Google Research Builds a Multi-Agent AI Video Co-Director

Google Research introduced a multi-agent framework for generating longer AI video narratives while reducing identity drift, continuity problems and cascading errors between shots. The system uses Gemini and Veo with planning and visual-refinement agents that treat long-form generation as a broader optimization problem rather than a chain of isolated prompts. The work is research-focused, but it points toward more automated workflows for creators producing multi-shot video. #GoogleAI #GenerativeAI #VideoAI