OpenAI Integrates Intelligent Image Generation Powered by GPT-4o into ChatGPT

OpenAI has now integrated a powerful image generation tool directly into ChatGPT. From today onward, users can create detailed visuals using the updated GPT-4o model, which excels at understanding context, handling intricate requests, and even generating perfect text within images.

Earlier versions of ChatGPT relied on the DALL-E 3 neural network, but the new system — dubbed Images in ChatGPT — operates on the multimodal “universal” GPT-4o. This means the model simultaneously processes text, audio, video, and images, dramatically boosting accuracy. Now, the AI can accurately associate up to 15–20 objects in a single image, avoiding the errors that previously occurred with as few as 5–8 elements.

One of the key improvements is the ability to produce images with perfectly legible text — a challenge that had long been the Achilles’ heel of generative models. ChatGPT can now generate restaurant menus, logos with captions, or infographics without any typos. Additionally, users now have an editing feature: any uploaded image can be modified by simply adding or removing elements through a text prompt.

This new feature is available to all ChatGPT subscriptions, including the free tier, though similar limits to those of DALL-E 3 still apply. Meanwhile, DALL-E will continue to be available through custom GPT modules. OpenAI has also paid special attention to safety: all images are marked with C2PA metadata to identify AI-generated content, and the system blocks requests related to violence, deepfakes, and other potential abuses.