As I’m trying to get to grips with the specifics of sculpting, I feel I don’t understand retopology well enough to make good decisions. I have experience with parametric design and some experience designing assets and levels for games/real time rendering. While sculpting is similar in some areas, it’s very different in many others. Many concerns you have in those disciplines don’t concern the sculptor and vice versa. Please understand that may not be using correct terminology, but do correct me so I can learn.
As I understand it, retopology is the process of increasing or decreasing the complexity of the mesh for the sake of adding resolution or detail or removing unwanted complexity, usually for the sake of efficiency. In Nomad sculpt there are three main ways of doing this:
-
Multiresolution topology: this method allows you to change the mesh complexity in subdivision steps, creating various levels of complexity. I understand the benefit of doing this is that you can easily work on large scale shapes, and then subdivide and do finer details, and going back and forth between them as needed. Shapes stretched out too far can be relaxed with the smoothing tool to prevent things from going askew too much.
-
Voxel retopology: a three dimensional grid of voxels of user defined size is used to ‘chop up’ the model into points attached to that grid. The user can hit the retopology button after every edit, creating a very homogeneous and mesh every time. A downside of sing voxel retopology is that you lose your multiresolution or dynamic mesh.
-
Dynamic topology: a simple mesh is automatically made more complex as shapes require it. This feels like the most modern option as it eliminates much of the worry about the specifics of the mesh, bringing the experience closer to just working on a piece of virtual clay. Downsides seem to be that lots of intricate details and mesh complexity can be added inadvertently, ultimately causing large models, slowdowns and crashes.
Some questions I have are:
-
Am I correctly observing that there is some non-linear interpolation is taking place when multiresolution subdividing, like when curves are subdivided? How are these subdivisions being handeled precisely?
-
Considering there is no underlying ideal model like in parametric modelling, do I assume correctly that voxel retopology is a lossy process as points are replaced by the nearest points on the new voxel grid?
-
Assuming only a small part of the model has been edited and assuming the same voxel size every time, do repeated voxel retopology cause losses in the rest of the model or are they retopologized to the exact same place on the voxel grid every time?
-
Do I correctly assume voxel retopology with different voxel sizes every time is much more lossy because information is lost every first time a new voxel size is used? I.e. the current model is being ‘translated’ to the new grid every time, introducing something akin to rounding errors?
-
Doing some hard surface experiments I find that when cutting shapes using trim or split yield much more accurate and lower poly shapes than voxel merging two shapes. However, having seperate shapes limits flexibility. Conversely, I find voxel remeshing often results in jagged or blocky shapes where straight lines are required, even with fairly high density meshes. Am I missing a trick? I feel my parametric experience may be leading me astray here, as may my tendency to seek out low poly shapes.
-
When cutting curves out of comparatively low poly multiresolution volumes I often end up with relatively complex faces, unless I do a voxel remesh leading to above issues. I have trouble simplifying these complex parts of multiresolution volumes. Is there a way to easily reduce complexity? Not worrying about the resulting complexity occasionally leads to issues when trying to do things like move, drag or push, which are less fussy on less complex meshes.