The skull model was actually generated from around 70 pictures (probably overkill) taken from a Samsung Galaxy cell phone (I think it’s 14 megapixels?). I’ve read that anything over 8 megapixels works, but the higher the resolution, the more PhotoScan has to work with to generate spacial data from. I use my Canon EOS Rebel T5i that I bought years ago for $800 (think you can get it for $650 now on Amazon). I use the stock lens it came with, but I can’t remember the specs on it off the top of my head. You don’t want a zoom lens, and you don’t want to use a fish eye lens. I’m not sure if you can use a macro lens (if you wanted to model stuff like bugs and such).
You want the f-stop to be between 8 and 16 (from Making - Tested, a decent starter on the topic). That way, most of the object you’re taking photos of should be in focus. If a little bit of the object is blurry, it shouldn’t matter, since every photo is going to have significant (40% or so) overlap. If shooting outdoors, only take shots when the sun is obscured by clouds, so the lighting is diffuse. I got tired of always waiting, so i bought a 3’x3’x3’ light tent to shoot the objects in. If you’re gonna use a light tent, get a stand for your camera, as you shouldn’t be using a flash. Anyways, the standard license for Photoscan is $170 I think, and I don’t think you’ll be disappointed.