From: Shack Dougall
I doubt that LL would go for anything that added more mesh points. Also, I think subsurf is best left to the modeling tool where you would have more control.
The Pro is that sculpties can become smoother, but there are two underlying factors that determine smoothness: the number of vertices and the precision with which they can be placed.
The subsurf idea increases smoothness by attacking both of these. It adds vertices and it increases precision.
But I think we have enough vertices in most cases, particularly with oblongs. The biggest impediment to smoothness appears to be low precision which is a consequence of encoding the pixels using only 24-bits.
So, I'm with Drongle. Give us more bits per pixel.
As you mention the main problem is the sculptmap resolution and getting more bits per pixel would be an option. On the other side i guess, that a "smooth" button (plus maybe a smoothing-slider, e.g. from 0 to 100 %) could be an option to actually save pixels instead of increasing pixel count... let me explain why:
As the allowed vertex locations are on a 256*256 grid we typically need to add a lot of work in order to "fit a sphere into a square". The best results are always achieved when the particular sculptie just "happens to be defineable within the grid". Now very often we do not need more vertices, we just need a finer control over the vertex locations. Here is an example:
The follwing 4 cylinders have been made out of meshes with 4, 8, 16 and 32 faces along the x axis:
So they pretty much look identical. The trick is: The vertex locations are calculated with floating point numbers.
Here is the mesh:
Now i have applied a vertex subdivision to all 3 cylinders, where i added 3 levels of subdivision to the lowest cylinder, 2 levels to the next higher cylinder and 1 level to the one just below the topmost cylinder:
It becomes very hard to see the differences...
So what is the conclusion:
1.) If we where allowed to place our vertices with higher precision, we would need less vertices to achieve smooth results. (Maybe better: We would need less work to arrange the vertices...)
2.) If we had a smoothing option, we would need less vertices to define the model.
3.) If we keep the current vertex position limitations, but add smoothing where the intermediate vertices could be placed outside of the sculptgrid limitations (precise calculation of the catmull-clark locations) we could possibly save half of the needed vertices (or even more).
Let me explain point 3.) : Assume, we had the low level model using only 4 faces along x (better said: 'U'). Then lets apply 1 level of smoothing. We would end with 8 faces in U. But due to the automatic fine placement of the calculated vertices we get rid of the so often seen jaggy surfaces of sculpties. Look at the first image above and there at the 2nd cylinder from the bottom. It is almost perfect! 8 faces compared to 32 faces used by a standard sculpty: 4 times less faces -> ~4 times less vertices for that dimension. If you apply the same thoughts to the other axis (V), you could easily end up in ~8 times less vertices with increased quality of appearance.
I know, this is a bit idealistic and needs more thinking. Especially it might be interesting to limit the subdivision mode to one axis (U or V) to avoid unnecessary creation of vertices where they are not at all needed.
Please feel free to take this contribution as a starting point for some more thinking. Maybe someone would even be able to add the subdivision mode to the viewer-sources just to play with it ? I don't know how to do that sort of programming
but i am pretty sure, that adding such an option just on the viewer side could become very interesting without breaking the current implementation of sculpties at the backend side
thanks for your attention.
have a god time
Gaia
