Beyond End-to-End Video Models: An LLM-Based Multi-Agent System for Educational Video Generation

TL;DR

We address the inadequacy of end-to-end video models for educational content by introducing LASEV, a hierarchical LLM‑based multi-agent system that generates educational videos via Executable Video Scripts. The Orchestrating Agent coordinates three domain‑specific agents—Solution, Illustration, and Narration—while a heterogeneous critique loop ensures semantic correctness, executable validity, and structural compliance, culminating in deterministic video synthesis from a structured script $\\mathcal{S}=(\\mathcal{P},\\mathcal{N},\\mathcal{A})$. Outputs are compiled through a deterministic pipeline $\\mathcal{V}= \\texttt{Render}_{\\text{vis}}(\\mathcal{P},\\mathcal{A}) \\parallel \\texttt{Synth}_{\\text{audio}}(\\mathcal{N})$, enabling scalable production of instructional videos with high quality and low cost. Experiments on Elementary Chinese Language Arts and Middle School Mathematics show that LASEV achieves near‑perfect publishable quality and up to ~1M videos/day throughput with ~95% cost reduction, validating both effectiveness and practicality for large‑scale educational deployment. The work demonstrates that transforming video generation into structured script synthesis and automated verification can reconcile pedagogical rigor with scalability.

Abstract

Although recent end-to-end video generation models demonstrate impressive performance in visually oriented content creation, they remain limited in scenarios that require strict logical rigor and precise knowledge representation, such as instructional and educational media. To address this problem, we propose LAVES, a hierarchical LLM-based multi-agent system for generating high-quality instructional videos from educational problems. The LAVES formulates educational video generation as a multi-objective task that simultaneously demands correct step-by-step reasoning, pedagogically coherent narration, semantically faithful visual demonstrations, and precise audio--visual alignment. To address the limitations of prior approaches--including low procedural fidelity, high production cost, and limited controllability--LAVES decomposes the generation workflow into specialized agents coordinated by a central Orchestrating Agent with explicit quality gates and iterative critique mechanisms. Specifically, the Orchestrating Agent supervises a Solution Agent for rigorous problem solving, an Illustration Agent that produces executable visualization codes, and a Narration Agent for learner-oriented instructional scripts. In addition, all outputs from the working agents are subject to semantic critique, rule-based constraints, and tool-based compilation checks. Rather than directly synthesizing pixels, the system constructs a structured executable video script that is deterministically compiled into synchronized visuals and narration using template-driven assembly rules, enabling fully automated end-to-end production without manual editing. In large-scale deployments, LAVES achieves a throughput exceeding one million videos per day, delivering over a 95% reduction in cost compared to current industry-standard approaches while maintaining a high acceptance rate.

Figures (2)

• Figure 1: Overview of the LASEV framework. The Orchestrating Agent coordinates three specialized working agents--Solution Agent, Illustration Agent and Narration Agent, through iterative quality critique and template-based video assembly. The educational videos are thus rendered from assembled executable video scripts (EVS).
• Figure 2: Several positive and negative cases produced by LASEV across real Chinese Language Arts and Mathematics problems.