Human-supervised AI video production pipeline for personalized instructional videos

Design · All designs
Evidence · 2 claims (1 for, 1 against) · 1 study (1 associational), q1 · 0 of 1 report an effect size · 2 claims rest on one study

Description

A custom software pipeline produced 300 AI-generated personalized videos (100 topics × three interest variants: business, technology, society/biology) for a large online undergraduate course. LLMs (GPT-4o, Claude 3.7 Sonnet) drafted five-paragraph scripts from an instructor-defined "ethos" prompt; the instructor and project scientist reviewed and edited all scripts and checked examples for correctness. Videos composited a HeyGen AI avatar built from a recording of the instructor, Unsplash images, bullet points, titles, credits, and music into 3–5 minute videos uploaded to YouTube and embedded in Canvas pages.

Design Implications

Context

Requirements

• Instructor oversight at multiple stages: script review, correctness checking of examples and images, and final video review

Constraints

• Producing and reviewing the scripts "was a time-consuming process, but still quicker than it would have been to write multiple versions of 100 scripts by hand"; numerous iterations were needed to adjust avatar expressiveness, intonation, lighting, and framing

Target Learners

• undergraduate students in a large online general education course

Learning Goals

• connections between information technology and global sustainability

Claims

• Students preferred AI-generated personalized videos over human-recorded non-personalized videos in direct comparison [+W]
• Viewing of the personalized AI videos declined sharply across the quarter, with views of the first week-9 video set suggesting fewer than 10% of the course population was watching voluntarily by the end of the course [-W]

Related Designs

• Semi-automated three-stage human-in-the-loop pipeline for generating dialogue-based TTS lesson videos
• Bespoke: a five-stage pipeline that regenerates a complete industry-personalized lecture video from a seed transcript

Examples

-

Key Sources

• Tomlinson, B., Black, R.W., Patterson, D.J., van der Hoek, A., Ferguson, J. & Bietz, M.J. (2026). Students prefer personalized, AI-generated educational videos over non-personalized, human-recorded videos. Scientific Reports 16:21804. https://doi.org/10.1038/s41598-026-52798-9

Page metadata

Field	Value
type	design
id	ai-personalized-video-production-pipeline
title	Human-supervised AI video production pipeline for personalized instructional videos
description	A custom software pipeline produced 300 AI-generated personalized videos (100 topics × three interest variants: business, technology, society/biology) for a large online undergraduate course.
status	review
generated	by process:settle-candidates on 2026-10-11
trust	unverified — no one has confirmed this page's content
sources	1 — tomlinson-2026

Frontmatter is the page's machine-readable record — see the Schema & Guide.