David CHEW
Department of Statistics and Data Science, Faculty of Science (FOS)
Chew, D. (2023). Creating teaching videos using AI-generated voices [Paper presentation]. In Higher Education Campus Conference (HECC) 2023, 7 December, National University of Singapore. https://blog.nus.edu.sg/hecc2023proceedings/creating-teaching-videos-using-ai-generated-voices/
SUB-THEME
AI and Education
KEYWORDS
Technology-enhanced learning, AI voices, videos, blended learning, transferability
CATEGORY
Paper Presentation
ABSTRACT
The advent of artificial intelligence (AI) presents a remarkable opportunity for various industries, and education is no exception. Within the realm of educational technology, a promising opportunity has emerged with the use of AI voices to create teaching videos. This innovative approach harnesses the power of AI to enhance educational content and delivery methods, revolutionising the way knowledge is imparted to learners.
In this talk, I describe an effort to make use of AI-generated voices to create teaching videos for the course ST2334 “Probability and Statistics”. ST2334 has an enrolment of 800 students every semester and is offered in a blended learning manner. Each week, students view at their own time 30 to 40 minutes worth of pre-recorded videos, before attending a “live” lecture delivered by the course coordinator. As the course is taught by different faculty members in different semesters, it was decided that the pre-recorded videos will be made with a “neutral” voice. An AI voice software Descript was then used to create the pre-recorded videos.
There are several ways you can use Descript.
(A) Use it as a video recorder cum editor
- Record your teaching videos using your own voice.
- Import the videos into Descript. Voice narrations will be automatically transcribed into text and aligned automatically to the audio. It is then easy to edit your videos in a word processor-like environment (Figure 1). Instead of working with sound waves (as with many other video editing software), the user can work on the script/words directly. Deleting words will automatically remove the associated video footage.
- If you like to replace (the audio of) a mispronounced or wrong choice of word, it is possible to select that word, correct it and have that word replaced using a trained AI voice that sounds exactly like you.
- Annotations/animations can be timed to coincide with text easily.
(B) Use it to construct your videos from scratch using an AI voice
- Import your slides/videos into Descript.
- Overlay the slides/videos with AI voices by typing out a script.
- You may use (i) a stock AI voice, or (ii) train and use an AI voice that sounds exactly like you.
Here are some advantages of using an AI voice software like Descript:
- The videos can be edited easily in the future, much like how one can easily edit a Word document or a PowerPoint file. Slides can be replaced, the script can be edited and audio regenerated easily in Descript.
- The videos are easily transferable. Colleagues taking over the course do not have to record new videos using their own voice, but can easily reuse these videos since they are made with a “neutral” stock voice. They can also choose to train and use their own AI voice.
The use of an AI voice to produce teaching videos holds tremendous potential. This technology is heavily utilised by podcast content creators. There are many aspects of harnessing AI that educators can learn from such content creators to produce teaching videos that are engaging and accessible to students.
REFERENCES
Descript (2020). Introducing Descript [Video]. https://youtu.be/Bl9wqNe5J8U
Descript (2022). Descript Storyboard: Preview & Demo [Video]. https://youtu.be/P7SfbmsEK24