New UCL study shows the benefits of using AI-generated videos for adult learners

Today, the University College London’s (UCL) Faculty of Education and Society is releasing a research report that investigates the effectiveness of AI-generated synthetic videos for adult learning. The study, undertaken by Zoe Ruo-Yu Li, a PhD student at the UCL Knowledge Lab under the guidance of Prof. Mutlu Cukurova, looked at the impact of AI-generated synthetic videos on adult learners' recall, recognition, and affective feedback compared to traditional learning materials such as human-generated text and videos. The study involved 500 adult learners and used a mixed-methods approach, with a pre- and post-test design.

There were four learning conditions considered:

• A video transcript about food safety in the workplace created by a human instructor. The transcript was used as the baseline for the study and delivered to participants as a static, text-only asset.
• The same text above was framed to participants as AI-generated.
• A video created by a team of instructional designers with traditional video tools and methods such as filming a person delivering the presentation above in a studio and editing the video using traditional software.
• An AI-generated synthetic video which faithfully replicated the human-made script and visuals but was created entirely in Synthesia using a stock AI avatar.

The researchers found that participants overwhelmingly preferred learning through video over text materials. Moreover, those who watched the AI-generated videos performed just as well in learning tasks measuring their recognition and recall performance as those who learned from human generated video. There was also no statistically significant difference observed in participants' affective responses between human instructors and AI avatars.

The UCL study suggests that AI-generated synthetic videos can be equally effective as traditional learning materials in terms of knowledge acquisition, while offering a more effective and engaging learning experience for adult learners.

Let’s look at how these findings track with insights reported by our customers in recent years.

Participants preferred mainly learning from videos instead of learning from text

There is no denying that the rise of TikTok, Instagram and YouTube has had an impact on the preferred medium through which people learn. For many people, looking for information about a topic now starts on video platforms rather than a traditional search engine.

So it was not surprising to see that 77% of the participants in the UCL study preferred learning from a video rather than text, and that many of the participants found the text materials boring and monotonous.

In fact, 94% of the participants expressed a desire to learn from video materials if their organizations used them as part of their training programs.

‎We’ve heard similar feedback from our customers: Spirit Airlines saw a 6x increase in engagement when they switched from text-based communication to Synthesia videos, reflecting the general preference that most people have for video over text. The airline’s HR team used to share wellness and benefits policies as written updates through the company’s intranet and sometimes record summaries of those updates with their smartphones. However, they found that long text was not suitable for flight attendants and other on-the-go employees to check in real-time, and the constant change and fluctuations made re-recordings hard to scale.

When they switched to Synthesia for internal communications, the HR team was able to create over 326 hours of video and distribute that content to 13,000 employees in just two weeks. Employee support inquiries over the phone also dropped by 76%, indicating improved retention of information.

Recall and recognition scores as well as affective feedback when learning through AI-generated video were not different to those for human instructor-generated video

Thanks to advances made by our R&D team, our Expressive Avatars are now able to replicate human emotion with near perfect accuracy, leading to videos where it’s becoming very difficult to tell whether the people onscreen are AI generated or not.

Expressive Avatars have significantly improved the quality of training materials, as they deliver a sense of realism which is likely to remove the distracting “uncanny valley” effect experienced by older generation avatars. 

The UCL study showed that videos made with Synthesia were just as good for learning as videos made with real presenters - a fact experienced by Zoom when they moved from filming training videos with subject-matter experts to Synthesia-generated content with AI avatars.

Zoom’s learning and development team relied on instructional designers and subject matter experts to create training videos for Zoom’s sales team of over 1,000 people. However, creating training videos was time consuming because the instructors had to record many takes to be satisfied, which took a lot of time. And if updates were needed in the training material, the team had to re-record the entire video.

By replacing traditional video with Synthesia, Zoom was able to create videos 90% faster than before, producing content in less than an hour. One instructional designer was able to create over 200 videos in less than six months. This new way of working also freed up Zoom's subject matter experts who no longer need to repeatedly record themselves, freeing up 15-20 hours each month to work on more productive or creative tasks, resulting in monthly cost savings of up to $1,500 per employee that were previously spent on creating training videos.

Participants watching the AI-generated synthetic video completed the training faster than those watching human-made videos

The UCL study found that there was a 20% drop in the average time spent on the training course when viewers watched the shorter length of AI-generated video compared to the human-created one, with no negative impact on the learning gains.

‎For large organizations, this number could bring significant cost and time efficiencies when multiplied with the number of employees. For example, if every employee in a 50,000-person organization had to watch a five-minute mandatory video on health and safety or cybersecurity awareness every year, the company would save 833 people-hours. Multiply that with the average salary of a worker in the private sector as of last month was just above $35, leading to an instant $30,000 saving just with one video.

But the advantages in cost savings don’t stop there - we also need to consider the time and money spent making those videos. For Synthesia, a minute of AI-generated video starts at $1.8 with a Creator plan. Meanwhile, filming and editing videos with traditional means easily catapults costs into the high four figures. Even if we assume a more conservative cost of $500-$1000 per minute, that’s still $5000 to make the five minute health and safety video above. That doesn’t take into account the costs to update the videos once created or to localize them.

BSH, the largest manufacturer of home appliances in Europe, had similar pains with their knowledge sharing. Using Synthesia, BSH quickly developed a scalable, AI video-based knowledge sharing program through which they achieved 30,000 views through their web portal. They achieved 70% cost savings in video production compared to using an external agency and a 30% increase in learning engagement.