Houdini 16, booleans and polybevel

   36076   92   10
User Avatar
Member
2062 posts
Joined: 9月 2015
Offline
This is cool vux;

Curious, is the Houdini Toolset your speaking about done in HDK or is it done in vex?

Been playing with vex to try the same result, seems like it's doable, but at the same time looks like it might be more ‘efficient’ to do at HDK level.
User Avatar
Member
538 posts
Joined: 12月 2006
Offline
1. I never used HDK.
2. HDA contains near 50 SOPs with Point Wrangles also
https://gumroad.com/alexeyvanzhula [gumroad.com]
User Avatar
Member
2062 posts
Joined: 9月 2015
Offline
Cool.

Thanks for the info
User Avatar
Member
1755 posts
Joined: 3月 2014
Offline
bardia
This is not even a Houdini limitation, no tool out there bevels beyond the immediate ring of polygons. Doing that would easily as difficult of the boolean if not harder.

That's correct. XSI doesn't do this either, but it does not cap the width value the user can input like Houdini does. The geometry, as seen below in the 1st image, is a mess and you wouldn't want to use it like that anyway, but in some cases, it's good to have the option to go crazy.

Whether the bevel tool should check beyond the first poly loop or not is something I haven't thought much about, although after a brief mental gear screech, one would probably need to clean the mesh anyway if it's for a model and not for a simulated mesh which can be triangulated w/o problems.

Vux's tool produces a nice result, but in some cases where the topology is crucial (quads or quads&tris only), it's not very useful as it is.

The best way to tackle this Sub-D modeling problem is like in the attached .hip file.
If there was a “clean orphan points” (points with only two adjacent edges) option on those Boolean ops, the result would be much cleaner - now one has to roll up his sleeves and manually clean those edges.

p.s. the mesh in the .hip file is messed up for some reason - I must've changed the polybridge group or something and saved after I took the print-screen, but the idea is there.
Edited by anon_user_89151269 - 2017年3月27日 15:00:24

Attachments:
xsi_bevel.jpg (407.7 KB)
cyl_manual_bevel.hipnc (94.0 KB)
h_manual_bevel.jpg (280.7 KB)

User Avatar
Member
1755 posts
Joined: 3月 2014
Offline
If there was a “clean orphan points” (points with only two adjacent edges) option on those bevel ops, the result would be much cleaner - now one has to roll up his sleeves and manually clean those edges.

Spoke too soon - I guess “edge threshold” does exactly that, but it's not quite there yet. It would be nice if only the points that don't belong to the remaining mesh (after a minus operation) would be purged. The excluded points from purging would be those that would otherwise change the hole in this case (the top part).
It's a bit complicated, so further study is necessary.
User Avatar
Member
1755 posts
Joined: 3月 2014
Offline
Here's a much cleaner result using “edge length threshold”.

Attachments:
h_manual_bevel.jpg (331.8 KB)
cyl_manual_bevel2.hipnc (83.9 KB)

User Avatar
Member
538 posts
Joined: 12月 2006
Offline
Vux's tool produces a nice result, but in some cases where the topology is crucial (quads or quads&tris only), it's not very useful as it is.
I think this is a hasty conclusion. Wait for a final HDA video, where I'll show how it works with different kinds of topologies and trim types

PS Show us how your result looks without wireframe in smooth shade mode, from different view points)
Edited by Alexey Vanzhula - 2017年3月27日 15:50:04
https://gumroad.com/alexeyvanzhula [gumroad.com]
User Avatar
Member
1755 posts
Joined: 3月 2014
Offline
Yeah, no, I didn't intend to imply that I know how your tool works, my comment was based solely on the image you shared, that's all.
However, there's one thing we ought not forget - there's no possible tool that produces as good of a result as one would obtain with planning ahead - number of poly loops, positioning, etc.

As far as modelers are concerned these tools (booleans, etc) are starting points and some of them are so good that will get you 90% or more of the way, but in the end in most cases you still have to manually intervene. That is if you care about topology and if you don't, you make the mesh dense enough so that any shading artifact (to which you alluded) is no longer visible. I hope you do realize that the mesh I've attached, with some manual tweaking can be made to look perfect.
User Avatar
Member
1755 posts
Joined: 3月 2014
Offline
For bardia's interest foremost (will also submit a RFE if necessary).

I'll explain a bit what I meant earlier with excluding points from B only when doing a Boolean A-B. More precisely, “preferring” to purge from B not “only”, because in some areas points from B have to be preserved. I'll try to explain as best as I can, with images too.

The 1st image shows a very low “edge length threshold” (ELT), of about .0001. Points market with red are the ones that should be purged, since they belong to object B.

The 2nd image shows the resulting geo after increasing ELT a bit - .009. What you'll notice is that the geometry has been cleaned up but the purged points are not the preferred ones (from a topology/modeling standpoint), they belong to either A or B (the algorithm probably looks at edge length only and disregards the points' provenience) and this is not desirable. The marked purple edges that have been created after the “wrong purging”, are taking the topology into a less than optimal direction. Marked with a faint green are the edges that should've been preserved from the spared A's points.

