ChatGPT Sketch Turns Your Doodles Into AI Images

ChatGPT Images 2.5 arrived on September 8, and OpenAI says it generates images about twice as fast as the model it replaces. Its headline addition is ChatGPT Sketch, a drawing canvas that turns a rough doodle into a finished AI image. Type "@Sketch" in the message box, draw something with a mouse or a finger, add a short description of what you want, and ChatGPT builds a complete picture around your shapes. The update is rolling out across every ChatGPT plan, plus ChatGPT Work and Codex.

What is ChatGPT Sketch and how does it work?

Sketch is a drawing canvas built directly into the chat window, so the doodle and the finished image never leave the conversation. The canvas looks a little like MS Paint from the 1990s, simple enough to sketch with a finger or a mouse. No drawing skill is required to get a usable result. Sketch the layout of a room, the outline of an outfit you're planning, or a quick doodle of an idea. Add a sentence describing the style, and the model fills in the rest.

"I briefly tested it, and it successfully turned my bad doodle of a cat that I drew with my computer mouse into a realistic photo, per my instructions." Jay Peters, Reporter, The Verge

Peters also used the new Comment tool afterward to change the cat's eyes from blue to green, without redrawing anything.

What else changed in ChatGPT Images 2.5?

Three smaller changes matter just as much as Sketch in daily use: guided templates, targeted comment edits, and a faster underlying model. Templates now walk you through popular formats like posters and merch with a series of follow-up questions instead of a blank prompt box. The Comment tool lets you click a specific spot on a generated image, type what should change there, and only that part updates once you hit send. Every edit is also saved as its own version, so you can scroll back through earlier attempts instead of losing them. OpenAI says the update produces more natural lighting and richer textures and follows multi-turn editing instructions more reliably than before, though how much that shows up depends heavily on the image.

How does it compare to Google's Nano Banana 2?

Sketch has no direct equivalent in Gemini. Google's tool works from an uploaded reference image instead of a freehand drawing. In one side-by-side illustration test, a Lifehacker reviewer preferred Gemini's output for a children's-book style image. The same reviewer rated ChatGPT's photo edits as more natural in a separate forest-cabin test. Neither model won outright. According to Lifehacker's overall verdict, ChatGPT Images 2.5 has a slight edge, specifically because of Sketch and the ability to save prompts as reusable templates. Which tool looks better in practice depends more on the specific prompt than on the brand behind it.

Tips for getting better results from Sketch

The description you type alongside your drawing matters more than the drawing itself.

"I played around with the tool for a bit, and the results were quite impressive. Though it is being promoted as a sketch-to-AI Image tool, it appears that the prompt you pair with your sketch does much of the heavy lifting." Jibin Joseph, Reporter, PCMag

Keep the shapes rough and the composition clear. The model reads structure before detail, so a messy sketch with a clear layout works better than a careful drawing with none. Spend your real effort on the prompt instead: name the material, the lighting, and the mood you want, the same way you would for a text-only image request. If the first result is close but not quite right, use the Comment tool to fix one element at a time instead of redrawing the whole sketch.