OpenAI has taken a significant leap forward in the field of artificial intelligence with the introduction of Sora, a cutting-edge image generation capability powered by GPT-4o, which is now seamlessly integrated into ChatGPT. This innovative feature, dubbed "Images in ChatGPT," enables users to generate images directly within the popular chatbot platform.

Sora was initially announced as an AI-powered video generator, but this initial release focuses solely on image creation. The feature will be available across all ChatGPT subscription tiers, including Free, Plus, Pro, and Team. While the free tier's usage limit is the same as DALL-E, OpenAI spokesperson Taya Christianson revealed that the company didn't have a specific number to share, adding that these limits may change over time based on demand.

According to research lead Gabriel Goh, Sora's image generation capabilities are a significant improvement over previous models. The new system boasts enhanced "binding" capabilities, which refer to the AI's ability to maintain correct relationships between attributes and objects. Goh explained that most image models struggle with this, often mixing up colors and shapes when asked to render multiple items. In contrast, Sora's new image generation can correctly bind attributes for 15 to 20 objects without confusion, representing a substantial improvement in accuracy and reliability.

Another notable improvement is Sora's text rendering, which makes it easier to generate coherent text without typos on an image. Goh acknowledged that getting text rendering right was a significant challenge, requiring many months of iteration to achieve consistently usable results.

The system employs an autoregressive approach, generating images sequentially from left to right and top to bottom, similar to how text is written. This technical difference may be the key to Sora's better text rendering and binding capabilities, Goh speculated. In a briefing before the feature launch, the OpenAI team demonstrated several examples showcasing the system's capabilities, including scientific diagrams, multi-panel comics, and informational posters with accurate text.

ChatGPT multimodal product lead Jackie Shannon highlighted the practical applications of Sora, such as creating transparent background images for stickers, restaurant menus, and logos. Shannon emphasized that the model brings world knowledge to the equation, allowing users to generate images without explaining the context.

While the new system takes longer to generate images than before, OpenAI believes the improved quality and capabilities make up for the additional seconds users will spend waiting. The company also emphasized that the system includes robust safeguards to prevent misuse, including preventing watermark removal, blocking generation of sexual deepfakes, and refusing CSAM generation requests.

Notably, OpenAI's new image generation system does not include visual watermarks or indicators showing images are AI-generated. However, Shannon explained that all generated images will include standard C2PA metadata to mark the image as having been created by OpenAI, and the company will have internal tooling to look up images as well.

Ultimately, OpenAI's Sora integration into ChatGPT marks a significant milestone in the development of AI-powered image generation capabilities. As the technology continues to evolve, it will be interesting to see the innovative applications and use cases that emerge from this groundbreaking feature.