Deforum currently has 6 monocular depth estimation algorithms: Midas+AdaBins (old), Zoe+AdaBins (old), Midas-3-Hybrid, AdaBins, Zoe & Leres.
This is a comparison of all 6, changing only the depth algo. All other settings are the same.
The movement parameters, which were defined with Parseq, use a combination of rotation around the y axis and translation along the x axis to create the effect of rotating around a subject. As you can see, this works better in with some algorithms than others. :)
See here for the full Parseq config and Deforum settings: https://github.com/rewbs/sd-parseq/discussions/106
#parseq #deforum #stablediffusion #ai #3danimation #depth
