OpenAI has launched ChatGPT Images 2.0, an advanced version of its AI-powered image generator. The new iteration offers a maximum image width of 2,000 pixels in a wider range of aspect ratios, including formats suitable for infographics. Significant improvements have been made to text rendering, enabling clearer generation of Japanese, Korean, Chinese, Hindi, and Bengali text. The update also enhances the creation of small text, icons, and interface elements, introducing subtle flaws for added realism.
ChatGPT Images 2.0 utilizes a general knowledge dataset, last updated in December, to interpret prompts and fill in missing information. Paid plan users can activate "thinking" and "pro" reasoning modes, allowing ChatGPT to access public web data for enhanced output and image reasoning. This feature aids in reducing errors and manual revisions.
The tool can now generate up to 10 images from a single prompt, maintaining visual consistency or applying various design styles.
Alongside the image generator update, OpenAI introduced Codex Labs, a new technical training service. This initiative aims to help organizations adopt and integrate OpenAI's Codex programming assistant, offering workshops and support for developer tools.