ChatGPT Sketch turns your bad drawings into detailed AI images
OpenAI has released ChatGPT Images 2.5 and added a new Sketch feature that lets users draw a doodle inside ChatGPT and instruct the model to generate an image from that sketch. The Verge reporter tested the tool, reporting that a crude mouse-drawn cat sketch was converted into a realistic photo and that image-specific comments could be used to change details like eye color.
Why It Matters
Sketch gives users a direct, visual way to guide image generation and lets them mark parts of an image for edits, which could make iterative editing more intuitive. Combined with OpenAI's stated improvements to lighting, textures, edit-following, and latency, the update targets faster, higher-fidelity image creation within ChatGPT.
Key Facts
- announcement: OpenAI announced ChatGPT Images 2.5 on Tuesday
- new-feature: Sketch — draw a doodle inside ChatGPT to guide image generation
- how-to-activate: Type @Sketch in the chat box to open a drawing window
- reported-test: The Verge reporter turned a poor mouse-drawn cat sketch into a realistic photo
- in-image-comments: Users can leave comments directly on parts of images to request changes
OpenAI rolled out ChatGPT Images 2.5 and introduced a Sketch tool that lets users draw a simple doodle inside ChatGPT and then instruct the model to generate an image based on that sketch. The Sketch function is invoked by typing @Sketch into the chat, which opens a window where users can produce a drawing to serve as the basis for generation.
In a brief test described by The Verge, the reporter used a basic mouse-drawn cat sketch and asked ChatGPT to render it as a realistic photo; the resulting image matched the reporter's instructions. The platform also supports leaving comments tied to specific areas of an image, enabling users to request targeted edits — for example, changing the color of the cat's eyes.
OpenAI says Images 2.5 produces images with more natural lighting and richer textures, and that it better follows editing instructions over multiple turns. The company also claims generation latency has been reduced by up to 50% compared with Images 2.0.
According to OpenAI, ChatGPT Images 2.5 and the Sketch capability are available now for ChatGPT, ChatGPT Work, and Codex users across desktop, mobile, and web platforms.
Keep Reading
Meta bets on AI agent Muse to catch up in AI race
Nintendo isn’t scared of GTA VI
A new class action lawsuit questions whether Anthropic broke the law by misleading power users
OpenAI fought dirty on career-making math problem, says NYU mathematician
Original source: The Verge