OpenAI adds Sketch tool to ChatGPT Images 2.5 for doodle-to-photo creation

OpenAI announced the rollout of ChatGPT Images 2.5, which includes a Sketch function that transforms hand-drawn doodles into high-quality AI-generated pictures. Users activate the tool by entering @Sketch, prompting a canvas where they can sketch any shape; the system then interprets the drawing and creates a realistic image according to further textual guidance. The feature also supports direct comments on specific image regions for targeted edits, such as changing eye color.

According to OpenAI, the upgraded model delivers more natural lighting, richer textures, and better adherence to multi-step editing prompts, while reducing latency by up to 50% versus the previous version. The service is now live for ChatGPT, ChatGPT Work and Codex users on all major devices. Early testing showed a crude mouse-drawn cat being turned into a photorealistic feline with customized eye color.