JVT: Just Video Transformer makes pixel-space video diffusion simple
Paper โข 2611.18549 โข Published โข 15
This repository contains the checkpoints of JVT (Just Video Transformer), presented in JVT: Just Video Transformer makes pixel-space video diffusion simple.
JVT is a simple large-patch DiT-style Transformer for pixel-space video diffusion, without a pretrained autoencoder, auxiliary detailers, cascaded stages, or U-Net-style skip connections. It supports unconditional, class-conditioned, and camera-conditioned (RGB-disparity) video generation.
Project page: https://zhoutimemachine.github.io/JVT/