The 3D effect of a movie is always based on the same thing for VR, Anaglyphs, shutters, stereographs, etc. It is to show a different image to each eye. Most of these techniques make the film very difficult to see without the appropriate glasses.
Except for lenses of different opacities (pulfrich) in this case the scene needs a lot of movement, this only serves in 2 or 3 shots of 4232.
There are already video designed to choose our own way of interpreting 3D for example videos containing the two images one for each eye in a vertical or horizontal way.
These can be reproduced in mpv, VLC, mplayer, ffmpeg, bino, etc. And they are compatible with some 3D televisions.
The short film itself does not have the two images that are needed to create the 3D effect, so how to create them?
The best thing would be for your editor to have the possibility to export both images. But if not we can predict depth, this technique also creates internal volumes in the image.
gmic -i imagen.png fx_tk_make3D 22,10,-10,20,0,5,0,1.405,3,0,0,0,15,1.2,25,1,0,0,1,0,0,2,200
My first choice was to think of neuronal networks, there are several that do this, but they are monstrous ending up weighing GB and are not meant for cartoons.
So I decided by GMIC that it has fx_tk_video3D a program that allows to convert 3D images deducting the depth with a method called “study” most of my drawings work well.
# crear carpeta temporal
# exportar imagenes
ffmpeg -i video.webm $TMP/frame_%d.png
# convertir a 3D
gmic -i $TMP/frame_1.png fx_tk_video3D 22,10,-10,20,0,5,0,1.405,3,0,0,0,8,0,"$TMP",1.2,25,1,0,0,0,1920,1,0,0,1,"$TMP","frame_",1,$(ls $TMP/|wc -l),0,0,"echo",0,0,1,5,0,0,0,1,0,0
# juntar todo en un mkv
ffmpeg -i $TMP/frame_%d.png 3d_video.webm
I like mpv and use this setup that is compatible with most of the movies I found on the internet.
Reproducir el video:
mpv video.mkv -vf "stereo3d=in=ab2l:out=arch"
This is to see you with red and cyan glasses. But with the -vf stereo3d=help command you can see the other options.