OpenAI just released ChatGPT Images 2.0 — an image-generation engine built straight into the chat experience and accessible via API. What stands out is its knack for complex prompts: it can position items cleanly and nail tiny visual cues, and embedded text often reads correctly. You can also switch the aspect ratio with one click in the interface, which saves a lot of fiddly cropping.
Under the hood there's a blend of visual and reasoning models. The network inspects context, can pull up-to-date info from the web, and runs internal checks (i.e., basic consistency and legibility tests) before handing back results. It spits out multiple variations at once — e.g., different color palettes or layouts — so you can grab the direction you like and iterate.
The update is available now to all ChatGPT users at no cost. That means folks outside traditional design workflows can experiment with pro-level tools; I found it useful for quick mockups, though it's still smart to double-check details (especially text in images) before using outputs commercially.