Continuing the model , sooner or later I will have to smooth the model. However, there are parts, which needs to be smoothed, and some which don't. I found that for static meshes it is better to use NURBS (is it NURBS, isn't it). Do I have to smooth whole model if I want only to smooth certain part (front lights on image below). If I can do that, do I have to detach parts and attach them back after smoothing ?
Replies
Or are you talking about using polygonal smooth shading? Which is smoothing the vertex/surface angles to affect how the mesh is rendered?
If you intend to subdivde(NURMS) the mesh then your topology is far from ideal and will need a lot of either/and/or creasing/in-line control edges.
I recommend researching the fundamentals of how sub-d surfaces behave. Here is a great video that all sub-d virgins should watch:
https://www.youtube.com/watch?v=ckOTl2GcS-E
When you say 'only a 1/3 of the model needs it', it doesn't really make any sense.
It's hard to explain but sub-D modeling techniques like edgelooping/termination/routing edgeflow/control edges/etc are the basics of controlling how the sub-d algorithm behaves when you smooth the cage-mesh. Without understanding these basic tenets you will encounter a lot of heartache.
Also, I saw many videos where people didnt subdivide the mesh. Should I always subdivide if I am not making lowpoly ?
There are N-gons and triangles - which are not necessarily bad(as they ultimately sub-divide to quads) but can often subdivide unpredictably.
Little or no supporting loops/edgeflow control.
This is part of a Sub-D car I built a few years back. You can see where I've used edgeflow control to support the forms. Also, note how I've broken the panels apart as separate meshes(as they would be in real-life) to make it easier to model.
Okay, so bare with me in this part. I know I may be frustrating, but I want to get this straight. What is the best way to model the light part ? I created that shape using cut tool, then connected edges so it created quads. Afterwards I smoothed it using creasing with value 0 and 1. You pointed out that this creates mess in topology, so how would you do it ?
For parts like the headlight area where the direction of the topology changes a lot I would still use edgeflow control to turn corners. That way you are always in charge of your edgeloops and can easily add continuous loops/chamfers/etc. As you've discovered - from what I mentioned above - locally sub-dividing adds a topological mess.
You are trying to add all the components of the vehicle into the main topology. This is not a good way to model. Think of building the vehicle as it is built in the real-world: separately model parts that are actually separate. It adds to believability, and makes your life much easier as a modeler. Modeling everything contiguous as a rule - as you've been taught/told/heard - is not good practice, and certainly not a rule.