More than 5,000 assets and 65 rounds of iterations ultimately became a 120-second World Cup television commercial. For Mito, this demonstrated that the bottleneck in AI video is no longer merely generating images, but organizing many models, versions, and team feedback into a deliverable project.
This case comes from Google Cloud customer Mito, which produced a World Cup advertisement for Movistar Plus+, part of Spanish telecommunications company Telefónica. Generative AI (using models to automatically produce text, images, or video content) has entered advertising production, but the more assets it generates, the more complicated collaboration and delivery become.
What this is
Mito connected cloud services, background task queues, asset storage, and content distribution. Each AI video clip takes roughly four to five minutes to render, so the system runs time-consuming tasks in the background while the interface remains usable. The 5,000-plus assets included 1,000 videos and went through approximately 65 rounds of revisions.
Unified orchestration also solved another practical problem: different models require different prompts and parameters, while character appearance, lighting, and camera angles must remain consistent across hundreds of shots. Public figures’ likenesses were also brought under asset management, rather than repeatedly left to prompt-based trial and error.
Industry view
We believe this shows that AI video companies are shifting their product focus from “generating more realistic results” to “delivering projects on time.” Generation models can be outsourced to multiple providers, but versions, permissions, feedback, and rendering queues must be controlled internally. Otherwise, teams get trapped moving large numbers of files around.
The risks are equally clear: Google Cloud provides both infrastructure and model services, and Mito has demonstrated a smooth path, not a guarantee that every company can replicate it at low cost. Pay-as-you-go model APIs are also exposed to rate limits, regional throughput constraints, and price fluctuations. And the more assets produced, the harder it becomes to manage storage, copyright, and liability related to people’s likenesses.
Impact on regular people
For enterprise IT: AI applications increasingly require unified task orchestration, asset permissions, and cost monitoring. IT departments must now manage auditable production systems, not just office software.
For individual careers: Video, e-commerce, and professional services teams will gradually delegate storyboards, test renders, and version revisions to tools. Process design and human judgment will become more important capabilities.
For consumer markets: Production cycles for advertisements and short videos may shorten, increasing the volume of customized content. But generating more does not mean producing better work; brand consistency, licensing boundaries, and final quality will still determine commercial value.