Video AutoChecker: this interface uses computer vision algorithms to read the answers provided in the video. Students would need to somehow indicate or signal to the program that they have submitted an answer. For example, a box drawn around particular text (along with annotation provided in the program) could signal the algorithm to look at the contents inside the box.
The computer vision aspects of this project may be too ambitious at this time but similar systems could be developed that require students making videos to signal when they are covering something noteworthy (like an answer to a subsection of the problem).
This post will be further edited as more information is made available, but feedback is welcomed. Advice is specifically requested from students with knowledge in the realm of computer vision. This could be an interesting class project!