Grace Jones and Eva Pflomm

Abstract:

Over the summer, we created a series of short educational videos in mathematics and physics using Manim, a Python-based animation library, with support from ChatGPT. The main project was a video on unit conversion, designed to highlight both the conceptual foundations and common pitfalls, such as confusing linear and squared units. With limited programming experience and no prior knowledge of Manim, we relied on ChatGPT to accelerate the process (drafting scene templates, suggesting animation structures, troubleshooting errors). This collaboration made it possible to translate teaching ideas into animated visuals much faster than we could have managed alone. At the same time, working with ChatGPT revealed important limitations. Its responses were highly sensitive to how prompts were framed, and some solutions were incomplete or overconfident, requiring careful review. To address this, we developed strategies such as providing ChatGPT with an “identity”, asking for inline comments, and testing small pieces of code before building larger scenes. These practices helped ensure both accuracy and clarity. The presentation will showcase the unit conversion video, outline the production workflow, and share lessons learned about combining AI support with human oversight—offering practical guidance for others interested in using Manim and AI tools in education.