AI Course Video Generator

Research for:

"From Course Concept to Lecture Video: An AI-Powered System for Automated MOOC Development"

Authors:

Jacob Igo1, Huaiyuan Yao1, Wanpeng Xu1, Nadia Kellam1, Hua Wei1 1

Data Mining and Reinforcement Learning (DaRL Lab)

Arizona State University, Tempe, AZ, USA

Emails:

{jigo2, huaiyuan, wanpeng.xu, nadia.kellam, hua.wei}@asu.edu

Submitted and presented at ASU's LERN 2026 convening.

Built off of the foundational research paper:

Instructional Agents: LLM Agents on Automated Course Material Generation for Teaching Faculties

link here: https://arxiv.org/abs/2508.19611

This project automatically creates narrated course videos from AI-generated slides and scripts.

Given:

• A Markdown speaking script (.md)
• A LaTeX/PDF slide deck (.tex → .pdf)

it produces:

• Extracted slide images
• Cleaned frame-by-frame narration text
• Natural voice audio (gTTS or OpenAI TTS)
• A fully composed video with slides + synced narration

---

Technologies Used

• gTTS (moving toward OpenAI TTS in the future)
• pymupdf (splitting slide frames)
• MoviePy (for combining image and audio)
• FFmpeg (for clip concatenation)
• Regex (for complex parsing)