OpenAI has introduced a new feature called “Images in ChatGPT,” allowing users to generate images directly within the ChatGPT interface using the advanced GPT-4o model. This rollout, which began on March 25, 2025, enhances the text rendering capabilities and is available for all subscription tiers, including Plus, Pro, Team, and Free.
Users in the free tier can expect usage limits similar to those previously applied to DALL-E, although specific numbers were not disclosed by the company. Taya Christianson, an OpenAI spokesperson, mentioned that these limits may evolve based on user demand.
Gabriel Goh, the research lead, described this new model as a significant advancement over earlier versions. Utilizing the “omnimodal” nature of GPT-4o, it can generate diverse types of data, including text, images, audio, and video. One of the notable improvements is “binding,” which refers to the model’s ability to accurately maintain relationships between various attributes and objects within an image.
Goh explained that previous AI image generators often faltered in this area, routinely mixing up colors and shapes, especially when tasked with rendering multiple objects. In contrast, the new tool can effectively manage around 15 to 20 objects in a single image without confusion, representing a considerable leap in both accuracy and reliability.
Additionally, Goh provided examples of the model’s enhanced binding capabilities, indicating that users can expect higher quality and more realistic images than those produced by previous models. While DALL-E remains accessible to its fans through a custom GPT, this integration into ChatGPT marks a pivotal step in expanding the functionality and versatility of OpenAI’s offerings in the AI-driven creative space.
