How we produce LifuKids learning content for YouTube

Alongside the web platform, LifuKids content lives on YouTube: story-based lessons with colourful illustrations and gentle narration designed for young children. This post walks through how we produce that content — and why we built the pipeline the way we did.

What we publish

The LifuKids YouTube catalogue covers learning stories, counting and alphabet lessons, and topic-based videos around animals, vehicles, and everyday concepts. Every piece starts from an idea — “a story about sharing for 4-year-olds,” “a counting video with farm animals” — and ends as a finished, illustrated, narrated video ready for YouTube.

We keep the production values consistent: bright, age-appropriate illustrations, clear and warm narration, simple animated transitions. The goal is content children actually want to watch and parents feel good about.

The pipeline

What makes this pipeline unusual is that it is largely agentic. Our kids story generator — an agentic framework built on Google ADK with Gemini 2.5 Flash — handles the heavy production work across three stages:

Story generation. One agent drafts the text from a topic prompt and target age range, breaking it into pages or scenes. The output is a structured script, not freeform prose — each page has a visual description and narration text.

Illustration generation. A second agent takes the visual descriptions and generates matching artwork: characters, backgrounds, scene compositions. The illustrations follow a consistent style so episodes feel like they belong to the same world.

Audio production. A third agent produces the narration — text-to-speech voiced and timed to the story flow — ready to sync with the visuals in editing.

Every stage includes human review. An agent never publishes unsupervised. But the pipeline removes the bottleneck: the human reviews and approves, rather than creating from scratch.

Why agentic

We chose the agentic approach for two reasons. First, it is repeatable — once the pipeline is dialled in, producing the next episode takes a fraction of the effort of the first. Second, it is composable — the same pipeline architecture (specialised agents, staged review, structured output handoff) transfers to other content domains without rebuilding from scratch.

Growing the catalogue

The catalogue is still young, but the pipeline lets us grow it steadily without scaling a creative team. Each new topic idea feeds through the same stages, and the output consistency improves as we tune the agents.

If you are building a content pipeline — for YouTube, for a learning platform, or for any domain where volume and consistency matter — the same agentic pattern applies. Talk to us about what that could look like for your organisation.