Until recently, asking an artificial intelligence to draw a restaurant menu or an advertising poster usually ended in disaster: invented words, duplicated letters, and unreadable phrasesThat detail, which seemed minor, was actually the biggest obstacle to using these tools for serious work, from marketing to internal company documentation. With the launch of ChatGPT Images 2.0, OpenAI is trying to close precisely that gap between the spectacular and the practical, following the big leap forward in images.
The company presents a model that not only draws better, but also Reason about what you need to create, organize the information, and treat the text as a central part of the design.not as a secondary embellishment. The objective is clear: that what is generated can actually be used in professional environments, including in Spain and the rest of Europe.
From "burtulous" to usable menu: text as a turning point
In previous generations, it was common to ask for a restaurant menu and receive impossible dishes like "enchuita" or "burrto", with the typography all jumbled up. ChatGPT Images 2.0 represents a significant technical leap in the way text is drawn within images.from small labels to long blocks of prose, including menus, signs, or diagrams.
OpenAI claims that the model is capable of producing posters, menus, and editorial materials where The text is legible, grammatically coherent, and visually integrated.In internal tests and demonstrations, examples of food menus, academic posters, or magazine pages have been seen that, at first glance, could pass for work done by a human designer.
This advance is not limited to the Latin alphabet. One of the most notable aspects is that Images 2.0 handles scripts like Japanese, Korean, Chinese, Hindi, or Bengali better.For European companies with international operations, media outlets with editions in several languages, or educational institutions that prepare multilingual material, this capability opens up possibilities that were previously very difficult to automate.
More than just illustrating: images as a language and a working tool
OpenAI emphasizes an idea that sums up the product's shift well: “Images are a language, not decoration”In other words, the priority is no longer just that the result is visually appealing, but that it serves to explain something, sell a product, or structure complex information.
ChatGPT Images 2.0 can be generated infographics, maps, user interfaces, visual guides, storyboards and comics where both content and form matter. The model attempts to follow detailed instructions, place elements in the right place, and respect specific details indicated in the prompt: from brand style to the visual hierarchy of a presentation.
In a context like Spain, this means that a marketing team might ask, for example, a visual comparison between cities for remote work —Valencia, Málaga, and Bilbao— with icons, climate, cost of living, and quality of life, all organized in columns. Or that a small business can generate a social media poster with optimized text and a ready-to-publish format without needing to use more complex design software.
The "Thinking" mode: when AI thinks before it draws
The big new feature of ChatGPT Images 2.0 is the introduction of a mode of reasoning, commonly called Thinking or ThoughtThis option, available in paid subscriptions (Plus, Pro and Business), changes the way the model handles a request.
Instead of instantly generating the image from the text, the system can Structure the task, consult the web for updated information, and review your own results. before delivering it. In practice, this allows you to request, for example, an infographic with recent figures or the correct logo of a company, and for the model to be documented first to adjust the composition.
This mode is also capable of Analyze user-uploaded materials, such as PowerPoint presentations or strategy documentsFrom these files, you can extract the key points, respect the logos and corporate styles, and turn the information into internal posters, slides, or training materials that maintain the organization's visual identity.
The cost of this more "thought-out" approach is speed. OpenAI acknowledges that Creating a comic strip, a very dense infographic, or a detailed storyboard can take several minutes.For many creative teams and communications departments in Europe, this additional latency can be offset if it reduces the time spent manually retouching and back-and-forth on the design.
Visual coherence: several images, same story
One of the classic limitations of generative image models was the lack of continuity between scenes or panelsThey changed character features, key objects, or styles from one panel to another without much logic, making it difficult to use them for complete campaigns, comics, or coherent presentations.
ChatGPT Images 2.0 addresses this issue by allowing images to be generated in a single request. up to eight or even ten images while maintaining the identity of characters and objectsThis is useful for designing storyboards, manga sequences, interior design projects, or series of creative content for social media where the same protagonist, color scheme, and style must be maintained.
OpenAI explains that this continuity rests on an architecture capable of manage complex spatial relationships, 3D perspectives, and cross-references between scenesFor a marketing manager working from Madrid or Barcelona, for example, it can be a tool to quickly design a multi-format campaign that respects the same graphic concept in all pieces.
