Asianfin--ByteDance's Volcano Engine launched an AI innovation tour in Shenzhen, during which it introduced two new video generation models: Doubao Video Generation-PixelDance and Doubao Video Generation-Seaweed. This marks ByteDance's official entry into the AI video generation market, targeting the enterprise sector for initial testing. According to Volcano Engine, the Doubao video generation models are based on the DiT architecture, utilizing efficient DiT fusion computing units that allow seamless transitions in dynamic video and camera movements. These models feature multi-camera capabilities, including zoom, panning, rotation, and target tracking. The deeply optimized Transformer structure significantly enhances the models' generalization capabilities, supporting various styles such as 3D animation, 2D animation, traditional Chinese painting, monochrome, and thick paint. They are designed to be compatible with various devices, including movies, televisions, computers, and mobile phones. Currently, the new Doubao video generation models are undergoing limited testing in the "Dream AI" internal beta, with plans for a gradual rollout to all users in the future.