OpenAI announced ChatGPT Images 2.5 on Tuesday, September 8, 2026, alongside a new feature called Sketch that lets users draw a doodle inside ChatGPT and have it converted into a generated image.
Sketch can be activated by typing @Sketch into the ChatGPT chat box, which opens a drawing window where users can create a rough illustration. ChatGPT then uses that drawing, along with a text prompt, to generate an image based on the sketch. The Verge’s Jay Peters tested the feature and reported that it successfully converted a mouse-drawn cat doodle into a realistic photo. A separate new capability also allows users to leave comments directly on specific parts of a generated image to request targeted edits, such as changing an element’s color.
ChatGPT Images 2.5 brings several updates to OpenAI’s image generation, including what the company describes as “more natural lighting and richer textures,” improved ability to follow editing instructions across multiple back-and-forth turns, and image generation latency reduced by up to 50% compared with Images 2.0.
The updates are available now for ChatGPT, ChatGPT Work, and Codex users across desktop, mobile, and web platforms.
The Sketch feature may lower the barrier for users who struggle to describe what they want in text alone, offering a more direct visual way to communicate image requests to the AI system.
Source: The Verge