TL;DR: Given a portrait and a reference image specifying target attributes,
Durian generates an animatable 2D portrait video
with composable visual attributes in a single-stage pipeline.
Results
Note that all result videos are complete zero-shot inference results that are not included in the training set, without any additional training/optimization.
We present results of transferring attributes to the portrait while simultaneously animating it according to the facial keypoint video.
Hair Transfer
Hat Transfer
Beard Transfer
Eyeglasses Transfer
Applications
All result videos are complete zero-shot inferences, not included in the training set and without any additional training/optimization,
showing the transfer of multiple attributes simultaneously to the portrait or interpolation between attributes.