Image generation now occurs via GPT-4o instead of DALL-E.
OpenAI has today released a major update for ChatGPT’s built-in image generation tool. The company recently also released a tool for video generation: Sora.
Farewell DALL-E 3
Image generation has been renewed in ChatGPT. OpenAI has replaced DALL-E 3 with GPT-4o image generation. This means that generation has greatly improved, as GPT-4o is a large language model that better understands and processes text and images.
According to OpenAI, GPT-4o can draw twenty different items specified by the user. The model now clearly shows even readable text, something that often went wrong before. Users can also provide ChatGPT with reference images to modify. In an example, the model can create transparent backgrounds. Transparent backgrounds are interesting for graphic designers to combine images.
Wall Street Journal reports that OpenAI trained GPT-4o using publicly available data and assets licensed from partners such as Shutterstock. “We trained our models on the joint distribution of online images and text, learning not just how images relate to language, but also how they relate to each other,” OpenAI employees wrote in a
blog post.
The improved image generator is now available in the free version of ChatGPT, as well as in Plus, Pro, and Team. OpenAI will soon bring the feature to Enterprise and education subscriptions as well.
read also