It's about making a virtue of a necessity.
When Be wrote that demo the situation is that the other operating systems you might plausibly choose all have working video acceleration. Even Linux has basic capabilities in this area by that point. BeOS doesn't have that and doesn't have a road map to get it soon.
So, any of the other platforms can play full resolution video captured from a DVD for example, a use case actual people have, on a fairly cheap machine and BeOS won't be able to do that without a beast of a CPU because it doesn't have even have hardware colour transform acceleration or chromakey.
But - 1990s hardware video acceleration can only play one video at a time, because "I want to play three videos" isn't a top ask from actual users. So, Be's demo deliberately shows several different postage stamp videos instead of one higher resolution video, as the acceleration is no help to competitors there.
And then since you're doing it all in software, not rendering to a rectangle in hardware, the transform to have this low res video render as one side of a cube or textured onto a surface makes it only very slightly slower, rather than being impossible.
Audiences come away remembering they saw BeOS render videos on a 3D surface, and not conscious that it can't do full resolution video on the cheap hardware everybody has. Mission success.