Developed and integrated the Video All-in-One Creation Engine (VACE) into the vllm-project/vllm-omni repository, enabling conditional video generation and supporting new text-to-video and image-to-video models for WAN 2.1. Leveraged Python, deep learning, and machine learning to enhance the video processing pipeline, accelerating content creation workflows. The work focused on a single, well-scoped feature, with code quality and collaboration demonstrated through comprehensive, signed-off commits and co-authorship. No major bugs were reported during this period, and efforts shifted toward stabilization and production readiness, laying the groundwork for future deployment and expanded utilization of advanced video generation capabilities within the platform.