Implicit Warping for Animation with Image Sets
Supplementary videos

TalkingHead1KH results
Driving video reconstruction with single source image
Source image Driving video FOMM fv2v Ours
Source image Driving video FOMM fv2v Ours


Driving video reconstruction with multiple source images
Source images Driving video FOMM fv2v Ours
Source images Driving video FOMM fv2v Ours


Cross-identity motion transfer
Source image(s) Driving video FOMM fv2v Ours
Source image(s) Driving video FOMM fv2v Ours


Keypoint location and strength visualization

TED Talk results
Driving video reconstruction with single source
Source image Driving video FOMM AA-PCA Ours
Source image Driving video FOMM AA-PCA Ours


Driving video reconstruction with multiple source images
Source images Driving video Ours (single source) FOMM Ours
Source images Driving video Ours (single source) FOMM Ours


Cross-identity motion transfer
Source images Driving video Ours (single source) FOMM Ours
Source images Driving video Ours (single source) FOMM Ours


Keypoint location and strength visualization

Attention visualizations
We dilate the attention maps with a square kernel of size 14x14 and only keep the regions with strength > 0.5.
These surviving regions of attention are visualized in the videos below.