Anyone know good and safe Edge Detail way for rbd destruction ?

[Ggradest 0]

Hi.I have been looking for a good edge detail way for Houdini destruction vfx.

  For Destruction, pieces are created by Voronoi split, but the straight cutting line is not good looking.

  EdgeDetail of RBDmaterialFracture node solves this, but it is very unstable for small or complicated geometry. Boolean-based segmentation (cutting) method is also less stable. so I want to distort the edges formed by Voronoi segmentation.And I don't want to use the shader-based method too much.

  The only way I found was https://vimeo.com/178206035, which is fast and stable, but for most geometries it forms strange polygons that don't fill tightly and can't keep the original geometry looking.

  This can be solved by the trick of TransformPieces only when the Lowres fragment moves by Sim, but the whole topology changes drastically, so it becomes very heavy if transferred to other software through Alembic etc. Motion blur is also difficult (especially not compatible with c4d)

  The problem of UV distortion is also serious.

  See the file for details of the problem.

  Does anyone know Edgedetail's method for a good Voronoi split that fills the gap exactly?

 

edgedetail.hiplc

[Midasssilver 11]

One thing you can do is add a lot of noise to the entire geometry of the fractured pieces, including the edges, then switch between the active shards and inactive shards, so that you keep the active shards that have the displacement, and the inactive ones that don't.

[Ggradest 0]

It was my blind spot! Thanks Joseph.

Deforming their mesh by active is certainly good for alembic.

From the point of view of data size, problem remains  because that deforming geometry can’t be outputted packed alembic geometry.

Anyone have static mesh idea?

 

[willow wafflebeard 15]

try gamedev uv transfer?

Edited September 16 by willow wafflebeard

[Ggradest 0]

Thanks willow! That node can give me more good result.
It can't be compiled because of attribute transfer so it's quite slow in my large scale destruction simulations, but it seems to be one of the good solution.

Edited September 17 by Ggradest