OpenAI has launched a major update to ChatGPT, giving users native image-generation capabilities through its GPT-4o model. This upgrade, announced by CEO Sam Altman during a livestream, marks the first major leap in ChatGPT’s visual functionality in over a year.
GPT-4o Brings Text-to-Image and Photo Editing to ChatGPT
Until now, GPT-4o powered ChatGPT only for text-based interactions. With this upgrade, the model can now generate and edit images directly inside the platform. This new feature is available today for Pro subscribers, who pay $200 per month. OpenAI plans to expand access to Plus, free users, and API developers soon.
GPT-4o replaces the DALL·E 3 model for image tasks. Although it takes slightly longer to render images, OpenAI says GPT-4o creates more detailed and accurate visuals. Users can also edit existing images using features like inpainting, which allows changes to people, backgrounds, and objects within the scene.
How GPT-4o Learns to Create Visuals
To support this new image-generation ability, OpenAI trained GPT-4o on public datasets and proprietary content from partners like Shutterstock. This approach helps the model produce more refined results while avoiding legal concerns around copyrighted material.
OpenAI’s COO, Brad Lightcap, emphasized that the company takes artists’ rights seriously. The system avoids generating images that mimic the styles of living artists. Creators can opt out of the training process through a dedicated form, and OpenAI honors web-scraping exclusions from websites.
Image Ethics and Industry Implications
Unlike some competitors, OpenAI enforces guardrails to prevent misuse. The company implemented policies to stop the generation of images resembling copyrighted characters or watermarked content.
This launch follows Google’s rollout of Gemini 2.0 Flash, which also supports native image output. However, Google’s tool faced backlash for generating unauthorized images, raising concerns over content security and copyright compliance.
With this update, ChatGPT enters a new phase as a multimodal AI assistant, blending text and visuals in one interface. For users focused on cybersecurity, privacy, and digital ethics, these changes also underscore the importance of transparent AI training practices and clear content boundaries.
