OpenAI has integrated a Sketch tool into ChatGPT Images 2.5, allowing users to upload rough doodles as prompts for image generation. The feature activates via the @Sketch command, opening a dedicated window where users can draw directly within the chat interface. A test of a mouse-drawn cat sketch successfully produced a realistic photograph matching specific instructions provided alongside the image. This update removes the previous requirement for text-only descriptions, enabling a more intuitive workflow for visual requests.
The change matters because it lowers the barrier for users who struggle to articulate complex visual ideas in words. By accepting imperfect sketches, the system accommodates a wider range of creative abilities without demanding precise vocabulary. The ability to add comments directly onto generated images also allows for iterative refinement, letting users point out specific areas that need adjustment.
* Activation requires typing @Sketch into the chat box
* Users can draw directly inside the interface using a mouse or touch
* Commenting on specific image sections enables targeted edits
