By “adaptive subdivision”, I assume you mean changing the subdivision process to subdivide less where the geometry is already dense. I believe Open Subdivision can be made to do this, but I haven’t looked into that. And I agree that a model that has bevels with a lot of segments in it (including especially custom profiles) would likely be a good case where this would help.
I see that as a fairly separate target from anything to do with Bevel, however. It should be a modeling project on its own and prioritized among the things we doing in the modeling space.
I think that I didn’t explain properly. I talk about the system to make the bevel more dense where the curve need more subdivisions (in curves) and with less subdivisions where the curve is near to straight.
The same that people ask to the splines. TO reduce the geometry, something like this
Oh, so you are talking about the sampling of points along the curved custom profile. That is not currently planned for custom profiles, but it certainly makes sense for something to consider in the future.
If there are enough samples, give one point to each edge so each control point will have a sample
If there are enough remaining samples, give a common amount to either
A. All edges if “Sample Straight Edges” is enabled, or
B. Curved edges if it’s not enabled
Give the remainder of the samples to the most curved edges.
A truly “adaptive” sampling method would probably switch out steps 2 and 3 for spreading the remaining samples out by giving each edge a relative weight of its “curved-ness” and spreading out the samples based on that weight. If I have time I’m happy to try that because it sounds like an improvement over the method I’m using.
Interesting idea, Wazou. If I understand it correctly, you propose that the vertex weight multiply the width spec at that end of beveled edge. So that one can get tapered bevels. Another way would be to have a “taper” parameter, but that has the problem that you don’t know which end is which (a similar problem to the “which end is up” we have for custom profiles in the GSoC project). Your idea solves that.
As you probably can tell, the current use of the vertex weight is only to be an “on/off” switch. If the value is < 0.5 then it is “off”. If we used your proposal without making it optional, it might break older models that use funny values for vertex weights. Though this seems unlikely to me.
There have been a couple of people on RCS asking for something like this, but this would be the most appropriate one I think:
Some kind of fall off system. This gif shows a widget being used for that, but I don’t know how we would seamlessly incorporate that into blender, as I don’t know modo at all, and am not sure at what step in the bevel process that the widget should show up. I suppose if we wanted to take that route it could be a tick box in the operator pop up box, and in the modifier. If ticked, then a widget appears.
Widget or not, I agree that I like the idea of having the bevel weights not being an “on/off” switch. If they correspond more to a 0-1 float range, then we could have excellent control. It might not be the speediest method in practice, but it sure would be flexible.
I would image in the future if we had a vertex group node, we could randomly assign bevel values to different vertices, and get some pretty wild results!
Actually, after thinking about it, I don’t like the widget idea, it reduces flexibility. With a widget like the one in the gif in the link, you can’t create bevel profiles that grow and shrink within the same bevel. It’s also fairly quick to go through the vertices manually and adjust the weight just like that.
We tend to get the same suggestions repeated and I myself have sometimes forgotten that they are already on that list. OK for this thread to try to raise my awareness of the importance of some of these suggestions, but we needn’t go over them in great detail again if they are already on the list.
Are there any plans to get this functionality incorporated sometime soon: https://developer.blender.org/D4328
Seems like it’d be incredibly useful for non-destructive modeling. I think it would be even better if the boolean modifier would output a new vertexgroup (for the new edges) that could then be used for the bevel modifier for even more precise control.
I see what you mean. The suggestion I were talking about was only tangentially related to the bevel modifier, so I think there might have been a bit of a misunderstanding. What I were asking for was for the boolean modifier to mark the border of the intersections it creates to be made part of a vertex group, so that this border could then picked up by a bevel (modifier) that is limited to only the intersection itself. So the bevel modifier wouldn’t need to make any new vertexgroups. This is similar to what the patch I linked to talks about (only this patch talks about giving the new border a bevel weight, instead of vertex group).
I’m not sure what you’d want to use a vertexgroup from all the bevelled edges for to be honest, but that seems like a completely separate thing from what I’m talking about.