ReCapture: Generative Video Camera Controls for
User-Provided Videos using Masked Video Fine-Tuning
CVPR Submission 1616
We present ReCapture, a method for generating new videos with novel camera trajectories from a single user-provided video. Our method allows us to re-generate the source video, with all its existing scene motion, from vastly different angles and with cinematic camera motion.
Gallery
The outcomes of various camera trajectories following ReCapture.
Zoom In w/ Rotation
Source Video
ReCapture
Source Video
ReCapture
Zoom Out w/ Rotation
Source Video
ReCapture
Source Video
ReCapture
Source Video
ReCapture
Source Video
ReCapture
Orbit Shot
Source Video
ReCapture
Source Video
ReCapture
Source Video
ReCapture
Source Video
ReCapture
Source Video
ReCapture
Source Video
ReCapture
Source Video
ReCapture
Source Video
ReCapture
Tilt Up
Source Video
ReCapture
Source Video
ReCapture
Tilt Down
Source Video
ReCapture
Source Video
ReCapture
Pan Left
Source Video
ReCapture
Source Video
ReCapture
Pan Right
Source Video
ReCapture
Source Video
ReCapture
Effectiveness of Masked Video Finetuning (Stage 2)
Our Masked Video Fine-tuning effectively eliminates strong artifacts, minimizes temporal inconsistencies, fills invalid regions in the anchor video, and regenerates a high-quality video.
Anchor Video
After Stage 2
Anchor Video
After Stage2
Detailed Ablation of All Components of Our Method.
Each component of our masked video fine-tuning contributes to performance improvements.