OpenAI has announced "ChatGPT Images 2.5," a new image generation model that improves image quality and editing precision. This model enables more natural lighting and richer texture expression, while also enhancing the ability to maintain the subjects of reference images. Furthermore, generation latency has been reduced by up to 50% compared to the previous Images 2.0.

ChatGPT will introduce a "Sketch" feature, allowing users to use their own sketches as guides for image generation, as well as a template feature to select formats such as flyers or product photography. Additionally, features have been added to allow users to leave comments directly on images to instruct edits or share the prompts used. These features will be provided to all users of ChatGPT, ChatGPT Work, and Codex across desktop, mobile, and web platforms.

For developers, two new models will be available via API: "GPT-Image-2.5 Flare," which balances high-quality image generation with fast response times, and "GPT-Image-2.5 Sunburst," which is suited for more sophisticated creative work. OpenAI stated that it has applied a multi-layered safety stack to these new models to detect and block policy violations in both prompts and generated images.

Source: