OpenAI has introduced an upgraded image-generation feature on ChatGPT, powered by its advanced GPT-4o model. This enhancement is available to Free, Plus, Pro, and Team users, as well as developers through the API. According to OpenAI, the model can produce images that are not only visually appealing but also practical for various uses.
Sam Altman, OpenAI’s CEO, expressed his excitement about the new capability in a post on X, stating that he was amazed by the first AI-generated images from this model. He believes users will appreciate its creative potential.
Previously, ChatGPT used separate models for text and image creation. However, GPT-4o combines both functions into one, resulting in quicker image generation and better contextual understanding. This leads to more lifelike, high-quality visuals. With this upgrade, graphic designers and artists can create more detailed and precise images with greater ease.
GPT-4o uses advanced techniques like text rendering, multi-turn generation, instruction following, and contextual learning to improve image quality. By integrating its extensive world knowledge, the model generates clearer and more realistic visuals with improved style accuracy.
Despite its impressive capabilities, OpenAI acknowledges some limitations. The model may occasionally create awkwardly stretched images, tightly cropped visuals, or display minor inaccuracies, especially with complex information-based visuals like charts. The company has promised to fix these issues in future updates.
To ensure transparency, images created by GPT-4o will carry metadata from the Coalition for Content Provenance and Authenticity (C2PA), making it clear that they are AI-generated. Additionally, OpenAI has enforced strict content guidelines, preventing the creation of harmful or inappropriate images, such as explicit content or deepfakes.
