DynamicLecture

Abstract

Producing high-quality lecture videos requires multiple rounds of edits to the original recording, for example, fixing errors, revising content, or incorporating feedback. However, instructors often lack the time and skills to apply the edits to the videos. To facilitate easier editing of lecture videos, we developed DynamicLecture, an object-oriented video editing interface that allows users to directly update the content of slide-based lectures within the video. The editing interface is powered by a computational pipeline which extracts slides, texts, images, and transcribed speech from the raw video. In a user study, students and instructors identified a list of items to edit from the instructors’ lecture video. Using DynamicLecture, instructors successfully edited 31% of the items and marked an additional 27% of the items as achievable with more time. Both instructors and professional video editors could edit lecture videos faster and easier using DynamicLecture compared to existing tools.