The Blueprint: Orchestrating 5,000+ AI assets for a World Cup TV spot

Mito, an AI-powered video creation platform, faced bottlenecks when scaling a single five-second clip into a 120-second national TV spot for Movistar Plus+ during the World Cup. The project required generating over 5,000 assets, including 1,000 videos, across 65 iteration rounds, revealing unsustainable workflows for managing media libraries and feedback loops. Traditional methods like manual downloads and fragmented communication slowed progress, while AI rendering delays of four to five minutes per clip disrupted creative workflows and froze interfaces.

To overcome these challenges, Mito adopted a serverless, asynchronous architecture on Google Cloud, running services on Cloud Run and decoupling job execution with Google Cloud Pub/Sub. Long-running video generations were offloaded to background workers, allowing creators to continue working uninterrupted. Media assets were stored in Google Cloud Storage and delivered via Cloud CDN, with expiring key signatures ensuring pre-release content remained private.

The platform leveraged foundational models like Omni, Veo, and Nano Banana through Vertex AI, alongside a structured frontend enforcing a 'Rule of 6' to limit prompt variables and prioritize still-frame generation for cost efficiency. A conversational AI Director translated natural language instructions into precise model parameters, maintaining consistency in lighting, style, and visuals across scenes.

The Movistar campaign aired across Spain’s major television channels, with 1,441 assets live on a shared canvas near launch. Mito’s internal studio, Mito Films, continuously tests tools on high-stakes projects, ensuring direct feedback informs engineering improvements. The architecture is designed to reduce rendering times from minutes to seconds, supporting future advancements in AI models and personalized content delivery.