Skip to content
No description, website, or topics provided.
Branch: master
Clone or download
Pull request Compare This branch is even with NVlabs:master.
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Type Name Latest commit message Commit time
Failed to load latest commit information.

Few-shot vid2vid

Project | YouTube | arXiv

Pytorch implementation for few-shot photorealistic video-to-video translation. It can be used for generating human motions from poses, synthesizing people talking from edge maps, or turning semantic label maps into photo-realistic videos. The core of video-to-video translation is image-to-image translation. Some of our work in that space can be found in pix2pixHD and SPADE.

Few-shot Video-to-Video Synthesis
Ting-Chun Wang, Ming-Yu Liu, Andrew Tao, Guilin Liu, Jan Kautz, Bryan Catanzaro
NVIDIA Corporation
In Neural Information Processing Systems (NeurIPS) 2019

Example Results

  • Dance Videos

  • Talking Head Videos

  • Street View Videos

Code Coming Soon

You can’t perform that action at this time.