OpenAI announced ChatGPT Images 2.5 and the Sketch feature, which allows users to draw directly within ChatGPT and then describe how the final result should look based on that drawing. To launch the feature, users must type @Sketch in the chat field, which opens a window for creating the drawing.
This innovation shifts part of the generation control from text descriptions to a simple visual draft. Users can first sketch an object and then supplement that sketch with instructions on how to transform it into an image.
The practical value of the feature depends on how accurately Sketch preserves the shape and key elements of the original drawing. The presented material contains no data regarding the feature's availability to all users, its limitations, the quality of results in different scenarios, or its rollout timeline.
editorial commentary
Why it matters
A likely consequence is more visual control over generation for users who find it difficult to precisely describe composition with words. The next observable signal will be information about Sketch's availability and examples of the feature working in different scenarios. Significant uncertainty remains regarding the quality of sketch conversion, limitations, and the scale of the launch.