The growing demand for high-fidelity video generation from textual descriptions has catalyzed significant research in this field. In this work, we introduce MagicVideo-V2 that integrates the text-to-image model, video motion generator, reference image embedding module and frame interpolation module into an end-to-end video generation pipeline. Benefiting from these architecture designs, MagicVideo-V2 can generate an aesthetically pleasing, high-resolution video with remarkable fidelity and smoothness. It demonstrates superior performance over leading Text-to-Video systems such as Runway, Pika 1.0, Morph, Moon Valley and Stable Video Diffusion model via user evaluation at large scale.
@article{arxiv.2401.04468,
title = {MagicVideo-V2: Multi-Stage High-Aesthetic Video Generation},
author = {Weimin Wang and Jiawei Liu and Zhijie Lin and Jiangqiao Yan and Shuo Chen and Chetwin Low and Tuyen Hoang and Jie Wu and Jun Hao Liew and Hanshu Yan and Daquan Zhou and Jiashi Feng},
journal= {arXiv preprint arXiv:2401.04468},
year = {2024}
}