Reangle-A-Video: 4D Video Generation as Video-to-Video Translation
Anonymous authors
Press Enter to pause or resume all videos.
TL;DR
We introduce Reangle-A-Video, a unified framework for generating synchronized multi-view videos from a single input video.
We achieve (a) Static view transport and (b) Dynamic camera control without relying on any multi-view generative priors, where both approaches offer six degrees of freedom.
(a) Static view transport: Regeneration of the input video shooting from target viewpoints.
Input video
View from Orbit right
View from Orbit down
View from Orbit left
View from Dolly zoom in
View from Dolly zoom out
(b) Dynamic camera control: Regeneration of the input video following target camera movements.
Input video
Camera movement: Orbit up
Camera movement: Orbit down
Camera movement: Orbit left
Camera movement: Orbit right
Camera movement: Dolly zoom in
Camera Visualizations
We demonstrate six degrees of freedom in both (a) Static view transport and (b) Dynamic camera control.
Here we visualize the transported viewpoints and camera movements used in our work.
Gallery: (a) Static view transport
Input video
View from Orbit left
View from Orbit right
View from Orbit down
View from Dolly zoom in
Input video
View from Orbit left
View from Orbit right
View from Orbit up
View from Orbit down
View from Dolly zoom in
Input video
View from Orbit left
View from Orbit right
View from Orbit up
View from Dolly zoom in
Input video
View from Orbit left
View from Orbit down
View from Orbit up
Input video
View from Orbit left
View from Orbit down
View from Dolly zoom in
Input video
View from Orbit right
View from Orbit down
View from Orbit up
Input video
View from Orbit left
View from Orbit down
View from Dolly zoom in
Gallery: (b) Dynamic camera control
Input video
Camera movement: Orbit down
Camera movement: Orbit up
Camera movement: Orbit left
Camera movement: Dolly zoom in
Camera movement: Dolly zoom out
Input video
Camera movement: Orbit left
Camera movement: Dolly zoom in
Camera movement: Orbit down
Camera movement: Orbit up
Input video
Camera movement: Dolly zoom in
Camera movement: Dolly zoom out
Camera movement: Orbit left
Camera movement: Orbit right
Camera movement: Orbit down
Input video
Camera movement: Orbit right
Camera movement: Dolly zoom out
Input video
Camera movement: Orbit left
Camera movement: Orbit up
Camera movement: Dolly zoom in
Input video
Camera movement: Orbit right
Camera movement: Orbit up
Camera movement: Dolly zoom in
Input video
Camera movement: Dolly zoom in
Input video
Camera movement: Orbit down
Input video
Camera movement: Orbit down
Input video
Camera movement: Dolly zoom in
Comparisons (1/2)
For (a) Static view transport, we compare with Generative Camera Dolly, and Vanilla CogVideoX I2V which employs the same input frame as ours.
Input video
View from Orbit left Ours
View from Orbit down Ours
Input video
Generative Camera Dolly
Generative Camera Dolly
Input video
Vanilla CogVideoX I2V
Vanilla CogVideoX I2V
Input video
View from Orbit down Ours
View from Dolly zoom in Ours
Input video
Generative Camera Dolly
Generative Camera Dolly
Input video
Vanilla CogVideoX I2V
Vanilla CogVideoX I2V
Input video
View from Orbit up Ours
View from Orbit right Ours
Input video
Generative Camera Dolly
Generative Camera Dolly
Input video
Vanilla CogVideoX I2V
Vanilla CogVideoX I2V
Comparisons (2/2)
For (b) Dynamic camera control, we compare with NVS-Solver and Trajectory Attention.
