OpenAI releases Images 2.6: one photo turns into a GIF, image generation speed increased by 50%

Beating AI News Flash: OpenAI releases image generation model Images 2.6, and ChatGPT can finally directly generate GIF animations. Enter a description or upload a photo to make the image move.

Users can also adjust animation speed, edit frame by frame, and even turn simple hand-drawn sketches into animated cartoon characters. These operations previously could not be completed in one step in ChatGPT.

In addition to making images move, the new version also solves a common problem in image generation: repeated edits can easily distort people, products, and even brand logos beyond recognition. Images 2.6 strengthens consistency across multi-turn edits while increasing image generation speed by 50%.

Developers can use two new models via the API. Flare prioritizes speed, generating images in as fast as 6 seconds at low quality; Sunburst focuses more on image quality and detail, making it suitable for complex designs that require repeated edits.

In addition, the API now also supports circling the area to be modified on an image, exporting SVG vector graphics, and adjusting how much effort the model puts into handling complex scenes.

This update has been made available to ChatGPT and Codex users across all subscription tiers worldwide, and the two models are also available via the API.

OpenAI disclosed that users worldwide now generate about 4 billion images per week through ChatGPT, Codex, and the API. Compared with 3 billion in September this year, that is an increase of about 1 billion in just two months.