ChatGPT-4o Now Generates ‘Beautiful and Useful’ Images
ChatGPT-4o Delivers Faster, More Realistic, and Higher-Quality Image Generation
ChatGPT-4o Now Produces ‘Beautiful and Useful’ Images
ChatGPT-4o Produces Faster, More Realistic, and More High-Quality Image Generation
On Monday, OpenAI announced a new image-generation feature on ChatGPT, backed by GPT-4o. Accessible to Plus, Pro, Team, and Free users, and developers through API, it is reportedly capable of producing images that are “not only beautiful but useful.”
It’s a phenomenal technology/product. I recall seeing some of the initial images produced by this model and having trouble believing they were actually created by AI. We believe folks will adore it, and we can’t wait to see the resulting creativity,” OpenAI CEO Sam Altman posted on X.
Although ChatGPT has been generating images for some time, the previous versions had a different model for generating images and text. The new GPT-4o combines both functions into a single entity, leading to quicker image generation and better contextual comprehension to create more realistic and higher-quality images. Graphic designers and artists can now create images with more detail and precision.
GPT-4o employs methods such as text rendering, multi-turn generation, instruction following, and in-context learning to improve picture quality. With the world knowledge of GPT-4o, this integration connects text and image knowledge to deliver more accurate and photorealistic images with improved style.
But OpenAI does admit that the image generator is not flawless. Sometimes, it creates stretched-out images that are too narrowly cropped, introduces errors through hallucinations, and fails to produce knowledge-based images like the periodic table. The firm has indicated that these will be fixed after release.
Images created by GPT-4o will contain Coalition for Content Provenance and Authenticity (C2PA) metadata for transparency. The system also cannot produce images that break content policies, including child sexual abuse content and sexual deepfakes. It also has controls to avoid producing images with nudity and graphic violence.
