FitAD: A Framework for Enhancing YouTube Fitness Videos with Audio Description

Abstract

Audio descriptions (AD) are concise descriptions of visual elements that make videos accessible to blind or visually impaired (BVI) people [1]. The availability of videos with AD is limited because the process of adding AD is tedious and costly; thus, many BVI individuals find it difficult to consume video content independently [2]. In particular, one category of videos with limited accessible content is YouTube home fitness videos. These videos are often heavily demonstration-based with insufficient verbal direction. While many people follow YouTube workouts as a remote method of exercising, not all BVI individuals are able to enjoy the diverse range of free exercise content available. Thus, this thesis proposes, implements, and evaluates an audio description authoring framework called FitAD that enables users to enhance existing YouTube workout content with audio descriptions. Our approach expands on previous AD authoring frameworks and incorporates features tailored to exercise content, such as default descriptions for exercises and an audio timer countdown sound. A proof-of-concept, based on a JavaScript web-based application, demonstrates the viability of FitAD.