Third Time's the Charm? Image and Video Editing with StyleGAN3
This work addresses challenges in generative model editing for researchers and practitioners, but it is incremental, building on existing StyleGAN architectures.
The paper analyzes StyleGAN3 for image and video editing, showing that aligned data training doesn't hinder unaligned generation, W/W+ spaces are more entangled than in StyleGAN2, and proposes an encoder for unaligned images and a video editing workflow to reduce texture sticking and expand field of view.
StyleGAN is arguably one of the most intriguing and well-studied generative models, demonstrating impressive performance in image generation, inversion, and manipulation. In this work, we explore the recent StyleGAN3 architecture, compare it to its predecessor, and investigate its unique advantages, as well as drawbacks. In particular, we demonstrate that while StyleGAN3 can be trained on unaligned data, one can still use aligned data for training, without hindering the ability to generate unaligned imagery. Next, our analysis of the disentanglement of the different latent spaces of StyleGAN3 indicates that the commonly used W/W+ spaces are more entangled than their StyleGAN2 counterparts, underscoring the benefits of using the StyleSpace for fine-grained editing. Considering image inversion, we observe that existing encoder-based techniques struggle when trained on unaligned data. We therefore propose an encoding scheme trained solely on aligned data, yet can still invert unaligned images. Finally, we introduce a novel video inversion and editing workflow that leverages the capabilities of a fine-tuned StyleGAN3 generator to reduce texture sticking and expand the field of view of the edited video.