ChatGPT Sketch turns your bad drawings into detailed AI images
OpenAI rolled out ChatGPT Images 2.5 on Tuesday, bringing with it a new input method called Sketch that lets users guide image generation through freehand drawings made directly inside the chat interface. Rather than relying solely on text descriptions, users can now pair a rough visual with their prompt to give the model a clearer sense of composition or subject matter.
To use Sketch, you type @Sketch into the chat box, which opens a drawing canvas within ChatGPT. From there, you can draw whatever you have in mind - even a loose, imprecise doodle - and then describe how you want the final image to look. The model uses both the drawing and the text prompt together to produce the output. Early testing by The Verge found that a rough mouse-drawn cat sketch was successfully interpreted and rendered as a realistic photograph.
The update also adds the ability to leave comments directly on specific parts of images, which points toward a more iterative editing workflow. Instead of rewriting an entire prompt to adjust one element of a generated image, users can annotate the area they want changed - a more precise and intuitive approach to image refinement.
These additions reflect a broader trend in generative image tools toward multimodal input, where text, images, and now drawings can be combined to better communicate intent to a model. Sketch in particular lowers the barrier for users who struggle to describe visual ideas in words alone, making spatial and compositional cues easier to convey. With Images 2.5 as the underlying model, OpenAI appears to be focusing not just on output quality but on expanding the ways users can interact with the generation process.