Will be watching this thread in case there are questions or I've messed up something.
Edited by anon_user_89151269 - 2017年3月28日 11:30:21

Attachments:
boolminus_marked_points.jpg (310.0 KB)
boolminus_marked_edges.jpg (352.3 KB)

User Avatar
Member
65 posts
Joined: 2月 2017
Offline
Just started learning houdini, but thought I'd give it a try. Far from perfect, but I used a polywire on the boolean seam. But I have no idea how to merge it with that nice curve in, like the awesome screenshots so far.
Edited by aoakenfoArchiact - 2017年3月30日 22:45:26

Attachments:
boolean_seams.hip (68.7 KB)
boo_seam.png (315.5 KB)

User Avatar
Member
28 posts
Joined: 3月 2014
Offline
I have slowly been sucked in by this topic. I have used FormZ for years, and this kind of thing is their bread and butter. Three clicks to make this object. Other things are quite complicated in FormZ, but booleans and rounding it does well.

Then I found this video on MeshFusion:

https://vimeo.com/87854599 [vimeo.com]

The new booleans in H16 are really nice, but clearly need work in the seams. I hope they add something like MeshFusion.

Or maybe vux will save the day.
Edited by Reynold Dodson - 2017年4月3日 20:22:10

Attachments:
BoolAndFillet.jpg (239.0 KB)

User Avatar
Member
538 posts
Joined: 12月 2006
Offline
Reynold Dodson
I have slowly been sucked in by this topic. I have used FormZ for years, and this kind of thing is their bread and butter. Three clicks to make this object. Other things are quite complicated in FormZ, but booleans and rounding it does well.

Then I found this video on MeshFusion:

https://vimeo.com/87854599 [vimeo.com]

The new booleans in H16 are really nice, but clearly need work in the seams. I hope they add something like MeshFusion.

Or maybe vux will save the day.


In any NURBS\Parametric modeler it is very easy. But not all can give you feature to change source geometry after trimming. Mesh Fusion can, but it is buggy with lots of intersections.
Edited by Alexey Vanzhula - 2017年4月4日 07:09:24
https://gumroad.com/alexeyvanzhula [gumroad.com]
User Avatar
Member
98 posts
Joined: 8月 2014
Offline
vux
nicoM, nice tool! But it looks like it doesn't fuse the seams between fillets and source surface. Is it true or am I mistaken?
As You said, it doesn't. I mean, I put this thing on internet, so it's hundred percent sure . I think this is clearly mentioned in description of tool. Also, hopefully for me now there is no need to work further on my Mesh Blend thing.

By the way, just to add to wishlist: would be great to have a variable distance fillet as an option, let's say if user would be able to define two points on intersection curve, as min - max blend distance. Imho this should make possible to describe shapes, almost impossible by fixed distance, like body - wings interface on airplanes, so on. Afaik, should be easy to implement such option (while I have no idea how easy is to implement user's interaction of this, in H).
User Avatar
Member
538 posts
Joined: 12月 2006
Offline
By the way, just to add to wishlist: would be great to have a variable distance fillet as an option
Already in TODO list
https://gumroad.com/alexeyvanzhula [gumroad.com]
User Avatar
スタッフ
6367 posts
Joined: 7月 2005
Offline
I guess this answers the question if anyone uses the Round option in Surfsect…?

The Round SOP and the round option in Surfsect are scheduled for dismantling as they rely on pasted surfaces to generate offset surfaces. So if you are using them, we'd like to know :>

The approach used in these operators is:
1) Inflate your surface by the round distance
2) Intersect the two
3) Pull back to the original size
4) Fillet between the two networks

The problem (other than the usual problem of surfsect failing, or not correctly joining trim loops) is that how to connect the new boundaries can get quite complicated and ill-defined. Nonetheless, I'd like us to try something like that for a better bevelling option with boolean.

An alternative approach is:
1) Intersect your surfaces
2) Build a tube from the intersection curve
3) Intersect the tube with your surfaces
4) Fillet between the boundaries created by #3.

This second approach seems better for polygonal modeling as inflating surfaces is its own hard problem.

Attached shows the surfsect sop with carefully chosen geometry so it actually works.

Attachments:
tubeisect.hip (56.7 KB)

User Avatar
Member
538 posts
Joined: 12月 2006
Offline
https://vimeo.com/211722563 [vimeo.com]
https://gumroad.com/alexeyvanzhula [gumroad.com]
User Avatar
Member
98 posts
Joined: 8月 2014
Offline
jlait
I guess this answers the question if anyone uses the Round option in Surfsect…?

The Round SOP and the round option in Surfsect are scheduled for dismantling as they rely on pasted surfaces to generate offset surfaces. So if you are using them, we'd like to know :>


I've played a bit in times of Houdini 13, they were looked more like prototype of nurbs fillet/blend engine, definitively not competitor to great and affordable other tools available today, like Autodesk Fusion 360, MoI, even old fashion nurbs tools in Maya LT (old fashion, but still strong and reliable).
Round from Surfsect SOP, Bridge SOP (blend) as well, were able to produce artifacts (bulges, highly uneven distribution) even on simple trims between two spheres, plain errors in case of closed nurbs surfaces (even trims were not intersecting with boundaries). Finally, I think this is only filleting engine today, or maybe in last 15 years, who is asking to manually define the marching steps.
Dismantle them…. Put some copy in museum, perhaps.
User Avatar
Member
538 posts
Joined: 12月 2006
Offline
https://vimeo.com/212326398 [vimeo.com]
https://gumroad.com/alexeyvanzhula [gumroad.com]
User Avatar
Member
538 posts
Joined: 12月 2006
Offline
https://vimeo.com/213208180 [vimeo.com]
https://gumroad.com/alexeyvanzhula [gumroad.com]
  • Quick Links