VividFace: A Diffusion-Based Hybrid Framework for High-Fidelity Video Face Swapping

Supplementary material

Original videos shown in the mauiscript

Original videos shown in Figure 1

Source
Target
VividFace

Original videos shown in Figure 6

Source
Target
VividFace
DiffSwap
FSGAN
REFace
SimSwap
Source
Target
VividFace
DiffSwap
FSGAN
REFace
SimSwap
Source
Target
VividFace
DiffSwap
FSGAN
REFace
SimSwap
Source
Target
VividFace
DiffSwap
FSGAN
REFace
SimSwap

Original videos shown in Figure 7

Source
Target
Texture: 0.2, Attribute: 0.2
Texture: 0.2, Attribute: 0.4
Texture: 0.2, Attribute: 0.6
Texture: 0.2, Attribute: 0.8
Texture: 0.2, Attribute: 1.0
Texture: 0.4, Attribute: 0.2
Texture: 0.4, Attribute: 0.4
Texture: 0.4, Attribute: 0.6
Texture: 0.4, Attribute: 0.8
Texture: 0.4, Attribute: 1.0
Texture: 0.6, Attribute: 0.2
Texture: 0.6, Attribute: 0.4
Texture: 0.6, Attribute: 0.6
Texture: 0.6, Attribute: 0.8
Texture: 0.6, Attribute: 1.0
Texture: 0.8, Attribute: 0.2
Texture: 0.8, Attribute: 0.4
Texture: 0.8, Attribute: 0.6
Texture: 0.8, Attribute: 0.8
Texture: 0.8, Attribute: 1.0
Texture: 1.0, Attribute: 0.2
Texture: 1.0, Attribute: 0.4
Texture: 1.0, Attribute: 0.6
Texture: 1.0, Attribute: 0.8
Texture: 1.0, Attribute: 1.0

Original videos shown in Figure 8

Source
Target
Without Occlusion Augmentation
With Occlusion Augmentation
Source
Target
Without 3D Reconstruction
With 3D Reconstruction