OpenAI has launched a new image-generation tool built into its GPT-4o model, aiming to make it more useful for practical applications such as graphic design and advertising. The update marks a shift toward a system capable of producing detailed, highly specific images that adhere more closely to creative direction.
The new generator addresses several long-standing technical limitations that previously made AI less viable for professional use in visual design. Notably, it improves upon “binding” — the ability to correctly associate and position objects within a scene.
Where earlier models often misrepresented spatial relationships or confused object placement, the upgraded version can now place, for example, a sign reading “ice cream” on the wrapper, rather than floating arbitrarily in the scene.
Better Text Rendering
Another key advancement is in text rendering. Previous models struggled to produce legible, coherent words, typically generating jumbled approximations that resembled captchas more than readable text. The new version of ChatGPT demonstrates far more consistent performance in this area, making it significantly more relevant for tasks involving packaging, branding, or signage.
This evolution reflects a broader trend in AI, as models originally designed for text are increasingly being integrated with multimodal capabilities. While ChatGPT first emerged in 2022 as a text-only tool, OpenAI has gradually expanded its functions — first with code generation, then image generation via the DALL·E model. With GPT-4o, OpenAI now offers a unified model that can process and generate across text, images, voice, and even video input.
The company begins rolling out the updated generator to users this week, with availability expanding across all ChatGPT user tiers in the coming weeks.
“For those who hold a special place in their hearts for DALL·E, it can still be accessed through a dedicated DALL·E GPT,” Open AI said in their announcement.
What sets the new system apart is its ability to handle complex, multi-part prompts with higher accuracy that should greater reflect what you’re trying to create.
In another example usage from OpenAI, it showed how users can describe an entire four-panel comic strip — complete with characters, dialogue, and scene changes — and receive a visually coherent output. This kind of precision opens the door for a range of commercial uses, particularly in areas such as advertising, marketing, illustration, and content production.
The model also supports image uploads for editing, and it will be available in both the video generator Sora and GPT-4o.
For advertising teams, the new model could speed up ideation, enable fast prototyping of campaign visuals, and reduce the back-and-forth on concept iterations — all within a single, conversational interface.
The update signals OpenAI’s growing confidence that its tools are ready to support professional creative work. Rather than serving as novelty generators of surreal or dreamlike images, these AI systems are becoming viable assets in design workflows — capable of taking direction, obeying constraints, and producing usable visual content at speed.
“For example, if you’re designing a video game character, the character’s appearance remains coherent across multiple iterations as you refine and experiment,” OpenAI said in their annoucement.
4o image generation rolls out starting today to Plus, Pro, Team, and Free users as the default image generator in ChatGPT, with access coming soon to Enterprise and Edu.
