CourseBlueprint pipeline boosts educational video engagement 4x with structured pedagogical contracts

CourseBlueprint addresses a critical gap in generative text-to-video systems: they produce visually fluent clips but lack pedagogical content knowledge (PCK) for effective instruction. The pipeline operates over a course corpus (BMED 2300) and generates a structured teaching blueprint in a single forward pass. It includes a scaffolding module that builds a stage-labeled prerequisite concept graph with deterministic cycle removal, an adaptive controller that assigns per-concept style specifications, and an engagement generator that produces narration following a fixed hook->retrieval->core->analogy->forward contract. A deterministic slide-image override grounds the video by reusing instructor slides when retrieval confidence is high.

In a five-topic ablation study, removing the engagement contract caused the engagement score to drop from 5.00 to 1.20, the adaptive score from 4.80 to 3.40, and Flesch readability from 38.0 to 19.8. Analogy and retrieval-prompt counts fell to near zero. The slide-image override converted a 0/9 corpus-grounding failure into 9/10 successful slide matches on the same topic. These results demonstrate that pedagogical video quality depends less on surface fluency than on explicit, typed instructional contracts that make scaffolding, adaptation, engagement, and grounding auditable. The researchers also released a reusable benchmark corpus and an evaluation harness combining LLM-judge scoring with regex-grounded metrics.