The project already used frame interpolation (through DaVinci Resolve) to double the number of frames in the video. Plus a low-tech "inside the wall" panned and blurred frame. Given the set-up, fancier scene reconstruction to build a 3D model of the environment could be combined with texture reprojection to improve the interpolation and allow the "inside the wall" frame to be perspective corrected, but that would be much more involved on the software side than what was done here (and beyond what I could do too).

The tech to look out for in this space is called NeRF (Neural Radiance Fields).

Just do a quick search on that term and you will find a whole bunch of videos that reconstruct 3d scenes using digital stills. I just saw a paper a couple of days ago showing advances in reconstructing human expressions that change over time using advanced techniques.

Yeah, the Frugal Man's Bullet-time is a DSLR, Instant NGP from NVIDIA, and COLMAP.

I want to show the creator this specific video on DAIN (Depth-Aware Video Frame Interpolation) https://youtu.be/sFN9dzw0qH8 !

https://github.com/baowenbo/DAIN