Story
September 8, 2026
ChatGPT’s Sketch Tool Makes Image AI Easier—and Its Creative Debt Harder to Ignore
OpenAI presents its new image tools as a way to give users more authorship over AI-made visuals, while critics see a familiar contradiction: easier creation built on creative labor many artists say was never fairly recognized or paid for.
On Tuesday, OpenAI rolled out ChatGPT Images 2.5, a new image engine aimed at fixing some of generative AI’s most visible weaknesses: unreliable likenesses, clumsy revisions and the sense that every output comes from the same synthetic style.
The upgrade’s practical pitch is straightforward. Users can sketch directly in ChatGPT, point to a specific area of an image and leave an instruction, rather than repeatedly trying to describe a change in text. Axios found that the tool better preserved people and pets, handled more detailed edits, and could generate images “up to 50 percent faster,” according to OpenAI.1
OpenAI’s leadership framed the release as an expansion of creative possibility. Greg Brockman called Images 2.5 “another step for creating whatever you can dream of.”
2 That argument is reinforced by the product’s focus on iterative control: turning a rough doodle into an image, adding an object with a sketch, or modifying only one element without rebuilding the whole scene.
Early testing suggests the mechanics are delivering. The Verge reported that its reporter’s mouse-drawn cat became a realistic image, while comment-based edits changed the animal’s eye color; OpenAI also says the model follows editing directions more reliably across multiple turns.3
But the same capabilities sharpen the ethical dispute. Axios noted that an editor can imitate recognizable styles—from The Simpsons to Sesame Street-like characters—because generative systems were trained on artists’ work, often “without consent or compensation.”1 For artists competing with outputs built from that material, better fidelity and faster production are not neutral technical gains.
OpenAI product lead Adele Li offers a different reading of Sketch and templates: they are meant to make people active participants rather than “a passive force” in creation. Her stated ambition is not a recognizable house look—“I don’t want to be able to see ChatGPT in the world”—but tools that let users express “their own individualism.”1
That is the fault line behind Images 2.5: more control for the person prompting, and still-unresolved questions over who supplied the creative foundation beneath the prompt.