OpenAI announced ChatGPT Images 2.5 on Tuesday, introducing a new way for users to guide image generation. A feature called Sketch lets people draw a doodle directly inside ChatGPT and then describe how they want an image created from that drawing. The feature is activated by typing @Sketch into the chat box, which opens a window for drawing.
The Verge tested the tool and reported that it turned a rough cat doodle, drawn with a computer mouse, into a realistic photo based on the writer’s instructions. The update also adds the ability to leave comments directly on parts of images.
Why it matters
Sketch offers an alternative input method for image generation, letting users convey visual intent through drawing rather than text alone.
Who should care
ChatGPT users who create AI images and want more direct control over layout and composition may find the new sketch and commenting features useful.