Mito used Google Cloud to orchestrate 5,000+ AI assets for a TV spot

From short clips to a national campaign

Mito was founded to reduce the technical friction involved in creating long-form narrative stories with generative AI. The Movistar Plus+ campaign, produced for Telefónica, exposed the limits of workflows designed around individual short clips.

According to Mito, the production generated more than 5000 assets, including 1000 videos, across around 65 iteration rounds. Local media libraries were difficult to scale, while feedback often depended on downloads, Google Drive uploads, chat applications and email. Standard AI video rendering also took four to five minutes per short clip, which could block the user interface and slow creative work.

The team also had to coordinate different prompting requirements across generative models while maintaining characters, lighting and camera angles. The campaign included the likenesses of public figures such as chef Ferran Adrià and journalist Juanma Castaño.

A serverless system for asynchronous generation

Mito built its platform on Google Cloud, with its API, AI Director and background workers running as Java services on Cloud Run. Google Cloud Pub/Sub acts as the central event bus, separating job execution from the main interface. Long-running video generations are sent to background Cloud Run workers, so creators can continue using the canvas while those jobs run.

Media assets are stored in Google Cloud Storage and delivered through Cloud CDN. Mito says expiring key signatures keep pre-release client assets private, while keeping storage close to the models accessed through Vertex AI reduces network transfer delays. Cloud SQL and PostgreSQL support an isolated database architecture for business intelligence and analytics workloads.

The platform uses models including Omni, Veo and Nano Banana through Vertex AI, together with Mito's custom system prompts. Its structured canvas applies a "Rule of 6", limiting prompt variables to World, Camera, Light, Texture, Actions and Character. Mito also prioritizes still-frame generation before video rendering to improve cost efficiency.

Collaboration, testing and model coordination

The finished campaign aired on every major television channel in Spain. Mito reports that 1,441 assets were live on one shared Canvas near launch, allowing distributed teams and brand clients to work together in real time. New generations were saved as separate assets rather than replacing earlier versions, enabling parallel creative directions.

Mito's AI Director translates natural-language instructions into model parameters while preserving context, lighting and visual style between scenes. The company built a virtual queuing system to manage third-party API rate limits during usage spikes, using bounded retries and a defined result for every job.

Because creative output cannot be assessed fully through conventional unit tests, Mito created an automated evaluation harness powered by Gemini. The system evaluates AI Director outputs against users' original intent. The AI Director can also route tasks among specialized models, including Sonnet, Gemini and ElevenLabs for audio.

What the production approach means for creators

The campaign shows how Mito is positioning AI video production as a collaborative, multi-model workflow rather than a sequence of isolated clip generations. Its shared canvas, asynchronous processing and asset versioning are designed to let filmmakers and clients explore more ideas without interrupting production.

Mito says its architecture is prepared for rendering times to fall from minutes to seconds, as well as for future world models, deeper memory states and personalized content delivery. The company also says its internal studio, Mito Films, uses the tools on client projects so feedback can flow directly back into engineering.

The Movistar spot deliberately included an AI misunderstanding. In one scene, coach Luis de la Fuente says "In the same milkshake" before the prompt is changed live to "In the same heartbeat." Mito says this imperfect outcome, alongside surreal settings such as stadium stands built into Galician cliffs, was used to support human storytelling rather than obscure it.