13:42Machine-translated from Azerbaijani by AIRead in Azerbaijani
OpenAI Unveils Images 2.6 Image Model That Creates GIFs and Is 50% Faster
OpenAI has unveiled its new “Images 2.6” model. ChatGPT, Codex, and the API can now create GIFs based on a written prompt or an uploaded image. Users can upload a simple sketch and turn it into a more detailed, animated character. The new feature is rolling out worldwide across all subscription tiers on the web, desktop, and mobile versions.
Users can change the playback speed of a generated GIF, view the animation frame by frame, and edit each frame individually. Finished animations can also be shared with others. The feature is intended for creating brand characters, product images, reaction GIFs, and memes. Users can also create animated storyboards from rough sketches and test a character’s movements before filming. Although “Images 2.6” creates GIFs, it is not a system for producing complete AI-generated videos. The model is intended for short, looping animations and animated storyboards. A post that OpenAI later deleted said uploaded images could be stored for reuse in future prompts. However, it is not yet clear whether this feature remains in the final version. According to OpenAI, “Images 2.6” creates images 50% faster than the previous “Images 2.5” model. More than 4 billion images are created each week through ChatGPT and the API. In September, that figure was 3 billion. During successive edits, the model more consistently preserves the appearance of characters, products, and brand elements.
Developers can choose between 2 models in the API. Designed for speed, “GPT-Image-2.6 Flare” can create an image in approximately 6 seconds at the low-quality setting. The model produces cleaner layouts for posters and product images. “GPT-Image-2.6 Sunburst” displays sharper details at medium- and high-quality settings and keeps results more consistent during edits.
The API allows users to mark the part of an image to be changed and create SVG files for websites, applications, and print. This makes it possible to edit only the selected area without altering the rest of the image. A new complexity setting determines how much planning the model performs before creating an image. Lower settings are better suited to minor adjustments. Earlier this year, OpenAI discontinued the little-used “Sora” video-generation application. The application was reportedly costing the company approximately $1 million (1.7 million Azerbaijani manats) per day. The company later shifted its primary focus to still images. “Images 2.6,” however, represents a partial return to moving images. Before this, the “Intelligent UI” feature for creating interactive visual interfaces had also been introduced in ChatGPT.
Users can change the playback speed of a generated GIF, view the animation frame by frame, and edit each frame individually. Finished animations can also be shared with others. The feature is intended for creating brand characters, product images, reaction GIFs, and memes. Users can also create animated storyboards from rough sketches and test a character’s movements before filming. Although “Images 2.6” creates GIFs, it is not a system for producing complete AI-generated videos. The model is intended for short, looping animations and animated storyboards. A post that OpenAI later deleted said uploaded images could be stored for reuse in future prompts. However, it is not yet clear whether this feature remains in the final version. According to OpenAI, “Images 2.6” creates images 50% faster than the previous “Images 2.5” model. More than 4 billion images are created each week through ChatGPT and the API. In September, that figure was 3 billion. During successive edits, the model more consistently preserves the appearance of characters, products, and brand elements. Developers can choose between 2 models in the API. Designed for speed, “GPT-Image-2.6 Flare” can create an image in approximately 6 seconds at the low-quality setting. The model produces cleaner layouts for posters and product images. “GPT-Image-2.6 Sunburst” displays sharper details at medium- and high-quality settings and keeps results more consistent during edits. The API allows users to mark the part of an image to be changed and create SVG files for websites, applications, and print. This makes it possible to edit only the selected area without altering the rest of the image. A new complexity setting determines how much planning the model performs before creating an image. Lower settings are better suited to minor adjustments. Earlier this year, OpenAI discontinued the little-used “Sora” video-generation application. The application was reportedly costing the company approximately $1 million (1.7 million Azerbaijani manats) per day. The company later shifted its primary focus to still images. “Images 2.6,” however, represents a partial return to moving images. Before this, the “Intelligent UI” feature for creating interactive visual interfaces had also been introduced in ChatGPT.