Tesselator Quad-Remesher


(Jeacom) #43

I had this plan but it’s hard to archive a good connectivity like in the source.
There’s a software called wrap3 that basically does this, if I understand how this kind of algorithms works, sure I will try out.

(Jeacom) #44

I am updating Tesselator for blender 2.8, soon it will be avaliable on blendermarket:

(Jeacom) #45

Okay, Its updated for 2.8, who already bought the addon should be able to download the new version in the orders history on the account settings.
Btw, this is temporary, I want to reimplement the code from scratch to make it faster on 2.8 so stick around, and if you like Tesselator please share, that helps alot!

List of addons that work with 2.8
(masterxeon1001) #46

awesome! Been wanting to get back to messing with this!

(Bohdan Lvov) #47

Nice, works like a charm in 2.8!

Btw, is that possible to add a feature which would allow for algorithm to consider an object as one shell?
Because now if you have intersecting geometry, you have to separate parts, then use Union Boolean, hoping that it won’t hiccup with the dense geometry and only then use Tesselator on a garbage geometry after boolean.
It would be so awesome, if Tesselator could do it straight away!

(kabu) #48

What you are pointing out here is a plaguing problem; the need of a robust algorithm / addon for calculating the outer hull of a multiple objects composition, while removing self-intersections.
Same problem here and I don’t know where to find it and if it exists and addon for that.
Maybe @Jeacom is able to do it, would be very big; especially because it has nothing to do with quad remeshing.

(Bohdan Lvov) #49

Yeah, that’s why it’s so interesting to me. I watched Jama Jurabaev’s tutorial about turning Grease Pencil into geometry (https://www.youtube.com/watch?v=VKmZNPpRXNM), and only one thing which for he switched to ZBrush is exactly that problem. ZBrush has Dynamesh, which able to resolve that issue. Tesselator come a long way and almost on pair with Dynamesh now, EXCEPT this unionizing feature.

(LeighAH) #50

If you can resolve it manually with a series of booleans, then it should be possible to make the software just perform those booleans for you without having to resort to complex mathematics.

(kabu) #51

One should try with some python programming. Problem is if the math is not robust self intersections won’t go away. Afaik, Blender booleans are not that robust. I didn’t try anyway, so I may be wrong.
Consider even libraries like CGAL have problems with that…

(kabu) #52

Interesting video, now it’s very clear. All those edges getting inflated into the others, and the resulting intersection mess. Evidently it takes some processing step for getting sketches out in a good way!

(LeighAH) #53

You may well be right, I don’t use many booleans myself. I’m going purely on the premise that there’s a straightforward manual workaround at the moment. If there isn’t, then it won’t work.

Edit: I just tried stacking up a few booleans myself and it was a total disaster. Looks like the positioning and order that you apply them is crucial, so yeah, not that simple.

(Tosky) #54

OpenVDB remesh ($$) let you do that: you can join the meshes then remesh and voitlà, no self-inersections issues with a dynamesh like result. Not for retopo of course.

(Bohdan Lvov) #55

yup, here’s another one: https://www.youtube.com/watch?v=AFXwG-l9F0I

And this is the specific moment where shown how it looks after converting to geo (middle one) and how Dynamesh is processing it.

(Bohdan Lvov) #56

holy molly, that’s amazing! thanks for the shoutout!

(kabu) #57

Ah, good to know that, thank you. I’m on Os X and there’s no version of it. I tried it on Windows but I must admit I wasn’t able to take anything meaningful out of it. Probably it takes more dedication… :face_with_raised_eyebrow:

(kabu) #58

Yes Yes Yes. I’ve just bought a new tablet, and I think I’ll definitely give this workflow a try. Sketching instead of box modeling. Yes.

(Jeacom) #59

Okay, so the features you guys requesit are:

  • Better topology pencil handling
  • more robust symmetry
  • self intersection handling

While all those features are hard to a non-computer scientist like me, by far the hardest is the self intersection problem. Its easy to deal with multiple solid overlapping objects but a single object intersecting itself is tough.

I guess zbrush uses some kind of obscure magic pixologic won’t reveal for obvious reasons LOL.

I am researching ways of doing it with a raycast particle system but I can’t guarantee it will be a fast algorithm, at the end I have the skills of a indie game developer.

but I wont give up, lets see what the future reserves.

(Jeacom) #60

Okay, Starting a new version from scratch.
After a week of frustration, I could understand a bit the instant meshes paper.
And after a day learning Numpy, I did implement a faster variation of the Field Smoothing algorithm.

Its mostly equal instant-meshes field and will support better the guides and constraints, but I did some heuristics and turned it more like a generic algorithm with some more randomness.

Its not particularly faster than Tesselator’s field generator, which does everything in a single pass but It’s waaay smoother and smarter.

(Metin Seven) #61

That looks promising! Good luck, and please keep us posted!

(Bohdan Lvov) #62

Majestic! Can’t wait to try it!