Learned Spatial Representations for Few-shot Talking-Head Synthesis

TL;DR

This paper introduces a new method for few-shot talking-head synthesis that disentangles spatial and style representations, leading to more accurate identity preservation and improved image quality.

Contribution

It proposes a novel factorization of subject representation into spatial and style components, enhancing identity preservation in talking-head synthesis.

Findings

Significant improvement in identity preservation over previous methods

Quantitative and qualitative enhancements in synthesized image quality

Effective disentanglement of spatial and style features

Abstract

We propose a novel approach for few-shot talking-head synthesis. While recent works in neural talking heads have produced promising results, they can still produce images that do not preserve the identity of the subject in source images. We posit this is a result of the entangled representation of each subject in a single latent code that models 3D shape information, identity cues, colors, lighting and even background details. In contrast, we propose to factorize the representation of a subject into its spatial and style components. Our method generates a target frame in two steps. First, it predicts a dense spatial layout for the target image. Second, an image generator utilizes the predicted layout for spatial denormalization and synthesizes the target frame. We experimentally show that this disentangled representation leads to a significant improvement over previous methods, both quantitatively and qualitatively.