Kids story generator — an agentic pipeline for children's stories

Problem

Producing a polished children’s story normally needs a writer, an illustrator, and a narrator — slow and expensive when the goal is a growing content catalogue. We needed a repeatable pipeline that could take a simple story idea and produce a finished, illustrated, narrated story without a production team behind every release.

Approach

We built the generator as an agentic framework on Google ADK with Gemini 2.5 Flash: a coordinated multi-agent pipeline where each specialised agent handles one stage of production.

The pipeline runs in four stages. A story generation agent drafts the text and breaks it into pages. An illustration agent creates matching artwork for each page. A narration agent produces audio — text-to-speech voiced and timed to the story flow. A coordination agent ties the stages together, passing each output through review before the next step begins, with everything persisted to SQLite so the pipeline is resumable and auditable.

The whole system starts from a single prompt — a story idea and a target age range — and returns a complete illustrated storybook ready for the LifuKids content catalogue.

Result

One prompt in, a finished children’s story out: text, illustrations, and narration packaged and ready for distribution. The generator feeds directly into the LifuKids content pipeline for YouTube and Spotify, letting us grow the catalogue steadily without scaling a creative team.

The same multi-agent pattern — specialised agents coordinated through a central framework — is reusable for client content-automation work. We built it once; the architecture adapts to any domain where a pipeline needs drafting, review, generation, and assembly stages.

Links

• LifuKids