Originally taught to me by RedDragon, this technique comes up in multiple projects as an optimization for baking a snapshot of a scene into a light-reactive picture for use as a low quality or far away LOD.

The technique:

  • Spawn a resonite camera
    • It should not be moved during this process
    • Capture only targeted should likely be selected
    • Orthographic should likely be selected, but this will depend on use case.
  • Pose entities in front of the camera
    • Once posed, the entities should not be moved.
    • If the entity is an avatar, turn off the VRIK component to “freeze” the avatar in place.
  • Change the texture of the entities to an unlit variant with the correct albedo
    • All parts of the avatar must be changed
    • The material tool can make this easier by holding a reference to the root of the entities and pressing secondary.
  • Take a photo with the camera
  • Change the entities to a world normal matcap
    • Paste this item link for a pre-configured material
  • Take a photo with the camera
  • Create a new PBR material
    • Set alpha to cutout
    • Assign the unit picture to albedo
    • Assign the world normal picture to normal
  • Apply the material to a quad or other flat surface.
  • Wiggle a light around the surface and see it react.

The world-space matcap material used:

resrec:///G-Earthenworks/R-E793EADA98C714A3B64FF55612742CF744470AA078378E5F3EFC7723AE4161A6

To use a folder link:

  • Copy the link to your clipboard
  • Paste the link inside of Resonite, this will spawn an inventory folder link.
  • Save the folder link to your inventory
  • Spawn the material from the folder.

The technique is very good for far objects where depth isn’t important, low quality LODs for expensive assets (like avatars) or providing “Windows” into other spaces.

RedDragon Wizardry

RedDragon has a more advanced version of this that prints perspective based light reactive flat-avatars.

The basic technique is the same as is used for other flat avatar bakes (such as used by Creator Jam): It works by taking dozens of pictures of a central point with the cameras looking at that point, bakes the prospectives into a tiled maps, and picks the correct tile based on the direction of the user versus the avatar.

RedDragon modified a version to use the above technique to make the flat avatars light reactive, which dramatically improves realism.