A Visual Guide to BMesh
With the release of Blender 2.63RC1, BMesh is now finally a part of Blender! Mike has published a nice visual overview of the new options.
As you may have noticed, I am somewhat excited about the addition of BMesh into the main Blender trunk in the upcoming release. What you may be wondering is why? In the past I explained what an ngon is, but without seeing them in action their benefit can be a bit hard to understand.
So that is exactly what this post is going to cover, it will demonstrate the new functionality that will be available in Blender 2.63 once BMesh support is added. Each animated image is actually 1080p in size, so click on it to see more details. Some of these features, such as the bridge edge tool, aren’t specific to ngons specifically ( and could be accomplished without BMesh ), however the internal changes to support BMesh made adding these features more feasible.
Link
The bevel an knife tool look efficient, but I hate that Ngons feature : On simple tests, when subdividing an edge, I get aVertex on a duplication of the original edge , not shared with the original one (disconnected faces). I am used to clean modelling with quads, and all I get with these Ngons is a bunch of messy faces that I spent much time converting in tris and quads and clean by hand ! I see mainly drawbacks in this system… Sorry, thats the first impression.
I can see only benefits, it is you who is in control of your mesh, it’s freedom in editing. Besides, it gives the oppertunity for a lot of new tools with a much more flexible API (I assume its flexible :-) ). It stimulates developers to actually develop new tools regarding editing the mesh. That’s my point of view though.
I really don’t get your “subdividing an edge” problem.
When you subdivide an edge, what it should do is add a vertex in the middle, and make n-gons out of the edges.
Then you can connect vertexes with “J” to split N-gon into tris or quads.
No “Vertex on a duplication of he original mesh”.
Just select in Edit mode two vertices of an edge. Hit subdivide button. Then select everything and convert into triangles. Select the vertex created by the subdivide operation and move it : you will notice that only one of the two adjacent faces separated by the originally selected edge has been cut at the middle vertex location. This new vertex and the two tips of the selected edge form an invisible face hidden along the original edge. You have to remove this face, remove the original edge and recreate new faces by hands to fill the hole ! And it is just a cube ! I can’t imagine working with this “tool” !
I just tried and on mine the vertex is connected and I don’t have any extra faces (build r45815). It’s definitely not supposed to do that, so I’d file a bug report if I were you.
Thanks xrg. Trying again, I disabled Beauty button and it worked fine.So I thought it was due to this Beauty fill feature. I tried again with Beauty disabled, and I got again an extra face… I will investigate a bit more, and report a bug if needed.
Reported a bug for you. (#31052)
While I can see some advantages to n-gons, I think it’s going to lead to some nasty surprises, especially for beginners. Even before n-gons, it was all too easy to end up with non-planar quads that gave weird lighting/shading on faces that you thought were flat… but with n-gons the problem is going to increase to the nth degree! I do hope there will be a ‘Make planar’ option somewhere to correct the inevitable mess…
Pretty much all the other 3d softwares have been using n-gons for ages, I think that’s a good thing that Blender goes the same way, it’s much better for editing geometry. And anyone should know that quads are always the best in the end, it’s the first thing you learn at school, or find in 99% of modeling tutorials. I agree that beginners will have to learn that, but that’s our duty to teach them, right? ;)
soooo good to read ur comment freem:) Blender is the only software using just quads for ages, that wasnt a challenge but a real advantage to so many modelling ppls who are intrested in important topolgy dependent workflows. As i know there is still nothing more important then quads consisting of two coplanar triangles for realtime rendering engines, These engines still tesselating evey incomming content to triangls. Maybe im wrong ? I guess theres still no standardisation of tesselating (nonplanar) Ngons
anyhow thx alot Blender Developers, ur realy good imo :)
Graphics cards only see triangles. All other primitive (quads, ngons, etc…) are entirely up to the software. Ngons in Blender are a new kind of solution to surfaces, not really a new kind of surface. One which saves a lot of headache and has been a proven industry standard for many years in other (professional) modeling packages.
ps. most game engines only work with triangle meshes since there’s no need to story any other information about each mesh. Only the most recent of game engines use bezier patches for tesselation (see Unigine heaven).
Thank you very much FishB8 ! I wonder how a so simple to find bug has not been seen before release : The first trial of subdividing a edge of the default cube shows the problem.
It’s not technically a bug. The mesh is still a valid mesh, but it’s not what most people would be expecting. It’s more a usability issue than has potential to inhibit workflow. What actually needs to be done is to rework the tessellation algorithm. (The algorithm that works out how to turn quads & ngons into trianlges)
Campbell closed it since it’s not really a bug, but acknowledged that it is an issue. You can read his response here if you like: http://projects.blender.org/tracker/?func=detail&atid=498&aid=31052&group_id=9
I’m happy that Blender is getting more friendly to the masses:) But when looking on these faces with 5+ verticles I always had this question: are the faces with 5+ vertices still converted to triangles internally to be able to render with OpenGL? At the end, only a 3-point geometry can be flat with any position of the 3 points… So this is basically speeding up the modelling process, but not changing the way computer thinks about our model: it’s still triangles, and designers will still have to remember that in order to create good models of complex geometry.