Formats, resolution and styles: more control over the result
Another area where the new model improves upon its predecessors is the management of formats and aspect ratiosChatGPT Images 2.0 supports a wide variety of aspect ratios, from 3:1 panoramic for web banners to 1:3 vertical compositions designed for mobile devices, as well as common formats such as 16:9 or 4:3.
In the gpt-image-2 API, images can reach Resolutions up to 2K or 4K, depending on the plan and parameters chosenWhile the standard resolution in the ChatGPT interface is somewhat more limited, especially on free accounts, this flexibility makes it easier to adapt the output for corporate presentations, advertisements, covers, social media posts, or educational materials without relying as heavily on subsequent cropping.
The model has also learned to be more faithful to the requested styleWhether it's realistic photography, cinematic aesthetics, pixel art, manga, European comics, or minimalist interfaces, for media outlets, teachers, freelance designers, or small agencies in Spain, this means being able to directly order a "technology magazine cover in Spanish, with a clean, minimalist style, ready for printing" and get a result closer to what they envisioned.
Up-to-date knowledge and "memory" of the world
OpenAI indicates that ChatGPT Images 2.0 is trained with Information up to December 2025This means that the model understands relatively recent references, technologies, current iconography, and design trends that are still relevant in 2026.
For cases where data from after that date is required—for example, recent economic figures, regulatory changes in the European Union, or breaking news—the mode of reasoning may Consult the website before composing the imageThus, an infographic about the labor market in Spain or a map with new European infrastructure is more likely to reflect the current situation.
Even so, the risk of errors or visual "hallucinations" remains. OpenAI itself admits that the model He still stumbles over tasks that demand perfect physical understanding.such as complex origami folds or certain spatial puzzles. Very small and repetitive details—like millions of grains of sand—remain a technical frontier where the result may not be entirely faithful.
Deployment, access, and business model
OpenAI has opted for a broad deployment from the outset. ChatGPT Images 2.0 is available to all ChatGPT users, both in free accounts and in paid Go, Plus and Pro plans, with differences in capabilities and speed.
Non-subscribers can access the basic model, which already includes a notable improvement in image quality and text handling. Those with paid plans, however, have access to... Advanced reasoning functions, web search, document analysis, and generation of multiple images in a single requestIt is at these levels that the "think before you draw" approach is fully exploited.
In parallel, the company has launched the gpt-image-2 API, with Prices vary depending on resolution, quality, and usage volume.This allows European companies to integrate the model into their own applications, from e-commerce platforms that generate banners in real time to internal documentation tools that turn reports into automated visualizations.
Security, copyright, and content labeling
The expansion of the visual generation is also accompanied by concerns about copyright, sensitive content, and misinformationOpenAI states that it has strengthened security protocols in Images 2.0 through filters, usage policies, and watermarking or metadata systems that indicate the synthetic origin of the images.
The company anticipates restrictions for Avoid the direct reproduction of protected works or copyrighted characters.This will affect those who try to create, for example, a manga based on well-known franchises. In Europe, where the regulatory debate on AI and copyright is particularly active, these measures will be analyzed by both regulators and rights holders.
The approach of labeling images with metadata generated by AI itself aligns with the lines of work being discussed in the European Union and other international forums, where it is valued that the public can to more easily identify which content has been generated or modified by AI systems.
Competition and positioning in the visual AI market
The launch of ChatGPT Images 2.0 comes in a highly competitive landscape. Models like Google's Midjourney, FLUX or Nano Banana They have carved out a niche for themselves in the artistic field, photorealism, or conversational image editing.
Instead of simply replicating that approach, OpenAI is trying to differentiate itself by presenting ChatGPT as a integrated environment where visual creation is part of a broader flow It combines text, code, data analysis, and now also structured design. The promise is that the user can move from an idea to a campaign, a report, or an interface without leaving the same ecosystem.
For professionals and organizations in Spain and the rest of Europe, this integration could be of interest if it is indeed It reduces friction between content, design, product, and technology teams.At the same time, it raises questions about vendor lock-in, data protection, and adaptation to future AI regulations in the European sphere.
The arrival of ChatGPT Images 2.0 marks a turning point in AI image generation: the focus shifts from isolated visual impact to practical utility, with legible text, controllable formats, prior reasoning, and coherence between scenesIt remains to be seen how users, companies, and regulators will respond, but the movement points to a scenario in which more and more of the visual content we consume—from restaurant menus to educational infographics or digital interfaces—may have been designed, at least in part, with the silent help of these types of models.
