There are a few major technical problems with my concept, in which I’ll be fixing them with a combination of concessions and libraries.
Regarding the magic circle detection, I’ve resolved to use text. Simply take the name of your preferred devil, and POOF! It shall appear. For that, I’m using Google Cloud Vision.
However, within this fix is another issue: how do I send an image to the Cloud if it’s warped? How do I even get a .png from Unity? I’m relying on perspective correction to address the issue, looking to this library as example.
By making some Game Objects to focus on the corners of the page (which I can make using Image Targets, currently represented by spheres) I can effectively get the vertices representative of the page coordinates, mapped to 2D. Then I can shove them through a homography matrix, to map that to a RenderTexture, then to a clean rectangular 2D Texture. Then I can turn that into a png that I can port to the Cloud.
Currently I’m trying to get the shader that actually does the homography calculations to work. It’s a work in progress.