WorldDreamer is an innovative video generation model that understands and simulates dynamic world scenes by predicting obscured visual tokens. It excels in various areas, including image-to-video synthesis, text-to-video generation, video repair, video stylization, and motion-to-video generation. Building upon the success of large language models, the model treats world modeling as an unsupervised visual sequence modeling challenge. It achieves this by mapping visual inputs to discrete tokens and predicting the obscured ones.