During a livestream on Tuesday, OpenAI CEO Sam Altman announced a major upgrade to ChatGPT’s image-generation capabilities, marking the first significant enhancement in over a year.
ChatGPT can now natively create and edit images using the company’s GPT-4o model, expanding beyond its previous text-only capabilities.
GPT-4o has powered ChatGPT for some time, but until now, image generation was handled separately by DALL·E 3. The new model takes slightly longer to process images but delivers what OpenAI describes as more accurate and detailed visuals. It can also edit existing images, including those featuring people, by modifying elements or “inpainting” missing details.
The upgraded image-generation feature is now available for ChatGPT and OpenAI’s AI video tool, Sora, but only for Pro subscribers paying $200 per month. OpenAI says it will soon roll out the feature to Plus and free-tier users, as well as developers using OpenAI’s API service.
To develop this feature, OpenAI trained GPT-4o on publicly available data and proprietary data from partnerships with companies like Shutterstock. However, OpenAI remains tight-lipped about its full training dataset, citing both competitive and legal concerns. Generative AI models often face scrutiny over intellectual property (IP) risks, particularly in how they handle copyrighted material.
“We’re respectful of artists’ rights in terms of how we do the output, and we have policies in place that prevent us from generating images that directly mimic any living artists’ work,” said Brad Lightcap, OpenAI’s chief operating officer, in a statement to The Wall Street Journal. The company also offers an opt-out form for creators who want their work excluded from its training datasets and allows website owners to block its web-scraping bots from collecting their content.
The update comes just as Google introduced native image generation in its Gemini 2.0 Flash model, which quickly went viral—but not necessarily for the right reasons. Users discovered that Gemini’s image tool lacked proper safeguards, enabling the removal of watermarks and the creation of copyrighted characters. OpenAI will likely face similar scrutiny as it expands its own image-generation capabilities.
