OpenAI’s image generator now uses web data

OpenAI has launched a significant upgrade to its AI image generation tool, now enhanced with new thinking capabilities that allow it to search the web for information. This latest version, ChatGPT Images 2.0, is designed to produce more sophisticated and detailed visuals by improving its ability to follow complex instructions and maintain specific user-defined details across multiple outputs.

The upgrade is powered by the new GPT Image 2 model, with its advanced reasoning features accessible to ChatGPT Plus, Pro, Business, and Enterprise subscribers. When users select the thinking model, the image generator can access real-time web data, create visual explanations based on uploaded files, and logically plan an image’s structure before beginning the generation process. A key new feature is the ability to produce up to eight coherent images from a single prompt, ensuring consistent characters, objects, and artistic styles across all outputs. This functionality is particularly useful for creating sequential content like manga pages, a unified set of social media graphics, or a complete series of interior design concepts for a home.

Beyond the premium thinking features, all ChatGPT users benefit from core improvements. The model now excels at capturing the distinct qualities of various visual formats, including pixel art, cinematic stills, and photographs. It supports higher resolutions up to 2K and offers a wider range of aspect ratios, from a broad 3:1 format to a tall 1:3 layout. Furthermore, OpenAI reports significant gains in the tool’s ability to generate accurate text within images, especially for non-Latin scripts. This means notably better performance for languages such as Japanese, Korean, Chinese, Hindi, and Bengali, expanding its global utility and accessibility.