Mora is a research framework that enables generalist video generation through a multi-agent approach using collaborative visual agents. It aims to replicate and extend the video generation capabilities demonstrated by OpenAI’s Sora, producing videos at resolutions like 1024×576 with durations exceeding 12 seconds. The system coordinates multiple specialized visual agents to handle different aspects of the video generation pipeline.