OpenAI revealed the ChatGPT Images 2.5 update on Tuesday, introducing a novel method for users to direct the platform’s image generation capabilities: by drawing. The new feature, named Sketch, enables users to create a rough draft inside the ChatGPT interface, which the AI then refines into a polished image based on accompanying text instructions.
To utilize the tool, users can type @Sketch into the chat input field, prompting a drawing window to appear. During an initial test, a rudimentary cat illustration drawn with a computer mouse was successfully converted into a realistic photograph after providing specific directions to the model.
Additionally, the update allows users to post comments directly on sections of generated images, further refining the collaborative creation process. The full details of the update are available via The Verge.
I drew a stick figure and got a photorealistic dog. The accuracy is genuinely scary impressive.
Finally! I’ve always hated trying to describe exact layouts with words. This is a game-changer.