Nanite Vegetation Study (Animated)

This was a project to test the boundaries of Nanite and Pivot Painter. The tree is fully created in Houdini using "Simple Tree Tools", and the hierarchy for Pivot Painter is equalized by a custom algorithm. As Pivot Painter 2 only supports 30k elements, the Houdini data packing script had to be adjusted to allow for virtually unlimited leaf counts. Here's an article on how that was done: https://www.linkedin.com/pulse/fixing-houdinis-pivot-painter-indexing-nanite-tilman-mielsch-pbdic/

No scanned mesh was used. The bark is done via a chain of height-lerps along an age gradient baked in the vertex colors of the mesh, using a texture atlas with varying age levels of bark, and then utilizes nanite displacement.The idea behind this was not just to have a high level of fidelity, but also the ability to have per instance random offsets of the underlying gradient, so asset repetition becomes less obvious.

No opacity masks are used, everything is directly cut out with geometry to eliminate overdraw completely.

Unfortunately, Nanite displacement and WPO don't go well together - you may see a few gaps here and there if you look closely.

The reasons for that are better explained here than I could:
https://x.com/BrianKaris/status/1579551563257118720

Pivot Painter, especially with high vertex counts and VSMs, is very expensive though. Reworking vegetation animation from the ground up using compute shaders and skeletons is a big project for the near future, which will make such levels of animation more feasible for games.