During Tuesday's live stream, Openai CEO Sam Altman announced his first major upgrade to ChatGpt's image generation capabilities in over a year.
ChatGPT can now natively create and modify images and photos using the company's GPT-4O model. The GPT-4o has long supported AI-powered chatbot platforms, but up until now, models have been able to generate and edit text, not images.
According to Altman, GPT-4o Native Image Generation today said Openai's AI video generation products ChatGpt and Sora are live for the company's $200 monthly Pro Plan subscribers. According to Openai, the feature is quickly rolling out to ChatGPT users and free users as well as developers using the company's API services.
The GPT-4O with image output is a little longer “thinking” than an image generation model that effectively replaces what OpenAI describes as a more accurate and detailed image. The GPT-4o can edit existing images, including images with those people. You can convert them, and “enter” details such as foreground and background objects.
To enhance the new imaging capabilities, Openai told the Wall Street Journal that it had trained the GPT-4o on “published data” and its own data from partnerships with companies such as Shutterstock.
Many generative AI vendors consider training data a competitive advantage and therefore retain the information associated with it near the chest. However, the details of the training data are also a potential cause of IP-related lawsuits. This is another hindering business to reveal a lot.
“We respect artists' rights in terms of how we output, and we have policies that prevent them from generating images that directly mimic the work of living artists,” Brad Lightcap, Chief Operating Officer at Openai, said in a statement to the Journal.
Openai offers an opt-out form that allows creators to request that they remove works from their training datasets. The company also says it respects its request to ban web scraping bots from collecting training data, including images from websites.
ChatGpt's upgraded image generation feature follows Google's experimental native image output for Gemini 2.0 Flash, one of the company's flagship models. The powerful feature has gone viral on social media, but not necessarily for the best reason. Gemini 2.0 Flash's image components have proven to have few guardrails, and can now remove watermarks and create images depicting copyrighted text.
This article was updated at 12pm Pt, which includes Openai's statement to the Wall Street Journal on GPT-4o training data.