OpenAI’s latest image generator, integrated into the GPT-4o model, marks a significant advancement in AI technology. This tool, designed for practical use in advertising and graphic design, excels in object binding with a 90% accuracy rate. It can correctly place objects like signs above food carts. The generator also improves text rendering, creating legible text in various formats, such as recipe cards and comic strips. It supports the creation of up to 12 discrete graphics within a single image. OpenAI plans to roll out this new model to all users over the coming weeks, replacing the older DALL-E system. The development of this generator involved analyzing millions of images to ensure text is displayed correctly at different angles and orientations. This progress sets a new standard in the industry, pushing other AI companies to innovate further.
Source: www.technologyreview.com















