Body

Full-Body Generation

Full-body Generation synthesizes body movements through 3D modeling using just a single video within 10 seconds.

By training a deep learning model on a person's full-body video, this technology can generate body shapes and alter movements by changing body parameters.

This technology is currently in the development stage and is expected to be commercialized in the near future.

Result

Our Full-Body Generation technology utilizes 3D-based avatar generation techniques to create 3D avatars using just a single video captured with a mobile phone, typically within 10 seconds.

Leveraging our unique deep learning lightweighting techniques, it achieves more than 10 times faster inference speed compared to existing models

We have developed a generation model that allows the features of diverse images to be encoded into separate vector spaces.

Additionally, we have created an image encoder corresponding to the generation model to accurately embed given images into latent vectors.