Progress on a new Bevel Geometry Node

Work has begun on a remake of Bevel as a Geometry Node. As well as rewriting in C++ for better future maintainability, the main new user feature is handling overlap collisions by merging geometry as the bevel offset edges start to overlap. The node will also have a collision-aware face bevel feature that will act as a smart inset.
This talk is about the progress on the new node, with an emphasis on how to specify what it does, and the algorithm used to deal with collisions by adapting the "Weighted Straight Skeleton" algorithm to the 3D mesh case.

[@charliejolly6022]

This really highlights the complexity of bevel and why it is not trivial. Hats off to Howard, Henrik and Hans. The 'H-Team'! ;)

[@Yassir.A.P.]

All new Triple-H team? 😁

[@jasonhurdlow6607]

As the author of a bevel plug-in for another package, it's nice to see someone share my struggles and pain.

[@MikkoRantalainen]

At last! The bevel (modifier or geometry node) is biggest limiting factor of Blender for nearly all the stuff I've done with Blender. The clamp overlap is really poor workaround because it often limits the bevel to nearly non-existant lengths. I hope Blender Foundation could sponsor this work so you guys could finish this as your day job.

[@Kavukamari]

i hope the laplacian vertex hole filling can be added generally as a tool because filling smooth grids is hard

[@howardtrickey]

Good idea. Henrik and I will think about this.

[@JeriKusumaa]

I just realized, how complicated bevel is. Love your work man.

[@LawrenceAaronLuther]

Yea, so much respect to these guys. I'll never again think "damn why can't they just have x feature" after realizing just how much thought and work goes into some of these functions.

[@tuanas458]

Lol

[@xanzuls]

He's the reason why Blender has the best Booleans as well.

[@4mb127]

Howard is doing great work.

[@Yassir.A.P.]

We don't deserve this. We users tend to see things as trivial, while the devs like Howard still hasn't finished solving this even after 5000 lines of code. Five thousand. And not done yet. Just for a "simple" bevel. Good Lord bless them developers.

[@mcgoose258]

I moved to blender from maya *specifically* because it's bevel was more stable and more intuitive. I was already happy and I'm thrilled to see how much work is still going into improving it. this is awesome

[@MotionApprentice]

Jeez oh man this guy is smart. I have a new appreciation for how complex these dcc apps are.

[@fikr1234]

yes yes yes, this looks amazing! I wish we could get sort of a "minimum viable product", while this gets refined. Merging a prototype into master might also give the devs a better idea of how artists use it and how to solve some of the edge cases.

[@black0ut76]

I asked exactly this at the conference and even in that scenario it would still be a couple of months out, so I think we all gotta just put our patient hats on and let the masters do their work :)

[@btn237]

For bevel, aren’t they all “edge” cases? 😅

[@chimpana]

This is a great example of misunderstanding of MVP. The V is the kicker: viable for what cases? When there are so many variations in use, there isn't a good way to define what 'minimum viable ' means... you just end up with loads of frustrated users who believe they're trying to do something simple and obvious (i.e. that should be part of 'minimum viable' scope).

[@noc2_art]

Howard is the mind behind the new Boolean system which came just in time. I've been saying thanx to Howard and his teams for it ever since. So, with Bevel V2, we're in absolutely good hands. God speed Howard 👍🤗

[@marthinus.x]

I have to say thank you to the amazing contributors who spend so much of their precious time making Blender even more amazing.

[@IrocZIV]

Glad such talented/smart people are working on Blender.

[@CybranM]

Really looking forward to when this gets added to stable, Bevel modifier is one of the best features in all of Blender, allowing for non-destructive workflow. Only issue is, like he talked about, those pesky UVs

[@TTTristan1]

Yes!!! As a user that frequently works in 3d text, I've been waiting so long for this.

[@artro398]

Great stuff, guys! Definitely looking forward to seeing this in Blender!

[@peterallely5417]

This is massively exciting, and I hope my saying’s so is encouraging to your efforts, lads. This would save so much tedious work in mitigation and cleanup, and allow way more time and energy to just being creative. Thanks for all you do.

[@mariokotlar303]

In the last 3 months of work, having this would have saved me a whole month I reckon.

[@disruptive_innovator]

I'm impressed that Blender bevel even tries to consider the UV coordinates. I would have just said new bevel is new geometry. Redo your UVs.

[@3dpprofessor]

I can not wait until this feature works. I do so much blender for 3D modeling, and having tight and clean bevels would improve so many things.

