Introduction

Generative video is often demonstrated through impressive examples, but educational use places different demands on the tools: clarity, correctness, pacing, and cognitive load.

This page exists to:

• Test generative image–video tools against concrete teaching needs
• Record what worked, what didn’t, and why
• Make trade-offs and limitations explicit rather than implicit

Initial experiment: Explaining PCA visually

Why PCA?

I’ve personally struggled to develop an intuitive understanding of eigenvectors and Principal Component Analysis since my undergraduate days. That experience made PCA a particularly motivating test case for exploring whether short, narrative visual explanations could help bridge the gap between mathematical formalism and intuition.

Principal Component Analysis (PCA) is also a concept that many students find abstract because:

• the intuition is geometric.
• the math (covariance, eigenvectors) comes early
• time and motion are hard to convey in static equation filled slides!

This made PCA a good candidate to test whether image–video generation can support conceptual understanding.

What I tried:

• Image-first workflows to establish clean geometric states using NanoBananaPro.
• Image → video transitions for controlled motion using Veo3.1
• Short, chunked clips instead of a single long explanation
• Narration used sparingly to add voiceover and enhance understanding.