Recent advances in artifcial intelligence (AI) have led to an increased focus on automating media production. One relevant application area for AI is using speech recognition to create subtitles and closed captions for videos. The AI methods based on machine learning are still not sufciently reliable in terms of producing perfect or acceptable subtitles. To compensate for this unreliability, AI can be used to build tools that support, rather than replace, human eforts and to create semi-automated workfows. In this paper, we present a prototype for including automated speech recognition for subtitling in an existing production-grade video editing tool. We devised an experiment with 25 participants and tested the efciency and efectiveness of this tool compared to a fully manual process. The results show that there is a signifcant increase in both efectiveness and efciency for novices in subtitling. Furthermore, the participants found the augmented process to be more demanding. We identify some usability issues and design choices that pertain to making augmented subtitling easier.
CCS CONCEPTS• Computing methodologies → Machine learning; • Computer systems organization → Embedded and cyber-physical systems.