[@stefanguiton]

Superb work! This is a great improvement

[@gabrielmoro3d]

That's so exciting! Not only implementing the function but fixing its flaws at the same time. This is going to save me so much time on modeling. Amazing, thanks to the devs!

[@qulzam685]

before that, i could never imagine this apparently simple thing had so many layers of complexity, i have huge respect for the developers after learning that, godspeed to them

[@galiafvelikiy4615]

it is just a Perfect Example of How Great Brains of Human beings work Together for GOOD! Thank you Blender Team!

[@World_Theory]

Excited to see this finished, when it eventually does. My opinion on how to handle cases where there are two or more valid potential solutions to handling a behavior, is to write both, and give the user a switch to choose which one it uses in their case. Also looking forward to having fine control of edges using selections, and preferably edge methods using selections too. Sounds terribly tricky to do all this though.

[@cekuhnen]

One of my most fav developers !!! Bevel I use to much as a product designer !

[@nirmansarkar]

Oh God YES!! Let there be Bevel.

[@RobbieTilton]

legendary! Thank you Howard!

[@olliveraira6122]

Where is the issue tracker & branch on the maniphest for this?

[@vstreet7583]

Absolutely fascinating. A great presentation. Thank you. Dg

[@sisqosnew]

Thank you for what you did.

[@Mr_Steve3D]

Where can I get current updates on this? please>?

[@Kavukamari]

was just thinking about this yesterday, it's more difficult than it looks I also want a "bezier profile" input gui module, but that's only marginally relevant (custom bevel profile)

[@BrentonWoods774]

I need this node now

[@FinalComp]

I would exchange all the eevee next new features for the bevel node in geometry nodes.

[@RunTheTape]

I'll probably stick to CAD programs for hard surface until this comes along. Looking pretty good on paper, gotta admit.

[@btn237]

How do CAD programs handle bevelling? I hear it’s a lot better but how does that work compared to Blender’s way?

[@RunTheTape]

@@btn237 try Fusion 360 for yourself and prepare your mind to be blown.

[@alihms]

@@btn237I am a CAD user not a developer, so take this with a heap of salt. CAD programs are Nurbs based modelling while Blender is polygon based. In Nurb based modelling, the geometries are mathematically accurate surfaces and splines. So, it is (relatively) easy to maintain curvature continuity when merging two surfaces together when you have the precise equations available. Polygon based geometries are just that; set of polygons in 3D space. You'll run into all sorts of discontinuity problems as described in this talk when merging two or more surfaces together.

[@spaceman-pe5je]

@Zeb sure, but also, for many use cases topology does not matter. a workflow in game art which is becoming very common is to model in CAD like Fusion, use Moi3D to export - because Moi3D has nice geo export, not "you can subdivide this" type of export but it will shade perfectly and then it can be used for baking. often dynamesh first to add any sculpted details or use a polish pass for tiny edges that are not manually filleted. all my point is that these CAD programs are seriously being used for AAA production in the case of first-person weapons at least, high stakes assets.

[@spaceman-pe5je]

@@huydinh3395 I can't really point you to anything specific. I'm in some online communities for game art and have myself witnessed people working at big companies like Treyarch talk about how they are transitioning to Fusion for the weapons team for example. You'll find on Artstation that posts for weapon art will sometimes show they were made in fusion in the tags at the bottom.

[@anthonylosego]

Part of the concern with bevel is that you are crossing two concepts, a mathematical function with artistic vision. In all cases that require creative artistry, it should be up to the artist of how that is handled. In some instances, some edges may not be included just because they fit the mathematical requirements. Where as in some situations, the artist may yet chose and altogether method entirely for the outlier confluences.

[@MorganWable]

I suppose it makes sense that it'll probably be a long while before it's ready. If you think about it, ALL beveling scenarios are edge cases. 🥁🥁📀

[@hnx9083]

This is a question that troubles many people. It would be great to develop a mesh fusion tool.

[@albarnie1168]

so exited for better bevel

[@anthonylosego]

new node. Mathematical 3D textures to 2D UV space node. Map textures to 3D space and then optimize to a 2D projection space at the end of the stack. That way bevel just works on geometry and the texture is through the 3D space. Only convert to 2D UV at pre-render time.

[@ed_halley]

It feels like the current Bevel and the Extrude Profile Along Curve code have a lot of similarities, maybe even share functions. My biggest annoyance with Extrude Profile Along Curve is the complete lack of mitering and miter options. If you extrude a circle along a curve, the profile of the CORNERS is a circle, but what people really want is the profile of the LINE SEGMENTS to be a circle. What you get today is a squished oval affected by the sharpness of any bends along the curve.

