09-06-2019, 12:42 PM
Hey,
thanks for the quick response! The person responsible for the curvy integration is not available anymore for our team and I want to find out, if Curvy could be an alternative for our current workflow, which I described earlier. Since I do not have any knowledge about Curvy or the Curvy generator system, I would like to find out what possible limitations are before diving into the tools. I do have some years of experience with Houdini and am not entirely new to procedural content creation. So I think I will understand what you are saying.
I got informed that curvy had been used here for example for mass-instancing of railway sleepers along splines but that it got too slow to work with at certain spline lengths that we are dealing with.
1- the sleepers have <300 triangles for LOD0 and <100 for LOD1. I don't know how long the build times get when 200.000 of those objects have to be cooked along a 115.000m long spline. But that's basically an anwer I'd like to get..
2- Great, there's potential for cooking time optimization
3- Yes, we make use of LOD prefabs. Do you have a roadmap for upcoming development of curvy?
4- perhaps that would be an option. I imagine to do many more things with curvy later - for example include cooking of large parts of the railway system. But for now, quickly placing those sleeper is my first goal.
5- read it. So it's a shader thing and I need to find out if -once the sleepers are instanced/placed- I could perhaps assign the GPU Instancer to the prefab and let this take care of the insane amount of draw calls.
6- this is an entirely different topic. We get the input splines as .dxf files, attach dummy objects to each point on the curve, export those objects as fbx files, put those points into a curvy spline container, where they get named and re-create the spline. Is there a better way? It expects JSON, right?
Best, Bodo
thanks for the quick response! The person responsible for the curvy integration is not available anymore for our team and I want to find out, if Curvy could be an alternative for our current workflow, which I described earlier. Since I do not have any knowledge about Curvy or the Curvy generator system, I would like to find out what possible limitations are before diving into the tools. I do have some years of experience with Houdini and am not entirely new to procedural content creation. So I think I will understand what you are saying.
I got informed that curvy had been used here for example for mass-instancing of railway sleepers along splines but that it got too slow to work with at certain spline lengths that we are dealing with.
1- the sleepers have <300 triangles for LOD0 and <100 for LOD1. I don't know how long the build times get when 200.000 of those objects have to be cooked along a 115.000m long spline. But that's basically an anwer I'd like to get..
2- Great, there's potential for cooking time optimization
3- Yes, we make use of LOD prefabs. Do you have a roadmap for upcoming development of curvy?
4- perhaps that would be an option. I imagine to do many more things with curvy later - for example include cooking of large parts of the railway system. But for now, quickly placing those sleeper is my first goal.
5- read it. So it's a shader thing and I need to find out if -once the sleepers are instanced/placed- I could perhaps assign the GPU Instancer to the prefab and let this take care of the insane amount of draw calls.
6- this is an entirely different topic. We get the input splines as .dxf files, attach dummy objects to each point on the curve, export those objects as fbx files, put those points into a curvy spline container, where they get named and re-create the spline. Is there a better way? It expects JSON, right?
Best, Bodo
