So much has modified within the AI trade within the 4 months since OpenAI launched ChatGPT Images 1.5. We have seen a heated race to build agentic tools, an unprecedented deal with the Pentagon and unending AI slop.
Now, OpenAI is again within the generative media sport. The corporate introduced on Tuesday that it is releasing ChatGPT Pictures 2, its next-generation picture mannequin.
ChatGPT Pictures 2 is supposed to create text-heavy designs, like on this matcha commercial and faux journal cowl.
It could appear unusual that OpenAI is releasing a brand new picture mannequin only a month after saying the shuttering of its once-viral Sora AI video app with the intention to deal with constructing enterprise-ready “core merchandise.” But it surely’s clear from how the brand new mannequin was constructed that OpenAI is not backtracking on that aim.
ChatGPT Pictures 2 is designed to provide text-heavy photos, together with infographics, scientific posters, examine guides and advertising supplies. The times of weird Sora videos and Studio Ghibli-inspired memes are over.
Now, the corporate is constructing AI that may do what it calls “economically priceless artistic duties.”
“The aperture and use instances for visible intelligence simply broaden so broadly, and we imagine that that is so important to ChatGPT’s imaginative and prescient for creating your individual private assistant, as a result of your artistic assistant is a big a part of who you’re as a person,” Adele Li, product lead for ChatGPT Pictures, informed reporters in a press briefing.
(Disclosure: Ziff Davis, CNET’s mother or father firm, in April 2025 filed a lawsuit towards OpenAI, alleging it infringed Ziff Davis copyrights in coaching and working its AI methods.)
In these examples, you possibly can see how significantly better ChatGPT Pictures 2 is at rendering legible textual content.
OpenAI has been chasing the dream of a super app, a one-stop store for all issues AI, constructed out of its Codex platform. ChatGPT Pictures 2 is bringing the artistic piece of that puzzle.
The brand new mannequin naturally improves typography, iconography and composition to provide extra skilled AI photos. It may well generate textual content in a number of languages. AI picture fashions have notoriously struggled with creating legible, factually right textual content. ChatGPT Pictures 2 is OpenAI’s finest mannequin for that but. Google beforehand improved its textual content rendering with Nano Banana Pro, however even that “better of the very best” mannequin struggled with accuracy.
ChatGPT Pictures 2 is rolling out to all customers now. Your era restrict relies on your plan: The extra you pay, the extra AI photos you possibly can generate.
Builders utilizing the mannequin within the API can create photos in 2K and 4K decision, although these larger resolutions are nonetheless in beta and could also be wonky. Paying customers also can create photos utilizing considering and reasoning fashions, which assist them search the net for info, compile it right into a readable design and double-check their work.
“Picture mannequin” would not appear to be fairly the proper time period for ChatGPT Pictures 2, although it’s technically right. ChatGPT would not seize the fantastical surrealism of AI imagery like Midjourney, nor supply wherever close to the modifying instruments of Adobe Firefly.
But it surely’s catering to a bunch of customers in the midst of the spectrum of Midjourney’s inventive fanatics and Adobe’s skilled creators: those that must create engaging content material.
Like Anthropic’s newly launched Claude Design, OpenAI’s ChatGPT Pictures 2 is geared toward working professionals. Lecturers can use it to create examine guides and illustrated lesson plans. Advertising and marketing managers can create social media posts and visible belongings.
You possibly can create as much as eight photos from a single immediate, like a three-page report, that keep visible consistency throughout all of them.
You can also make longer studies with ChatGPT Pictures 2, all matching pages.
That is the second half the AI-generated key lime pie recipe. Discover the visible consistency.
One draw back is that if you wish to tweak an AI picture, you may nonetheless must regenerate it. With extra text-heavy designs, that is extra more likely to be needed, so you may run via your credit faster. OpenAI mentioned it is centered on sustaining its iterative, prompt-based modifying stream to maintain it straightforward to make use of.
OpenAI’s security procedures have not considerably modified since its final picture mannequin. It nonetheless contains metadata via the C2PA customary, so AI photos’ origins may be recognized. Abusive and unlawful imagery remains to be prohibited in OpenAI’s insurance policies, an essential guardrail for AI corporations to successfully implement, given current examples of AI-generated deepfakes and nonconsensual intimate imagery.
