PURPOSE
Contemporary cohorts of medical learners are known to prefer visual
learning. Integrating media such as brief animations and videos in medical
education could improve student engagement and knowledge retention. While
this may require faculty to broaden their digital content creation skills to
use such resources effectively, generative artificial intelligence (GAI)
tools could work to decrease the burden on faculty to become content
creators.
METHODS
Using prompt-based model routing, a list was curated with 10 GAI tools
having the ability to generate videos or animations from text prompts,
images, or both. For each tool, a series of trials were conducted in which
identical text prompt or images were used to generate videos. The produced
videos were independently evaluated by two medical teaching faculty and a
medical illustrator for visual quality, accuracy of depicted content and second-order
perceptual evaluations.
RESULTS
Across all tools, median ratings for overall visual quality fell within the
upper quartiles, while medical accuracy showed greater variability. Outputs
from tools allowing greater control over style and pacing were judged to
align more closely with learning objectives but also required iterative
prompt refinement and greater parameter tuning by faculty. Inter-rater
agreement was lowest for granular accuracy judgments, underscoring the need
for content and visual-design expertise in the review of content before
delivery.
CONCLUSION
GAI outputs tend to include clear visual depictions, which can be improved
over iterative prompt revisions, that may increase learner engagement and
preserve contact time for higher order activities such as knowledge
application. However, realizing this potential requires faculty to possess
sufficient knowledge to plan, create, and effectively integrate such media
into their teaching. To ensure these benefits are realized, collaboration and
oversight from both content and technical experts remain essential.