Meta releases innovative generation model "Pippo" to achieve high-quality video generation in a single photo

Meta Reality Labs' research team recently released a generative model called "Pippo", an innovative technology that can generate multi-view videos at up to 1K resolution from a normal photo, marking a major breakthrough in computer vision and image generation. .

The core of the Pippo model lies in its multi-view diffusion converter design. Unlike traditional generative models, users only need to provide a photo taken, and the system can automatically generate physical and dynamic video effects without additional camera parameters or fitting the model. .

Currently, Pippo is released in code version and does not include pre-training weights. The research team provides complete models, configuration files and training code that developers can easily train and apply. Future plans include organizing code and launching inference scripts for pre-trained models to further optimize the user experience.

For more information, please visit the Pippo project page .