OpenAI announced the image generation AI "ChatGPT Images 2.5" on September 8 (local time). This model enhances the ability to modify specific areas of an image while maintaining the overall composition and other untouched parts. OpenAI stated that it is now easier to preserve previous changes and image quality even through repeated edits.

New features added to ChatGPT include the ability to provide editing instructions by writing comments directly on an image, a "Sketch" feature that generates images based on drawn drafts, and templates for items such as posters and product photography. Additionally, it is now possible to share the prompts used for generation along with the resulting images. The time required for image generation is reported to be reduced by up to 50% compared to the previous "Images 2.0."

For developers, two new models will be provided via API: the standard "GPT-Image-2.5 Flare" and "GPT-Image-2.5 Sunburst," which takes more time during generation to achieve higher editing precision. These features became available starting the same day for all plans of ChatGPT, ChatGPT Work, and Codex.

Source: