Blender Bevel

This is what bevel should be doing to a cube…

  • Count the edgeloops: three per edge.
  • Count the tris: none.But Blender 2.46 and even 2.47 can’t do this?

Why not? 2.44 could do it! (how? run single bevel twice)

Until Blender gets the ability to do this most basic of “correct” bevels, there is no way in hell my Maya colleagues are even going to consider taking Blender seriously as a Modeling tool. Proper edgelooping as such is one of the main things they’re looking for when judging entries to their speed modeling contests.

They would tell me that an “industry standard” bevel creates… three edges per sharp edge (e.g. two each side which allows smoothing, whereas four is excessive) and no tris.

Why can Blender no longer do this?

…or can it? If so, please point the way.

I suspect the bevel behavior has changed because in earlier versions of blender, you could only bevel the whole mesh and not a few selected edges…

PS: please don’t compare blender to maya or other 3d packages… it makes you and your request lamer… while i don’t have the most certain answer, i doubt a professional would help you if you approach them with this attitude…


Cheers dyf, and thanks heaps for at least answering. I don’t consider this request to be “lame” in the least. I’m using Blender in real world situations where I’m working alongside (in some cases competing against) others who are using those other packages.

I have raised this question before… without the comparison. Only one person looked into replying and their solutions didn’t actually work when applied.

I’m thinking this is a serious modeling flaw which has crept in to blender’s last few releases. I thought that surely the devs must be aware of it, so I let it lie for a while to see whether it would be fixed on the next overhaul release. It is not.

  • Bevel is key to making good renders on non organic models.
  • Tris are universally recognised as evil.
  • Tris don’t matter so much on hard (non organic) models but sometimes they do. Keeping edgeloops to the minimum needed for smoothing should be the goal for a basic bevel script, though I can’t seem to achieve this on the three different methods I can see to apply a bevel in Blender.I’m hoping someone out there might do a lot of hard (non organic) modeling and must know a workaround for what I’m seeking. You know… something obvious that makes my question look stupid… I would be grateful :smiley: because it would really speed up a few projects lately where I’ve spent hours on beveling by hand. :frowning:

…maybe it’s time for me to submit this on the Blender bug tracker?

Why wait, submit away!

Okay, Atom… done.

Submitted Bug#17520

I hope they don’t reject it as a feature request. This one’s been bugging me for ages.

this proposal looks nice and simple… good luck with your request

It’s possible to do by hand, but it’s a royal pain in the keester.

Add a cube, make a double loop cut in each direction, scale the loops out to near the edge of the cube. Select the corner vertices, scale them in a bit. Select the large faces, scale them out a little bit. I suppose you could put numbers on that scaling in and out part to make the method more accurate. In fact, someone could even write a script to implement this, then there would be a decent bevel a cube script.

I’m sure the general case is more complicated than just coding this by-hand algorithm. You’d have to move the large faces out along their normals instead of scaling. You’d have to worry about the four-pole and five-pole and n-pole case. But if it was working before, the code is out there somewhere, isn’t it?

Well, to be fair, the “before” method required bevel to be used twice… with the second time dragging until the very-edge faces squashed flat. Now there seems to be some safeguard in that the edges just plain stop when the first (corner) vertices meet each other, so the very edge edgeloops can never meet. I included two possible fixes for this on the actual bug report. It’s definitely a pain because I’m wanting to participate in generic (any 3D application) speed challenges, and “proper” edgelooping is one of the first things they look at. That, and my day to day modeling in general.

This problem recently made a one hour fridge turn into the best part of a days work… all because the edges had to be right.

Edit: Wo-hoo! The bugtracker has been assigned to somebody… hopefully this means it’s actually going to be addressed. :eyebrowlift2:

um, yeah the code will be out there.
if you want to use the old script from 2.44 (i do!)
Make a copy of the Script.
(open in text editor)
change the name in the header to Bevel1 or in this case Bevel Best 244.
Save the file as to your .blender/scripts folder.
now when you go to mesh in the scripts menu.
update menus
you should have Bevel Best 244 as a new menu entry.

Actually they are not. Their usage is recommended in low-poly organic models to improve joint deformation.

Tris don’t matter so much on hard (non organic) models but sometimes they do.

Can you give an example?

Thanks for that. I haven’t found the script though. It might be that I’ve use a pre compiled version of Blender… I just used the “W” key bevel.

Hi eye208. I’m not sure whether it’s wise for me to entertain a debate on this. I was meaning the kind of modeling I have my diploma in, and where triangles are very bad in the “if I find any you repeat the assignment” sense. I think you may have an exception in mind of the sort where bevel would no longer be practical.

Curved surfaces.

A diploma in modeling, hm?

I still don’t see why edge loops matter in solid non-organic models. However I do know that you can’t have “no poles” and “no triangles” at the same time. Looking at your sample picture, I prefer the bevel with triangles.

Of course you can be anal about it and tesselate any solid triangle into quads (just to have them re-tesselated to even smaller triangles again at the render stage). But frankly, I don’t see the point. People like Ray Baitt and Steven Stahlberg are well-known for using triangles even in organic models. For some reason they seem to get away with it. Not sure about their diplomas though …

Technically my diploma was more for Animation and Lighting, in that I majored in those, although Modeling was an important part. I didn’t really want to bring it up except that your questions were heading into areas where I could easily slip into an off-topic essay on, but then I would rather debate such points in separate discussion thread.

The reason I don’t want to use the “newer” bevel is because…

  • The triangles have broken the edge flow. This means I cannot simply “add edgeloop” within the bevel like with CRTL-R type function. The new loop would stop at the first corner triangle it hits so you would have to do every edge separately taking lots more time.
  • Edgeloop slide (if I want to manipilate the bevels) also distorts on the cube corners because the edgeloops are not clean.
  • The use of triangles is dangerous because they can sometimes show up under certain lighting conditions, especially on curves which is essentially what you would expect on a bevel. It’s known as face popping. Doesn’t happen quite as much with quads.
  • Ideally, I want to work with the least number of vertices / edges as you can to get a given shape. The new bevel with its four edges to each edge is excessive. Two is not enough because I should be able to apply “smooth” (or subdiv modifier) and still maintain the same shaped mesh… only with odd numbers of edgeloops does the corner / middle edge stay put on smoothing.
  • On the old method, I could remove edgeloops to get pretty close to my original mesh (slightly smaller admitted) whereas the edges of the newer bevel have been destructive to the original topology.
  • Probably if you play with it enough on edge by edge basis (similar to the workflow pointed out by Orinoco) you may be able to achieve similar effects… but why? It’s a huge waste of the companies time if I’m manipulating by hand all day when the bevel could & should follow basic edgeflow rules. The above was just a cube… what happens when I need to bevel a bookshelf? …like I said, it’s probably not good for me to start explaining / debating all this kind of stuff, but those are some of my reasons for prefering the older bevel.

There are many styles and a few do break conventions as you point out. In my case, I’m working with people who have what they would call an accepted “industry standard”. If I was to use the newer bevel in their pipeline, it simply wouldn’t make the cut.

I give the point to Lancer on this one.

Of course Lancer has a point. Anyone preferring Maya over Blender has a point.

But dyf has a point too. If you think Blender does things the wrong way, then by all means go and use Maya. If adding edge loops to bevels is an essential part of your workflow, go and use Maya.

For some reason Blender’s bevel works fine for me. I use it all the time. I’m sure it can handle bookshelves. I see it being used a lot by other people too. It’s not CAD-level accurate but it gets the job done. Nothing in Blender is 100% accurate. This is not a tool for product design.

Go ahead and file each feature that works different in Maya as a bug. Maybe in 20 years when all the “bugs” are fixed you will finally get your free Maya clone.


hmm I never used old bevel before cause it’s beveling whole mesh… now I use new bevel, where can I choose bevel by edges - I like it…
maybe it’s not so fancy because it sometimes produce triangles… but for me and my interior modeling is better then the old one
btw - about this tris problem, I think blender haven’t correct bevel tool becasue lack of true Ngons

Dude, you have totally the wrong idea about me. I’m upset about the new bevel, comparing it to the old one… of Blender.

This has nothing to do with my knowing Maya. What you have completely failed to recognise is that amidst a professional level industry in my small but highly competitive country… I’m the only one using Blender. What does that tell you about my preferences?

You have no idea of the battles I have had in terms of even getting hired because I choose the Blender way rather than the pirate Maya way which the general norm seems to be for newly graduated students over here. I could even recite you occasions where tutors sat me down to warn me against my insistence on using legal Open Source, but that’s completely not what I want this thread to turn into.

Who told you to enter this thread and turn it into a discussion about preference of software? You haven’t even correctly identified where my preference is.

That screenshot you posted looked good (looked very nice actually :yes:), but it’s kind of like when someone posts a model of a head they made… it looks good, until we ask to see the wires. Suddenly everyone gets “anal” about having edgeloops in the right place (concentric circles around the mouth… been raised on BlenderArtists countless times). Now, the modeler can either listen or they could say “it looks fine to me, I see no problem”. That problem, would soon become apparent if they needed to modify the shape, like… make the lips pucker. That’s why correct edgeloops (and a few other key principles) are so important. Sure, your model may not need to animate, but it should still be adjustable in anticipation that the client may want to, say, change dimensions… as this is pretty much the same deal. Again, I’m being an idiot pausing to explain all this because it’s detracting from the question I asked at the beginning of the thread.

That big list in my last post (and there were others) was nothing about which packages I use; it was about how the different bevel can impact on modelling workflow itself. The newer bevel breaks important modeling conventions. Those principles may not concern you at home, as your own boss, when working on your own projects, but they matter very much to me when I have to be accountable to a certain standard expected by an industry client.

Blender used to have a good “full mesh” based bevel. They now have a not-so-good “selective edge” bevel. I don’t mind having both, but they’ve thrown the baby out with the bathwater on switching to the new one. I think the recent changes have slipped a little. It is my hope that Blender maintains the high standard of its past iterations.

I’m fairly new to blender myself, but after a couple of pot shots at inorganic modeling I would agree that this needs to be fixed. I’m really surprised this hasn’t bugged more people, maybe I’m the only newbe that really took the edge loop tut to heart.
anyways, where can I track the progress of this? do I need to get an account over at the

I think you can track progress without an account, but you can’t report bugs you may find. Not that you’ll find many as a newcomer – actual bugs tend to occur in combining features in ways that don’t make much sense to the developers, so they weren’t tested rigourously.

In my three years of blending with the foundation release versions (not bleeding edge SVN builds) I’ve only found two real bugs. One is intermittent and I can’t duplicate it reliably, and the other was fixed within a week.

These dev guys are gooood. But, sometimes, as Lancer’s noticed, they get a little off track.

Hmmm, the link above (or this one: Bug#17520) should give you the bug report. They may have changed things because I think it now requires you to be logged in and I don’t think the tracker used to.

Last bug I reported, the dev at first challenged whether it was a bug. When I posted a sample file as evidence… wham! Bug was fixed overnight. I was incredibly impressed. :eek: