OpenAI has introduced a new image generation feature in ChatGPT using its latest GPT-4o model. The update, called “Images in ChatGPT,” allows users across various subscription tiers—including Free, Plus, Pro, and Team—to create AI-generated images directly within the chat interface. The free tier will have similar usage limits as DALL-E, though OpenAI has not disclosed an exact cap, stating that limits may change based on demand.
The new model brings notable improvements in accuracy, particularly in “binding,” a challenge where AI struggles to maintain correct relationships between attributes and objects. Previous models often mixed up colors and shapes, especially when rendering multiple items. With GPT-4o, the system can now correctly bind up to 15 to 20 objects without confusion, a significant leap from earlier limitations of handling only 5 to 8 objects reliably.
OpenAI research lead Gabriel Goh highlighted that GPT-4o is an “omnimodal” model capable of generating various data types, including text, images, audio, and video. This advancement enhances ChatGPT’s ability to produce more consistent and detailed images, making it a more reliable tool for creative and professional use.
The future of DALL-E remains uncertain, though OpenAI confirmed that users can still access it through a custom GPT. With the launch of GPT-4o’s image capabilities, OpenAI continues to push the boundaries of AI-generated content, improving usability and accuracy across its platforms.
