OpenAI releases GPT-4o image generation model, supporting multiple rounds of dialogue editing functions

On March 25, 2025, OpenAI announced that its latest image generation model was officially built into GPT-4o, and this breakthrough upgrade has brought users a stronger image generation and editing experience. It is reported that the feature has been gradually launched to all Plus, Pro, Team and free users of ChatGPT and Sora today. This news quickly attracted widespread attention from the technology community.

GPT-4o's image generation capabilities have made significant progress in several aspects. It not only can accurately render text, avoid the common garbled problems in AI-generated images in the past, but also can more accurately follow the user's prompt instructions to generate rich and realistic images. Users can quickly generate images that meet expectations through simple conversations, such as specifying screen scale, precise color (supporting hexadecimal code), or transparent background.

GPT-4o supports multiple rounds of dialogue modification functions. Users can gradually adjust the image content through continuous conversations, such as requiring "keep the character's hairstyle the same as the previous one" or "adjust the background color to blue", and the model can understand and automatically execute these instructions based on the context. This interactive editing method greatly improves the flexibility and practicality of image generation.

Currently, GPT-4o's image generation function has been deployed on ChatGPT and Sora platforms, covering all user levels. OpenAI said the feature will soon be expanded to Enterprise and Edu users, further expanding its influence. In addition, all images generated by GPT-4o will be embedded in C2PA metadata to identify them as AI-generated content, ensuring transparency and traceability.