Google RealFill AI camera will expand and repair existing images

Google recently registered a trademark for a new AI photography feature called “RealFill”. This technology, which is being developed with researchers at Cornell University, allows users to expand and repair existing images using up to five reference images. According to the official description, RealFill will provide “non-downloadable software that uses artificial intelligence (AI) to draw images” and “non-downloadable online software for creating generative models.” This suggests that the feature will allow users to seamlessly edit and enhance their photos, potentially filling in missing details or repairing damaged areas.

Augmenting images with reference images

The technology behind RealFill is based on a research paper titled “Reference-Driven Generation for Authentic Image Completion,” which discusses how this AI model can use multiple reference images to expand and repair real-world scenes. By using these reference images, the system can generate more accurate and authentic-looking results compared to other image augmentation solutions.

Potential opportunities and limitations

Although RealFill appears to be a promising AI-based photography tool, it is not without its limitations. The current implementation is reported to be slower than some other image generation models, and recovering the final image from the generated scene can be challenging. In addition, the technology may have problems accurately processing text in images.

Google has a history of integrating advanced computational photography features into its Pixel smartphone lineup, such as the popular Night Sight mode. It’s possible that RealFill will be integrated into future Pixel devices or the Google Photos app. Therefore, it will allow users to easily enhance their images with this AI technology.

Conclusion

Google’s RealFill technology represents a significant advance in AI-powered photography, offering users the ability to enhance and retouch their images more authentically and seamlessly. While the current implementation has some limitations, the potential for this feature to be integrated into Google’s ecosystem and broader digital media authentication efforts suggest that RealFill could be a valuable tool for photographers and image creators alike.