PDA

View Full Version : Additional Metanurbs Subdivision Method



sire
08-12-2003, 04:11 AM
In LW, there already are four flavours of sub-d surfaces to change from:

- Metanurbs
- Metaform
- Smooth
- Faceted

But only the first one is implemented as a non-destructive mode opposed to a destructive modeling step (and therefore can be carried out at render time with a user-defined number of patch divisions).

Actually, the last one is too. One only has to set the SubPatch Weight to 100% everywhere. The beauty of it is, since this is governed by a weight map, the same geometry can use both subdivision methods gradually weighted on different parts. And with a negative percent value, the behaviour can even be reversed.

Nice. But why not also adding Smooth and Metaform as sub-d methods? I mean, in the same manner Metanurbs already behave? This would mean the user could add another weight map which controls the weighting of the Smooth subdivision method on each vertex. The Smooth method has the strong advantage of having the resulting surfaces actually running through the cage vertices. Sometimes this is extremely useful. Really. I already was in situations where it would have made me cry out of joy, had this been possible! :)

Default would still be classic Metanurbs when hitting TAB. But among the VMap stuff there would be a new command which adds another SubPatch Weight Map (called Smooth Weight for example), which acts as a layer on top of the ordinary one. 0% would leave the vertex unchanged (that is, takes the value of the ordinary SubPatch Weight Map), 100% would mean only Smooth subdiv, 50% would mean half Smooth subdiv, half the ordinary SubPatch value. -100% would be, well, a -100% mix (let's just see what the mixing formula spits out... :))

Metaform, however, could be difficult (or impossible) to merge with the other methods because it creates another topology than Metanurbs when used on triangles. So at least this could not be mixed with the other methods, but could still be available as an own one. Although the results are practically identical to Metanurbs when used on quads, with triangles not only the topology is different, but also the resulting shape, which sometimes would be the better choice because it always gives round results, whereas Metanurbs often gives sharp peaks or edges at unwanted places which are difficult to get rid of. The Patch Subdivision setting also would rather mean a number of iterations than number of new vertices along an edge between two cage vertices. But this should hardly be a problem.

So, the subdivision algorithms, as I said, are already there, they only would have to be integrated into a non-destructive mode like Metanurbs. This would be a highly welcome feature, I guess... :) Even greater if implemented with weighting between compatible methods, as suggested! :D