OpenAI has introduced a new image generator, ChatGPT Images 2.0. dev.ua tested the tool and checked how it works with complex layouts, infographics, and text.
OpenAI has introduced a new image generator, ChatGPT Images 2.0. dev.ua tested the tool and checked how it works with complex layouts, infographics, and text.
OpenAI announced that ChatGPT Images 2.0 is a new version of the image generation tool in ChatGPT. The model is focused on creating complex visual materials — from posters and presentations to comics and UI layouts — with more precise adherence to given parameters.
According to the company, the key update is improved text rendering in images. Previously, generators often distorted or made labels unclear, now the model is able to more accurately form small text elements, captions, and interface components even in complex compositions.
OpenAI also claims to have improved language recognition, particularly in non-Latin scripts. This means the model should be able to more accurately reproduce text in Japanese, Korean, Hindi, and other languages where previous versions often made mistakes.
We tested the new capabilities of the image generator by asking it to generate an infographic of officially registered IT companies in Ukraine by the end of 2025. Here’s what came out of it:
As you can see, Images 2.0 copes much better with any text in Ukrainian, even small fonts.
Another change is the appearance of a «thinking» mode, in which the model can spend more time preparing the result. In this mode, the system analyzes the query, refines the image structure, and only then generates the final version. This approach is used for complex tasks, such as multi-frame comics or marketing materials with several elements.
In addition, OpenAI has added the ability to create multiple image variants from a single query, as well as support for different formats and sizes to prepare graphics for specific platforms.
The company has made the model available to developers through an API called gpt-image-2. It supports image generation, editing, and working with multiple resolutions, including more detailed options.
ChatGPT Images 2.0 is the next step in the development of image generation in OpenAI products, following previous models that worked on diffusion approaches. The company emphasizes that the new system better copes with complex scenes, small details and text — these are the elements that have previously most often caused problems for generative models. The new tool is already available in ChatGPT and via the API for developers.
Previously, dev.ua wrote about how OpenAI presented GPT-Rosalind, its first specialized AI model focused on the scientific sector. The new product is capable of complex logical reasoning in the field of biomedical research.
