Hierarchical Spatio-temporal Decoupling for Text-to-Video generation
Huazhong University of Science and Technology
Alibaba Group Fudan University
Despite diffusion models having shown powerful abilities to generate photorealistic images, generating videos that are realistic and diverse still remains in its infancy. One of the key reasons is that current methods intertwine spatial content and temporal dynamics together, leading to a notably increased complexity of text-to-video generation (T2V). In this work, we propose HiGen, a diffusion model-based method that improves performance by decoupling the spatial and temporal factors of videos from two perspectives, i.e., structure level and content level. At the structure level, we decompose the T2V task into two steps, including spatial reasoning and temporal reasoning, using a unified denoiser. Specifically, we generate spatially coherent priors using text during spatial reasoning and then generate temporally coherent motions from these priors during temporal reasoning. At the content level, we extract two subtle cues from the content of the input video that can express motion and appearance changes, respectively. These two cues then guide the model's training for generating videos, enabling flexible content variations and enhancing temporal stability. Through the decoupled paradigm, HiGen can effectively reduce the complexity of this task and generate realistic videos with semantics accuracy and motion stability. Extensive experiments demonstrate the superior performance of HiGen over the state-of-the-art T2V methods. Our source codes and models will be found here.
A cute fire dog walks on the sea.
A cute fire dog jumps high on the sea.
A young man with a mohawk, piercings, and an energetic vibe., impressionist style.
A yellow robot.
A yellow robot grows a pair of yellow wings.
A Egyptian tomp hieroglyphics painting ofA regal lion, decked out in a jeweled crown, surveys his kingdom.
A Marvel movie painting ofA wise old owl in a tweed jacket puffs on a pipe.
A rabbit, low-poly game art style.
A robot girl.
An iron man surfing in the sea.
An astronaut dances in the desert.
Wooden figurine surfing on a surfboard in space.
A teddy bear wearing sunglasses playing guitar next to a cactus.
A dog wearing vr goggles on a boat.
A small bird sits atop a blooming flower stem
A van parked on grass in Disney movie style.
A bear dancing and jumping to upbeat music, moving his whole body.
Man riding scrambler motorbike on the highway through the autumn forest.