Youtube examples

[matanox]

Thanks for this great research and implementation!

I am curious, whether there are any examples out there for video style transfers involving people speaking to a camera, that you would know of. Also, in a nutshell, how long it would take to process a video on a given example hardware. Looking into using this implementation for videos of people filmed speaking to a camera, which is a bit different than most of the examples I could eye out there.

Thanks in advance for your comments!

[manuelruder]

The closest we have published is this scene: https://www.youtube.com/watch?v=SKql5wkWz8E&t=4m28s

Processing videos are quite fast, a second per frame, given there is a pretrained model. We provide some in this repo, see the description. For arbitrary styles you either need to train a new model (takes up to 48h on a mid range GPU) or use "artistic-videos": https://github.com/manuelruder/artistic-videos That need several minutes per frame though.