[@RandomDude453]

Thank you, sir!

[@orphydiancg7759]

so the actual tool and modifier will get forgotten as always?

[@kangsan2014]

What if the mesh started out as vector and then when you want to bake it, it turns into quad mesh? That would be a good way to resolve quad mesh issues.

[@NEVERTRUMP-p3z]

My Fav Talk ❤

[@raulgalets]

oh boy did I fell in the blender rabbit hole at the correct time innit

[@davidrozier1126]

I think Face Inset should be a seperate node so we don't have to think about an Inset node in the future.

[@eshwarthegreat]

Awesome 👍

[@MaxPuliero]

amazing.

[@patrickwilson3d]

Oh this is great!

[@TheLordDino]

I know it's different but bevel it's the same as chamfer operation in CAD software that can be done by o 2 clicks. It sucks the amount of complexity that polygons modelling tools need to do a simply bevel. (BTw i work with Blender and Solidworks, I know both realities.)

[@sebbosebbo9794]

great. . .. text Modul need love to.... thx both devs....❤🧡💛 amazing .... thx for this talk...

[@thelawgameplaywithcommenta2654]

MY Man!

[@DigitalCapeTown]

What are the CAD softwares doing that allow their bevel functions to operate so 'smoothly'?

[@NoTengoIdeaGuey]

They're using NURBS not polygons. It's like the difference between using pixels in Photoshop and Vectors in illustrator. A NURBS algorithm uses fewer control points and interpolates between then with formulas to create the surface. Polygon modeling allows the user to access each individual control point. They have to do the interpolation in real time in the viewport and resolve the surfaces and how they interact.

[@anthonylosego]

By the way Howard, if you still would like some coding help, I'm still up for assisting. These cold months are going to get me stuck inside with extra time in the evenings.

[@YT775]

YES! YEEEEES!

[@lwdesign3743]

Wouldn't it be convenient to get rid of "Modifiers" and move all functions into Geometry Nodes? :D

[@marsmotion]

not for everyone and modifiers represent an easier initial user experience to get into for new people. imho

[@sanketsbrush]

what is he looking behind the camera ? what's behind camera ?

[@cnucnu3462]

wow，were to download this geometry node group？

[@YT775]

Is there a build available somewhere?

[@MoveBlocker]

do they do it yet?

[@NikitaBukoros]

Any updates on this?

[@JoeRussell-oj7xm]

Sadly it seems not, the developer thread for this project last had an update a year ago.

[@stephenwlodarczyk175]

I can't wait till I can stick my finger into the brain I want to download knowledge from... It still looks like gibberish to me but so nice to see how smart some people can be. One day I hope.

[@karaloop9544]

That sounds like something Dahmer would say. ;)

[@MetalGearMk3]

Has this landed in 3.6.3?

[@everr.r]

that guy with the long grey hair is looking at us 👁

[@rudypieplenbosch6752]

Wow would be great to get rid of this overlapping, that is a very annoying thing about bevels.

[@pauliusmscichauskas558]

I teach blender in an academy for Architects and interior designers. I have some requests: 1) Completely remove the "Clamp overlap" option. It's *completely* useless, and should never be turned on. All it does is prevents bevel from working. If the bevel modifier gives bad results, it becomes obvious that the user needs to fix their mesh and where the problem is. Communicating that is important. And sometimes the errors are so small, that they do not matter in a render. My students often get confused about why the bevel modifier doesn't work at all. 2) "Harden normals" should be turned ON by default. There are barely any scenarios where it would be turned off. 3) Instead of "Harden normals" giving a warning about autosmooth, it should just override it. The user is just needlessly given an extra step to do.

[@mod3528]

These are remarks of a very high level. Right now, development is still going on at a very low level of the question, something like: should it even work?)

[@pauliusmscichauskas558]

@@mod3528 These are things that can be done, very easily, on the current bevel modifier, as the first step of improvement. The new bevel refactor can come in any time later.

[@mod3528]

@@pauliusmscichauskas558 Then it's just a suggestion. This particular presentation was about the new bevel mechanism. This is not a refactoring.

[@pauliusmscichauskas558]

@@mod3528 By "Refactor" i mean "Rewrite". They are writing new code on how bevel works, and throwing the old code out, or leaving it as a "Performance option". Primarily as a bevel node for Geometry nodes, but the same can be used for the tool and modifier. Am I wrong?...

[@mod3528]

@@pauliusmscichauskas558 They don't throw anything away. This is not a new modifier. This is a node for geometry nodes.