16 frames per GIF, at 20 to 2,000 milliseconds a frame, and the default of nine frames at 200 milliseconds runs for just 1.8 seconds. They arrive about seven months after OpenAI announced the end of Sora, its video app.
OpenAI’s system card says GIFs bring risks still images may not show, such as sexualized movement, so GIF rules block some content more broadly. It also admits the safeguards may not catch every motion-related violation.
It’s the second time OpenAI made a 50% speed claim in about eight weeks. Images 2.5 arrived on September 8th promising waits up to 50% shorter than Images 2.0. OpenAI puts usage at more than 4 billion images a week, up from 3 billion in September.
The wait for an image is also where OpenAI planned its test of big picture ads for Free and Go users in the US.
ChatGPT can also keep reference images of you, your pets and your characters to reuse without uploading them again.
Developers get two models. OpenAI says GPT-Image-2.6 Flare can generate an image in as little as six seconds at low quality, while GPT-Image-2.6 Sunburst aims for more detail and more consistent edits. Both add SVG and GIF output. Their token rates match the Images 2.5 models: $8 (about Php 500) per million image input tokens and $30 (about Php 1,900) per million image output tokens.
Images 2.6 is available on all ChatGPT and Codex tiers on web, desktop and mobile. For now, GIF creation is still being rolled out in ChatGPT, so don’t panic if you don’t see it in your account yet – just check back again later.
