This tutorial shows how to use normal maps in RealityCapture to fake detail - from very high polygon models onto models with simplified or decimated geometry.
https://dev.epicgames.com/community/learning/tutorials/Rky5/capturing-reality-how-to-generate-optimal-normal-maps-in-realitycapture-on-simplified-models
This was very helpful!
Would it be true to say that the generation of normal maps is really only needed when projecting the texture from a high triangle count model onto a simplified model? It’s using the normal map that allows that simplified model to appear to have the detail of the original model, correct?
Thanks!
Hello, yes we might say that. If we have superpowerful computers we could use 1:1 3D models with all the details. Unfortunately, we are still bound by today’s hardware and computing power so that is why we need to use normal maps to fake the details for our eyes so real-time applications run smoother and also offline renderers run faster.
This is something that you can try yourself and experiment with. Place a couple of full-resolution scans from RealityCapture to Unreal Engine or Twinmotion and compare the performance with simplified models with normal maps. You might not notice the difference with one model so that is why you should try to build a scene from multiple models like you would for a game level or an architectural visualization.