How We Generated 7 Characters in 2 Cultures for Kellogg's in 11 Days
Inside the AI production pipeline that delivered a multi-character, bilingual FMCG campaign in under two weeks — and what it means for the future of brand work.
11
Days end-to-end
7
AI characters
2
Cultural versions
The Brief
A global FMCG icon, two distinct markets, one tight window.
Kellogg's came to us with a familiar but high-stakes challenge: launch a campaign featuring multiple iconic brand characters, adapted for two cultural contexts, and ship it in time for a fixed retail moment. Traditional production — casting, sets, costume, post — would have taken months we didn't have.
The mandate was simple: make it look like a studio production. Don't let the AI show. And don't compromise on the characters that millions of consumers already recognise.
The Approach
Three parallel pipelines, one creative spine.
We split the work into three concurrent tracks: character development, environment generation, and cultural adaptation. Each ran on its own AI stack but fed into a single creative direction document we kept obsessive control over.
The character pipeline focused on consistency — the same mascot needed to appear across multiple shots, angles, and emotional beats without drifting. We built reference grids, locked seeds, and built a custom review loop with the client so revisions happened in hours, not weeks.
The Pipeline
Reference grids, locked seeds, and obsessive QA.
Character consistency in AI is the hardest problem in the room. Our solution was a layered approach: a master reference sheet for each character, frozen generation parameters, and a human art-director sign-off on every hero frame before it moved into motion.
For environments, we built two parallel mood boards — one per cultural version — and used them as the visual constraint for every scene. The characters stayed identical; the worlds around them shifted authentically.
The Result
Studio-grade output, 90% faster.
Eleven days from kickoff to delivered masters. Seven characters, two full cultural versions, and a campaign that the client's regional teams could deploy across markets without a re-shoot. The work didn't just hit the window — it set a new internal benchmark for what AI-native production can look like at FMCG scale.
The bigger lesson: AI doesn't replace the creative process. It compresses it. The discipline, the taste, the brand stewardship — all of that matters more, not less, when production friction disappears.
What We'd Tell Other Brands
Three things to know before starting.
- Lock your characters first. Don't start motion work until your character system is bulletproof. Inconsistency is the only thing that breaks the illusion.
- Plan for parallel pipelines. AI lets you run tracks concurrently that traditional production runs sequentially. Resource accordingly.
- Keep the human review tight. The speed advantage evaporates if approvals take a week. Build a fast review cadence with the client from day one.