OpenAI has launched ChatGPT Images 2.5, an updated image generator focusing on precision editing. The model improves object identity preservation, ensuring key elements like people remain consistent when changing styles or scenes. It now more effectively targets specific image aspects, such as backgrounds, clothing, or text, without affecting other parts of the image.
The system's performance has increased, reducing request processing time by up to 50%. The model is now more responsive to complex prompts, including detailed compositions, text, and transparent backgrounds. A new Sketch function allows users to draw rough layouts directly in the interface to guide spatial composition.
The tool is available on ChatGPT, ChatGPT Work, and Codex via desktop, mobile, and web interfaces, as well as through an API. It comes in two versions: the universal GPT-Image-2.5 Flare and the high-precision, slower GPT-Image-2.5 Sunburst. To ensure transparency, all generated images include invisible watermarks and C2PA metadata.