You’re right. And it was already the case for quads, everything is rendered as triangles.
Is anyone else finding that the bevel just… Doesn’t work very well? Try doing a multi-level bevel on a few edges that make a corner. Mess around with the combinations of ‘Distance’ and ‘Even’. I can’t get a good-looking result at all :/
It’s obvious.
People have got to learn how to report a problem without actually complaining about it. It shows little gratitude for the hard work theses guys put into Blender’s development.
Anyways, nice development, guys. I’ll report any bugs I run across. For the most part, I’m loving the new Knife tool and I’ll checking out the new Cycles additions. Thank you very much!
For my own, even if I am totally conscious of the work behing the tools, I think that I have the right to say that I dont like the “improvements”. I have read “So this is basically speeding up the modelling process” in the lines above, and I don’t agree.
For what I have seen so far with the new modelling system, I have to redo by hands steps after steps the topology broken by Ngons. When creating Ngons, you don’t see what actually happens to the topology only after triangulating, and then, most of the time you can just cry and waste time to fix troubles by hand. A feature I didn’t want at all has just been implemented in Blender without the possibility of disabling it. So, yes, I complain.
Maybe after some correction and with the time I will find it useful, but right now it breaks my workflow and after using Blender for 8 years I feel like a newbie !
So, I say thank you for everything else, but not for Ngons !
A feature that is missing in the new NGONs function is that of being able to display the implicit subdivision of NGONs without actually having to break them apart (into triangles or quads) (3DMax has this display option). In this way, the user would know if he/she really needed to break an NGON face in order to achieve the desired result.
I don’t think there is any implicit subdivision, because you can explicitly subdivide it any way you want.
There is a roundabout way that works for me but only in object mode: Apply a Decimate modifer with no reduction in polys (Ratio: 1.000), adding the Wire option in Draw Extra shows the implicit tessellations in object mode, though not in editmode.
Ngon support and all the needed functions (i love that knife so much) that it will allow sooner or later in Blender are a very welcome addition, making the modelling process even better and making it future-proof. I have enjoyed a lot modelling since several revision when i tried them after the Bmesh merge.
But indeed there are lot of problems so far, some having been mentionned in the comments here, and some in the bug tracker too, but i’m sure these will be fixed and improved with future development version, it has always been the case when something big and new enter the fray, it’s rarely bug-free from the start.
Took several years for ngon support to make Blender more modern, but now finally the big start is there, thanks to everyone that made it possible, for the new era of functionality it will open to Blender.
I have just noticed something else… Ngons not only appeared… but Fgons disappeared ! I miss that feature. It allowed to select a bunch of face simultaneously, but without breaking the topology !
Please, I would really want to be told what are the advantages of Ngons in the modelling process (Not Bmesh, just Ngons). Currently, I only see drawbacks. And all provided demo so far include at the end a warning about risk of getting a bad topology with this tool, so I don’t think that it is only a fear in my head !
I actually agree with you, funny how every new feature ends with a big warning :).
The only good feature imo is the bevel tool… if it works right.
I remember the horrible start I had with 3d modeling, I used to create tons of bad meshes and spend afterwards hours to fix the ngons or created unnecessarily dense meshes. I read documents about the theory how to resolve ngons into 3s and quads etc. Then came blender and taught me to discipline myself how to create correct meshes right from the beginning and think in quads rather then just trying to form a shape. It felt a bit like learning to draw with a pencil or color theory by mixing colors. In a very beneficial way.
I guess (i might be wrong though) the decision by the developers to include ngons (bmesh) was just to make blender more mainstream or popular. That would be a valid point. The bad thing though is that most people think that what the majority does is the correct way.
Anyway, as long as blender doesn’t create ngons without me being aware of it, i’m fine with it. I feel happy though, that i had the opportunity to learn modeling with blender before bmesh :)
“The bad thing though is that most people think that what the majority does is the correct way.” Oh so true
The goal of bmesh is actually to make the mesh system more self aware. Vertexes keep track of what they’re connected to and several other things that are way over my head. It makes it easier to write modeling tools because they don’t have to iterate through the mesh manually to find everything. Ngon support is a bit of a side effect of that.
Not supporting ngons just means it has to either punch a hole in the
mesh or triangulate it whenever Blender encounters a poly with more than 4
sides – neither of those are ideal and it has always been broken functionality – you’ve just
gotten used to it and know to work around it is all.
As you said it took you awhile to learn good modeling techniques without ngons, it’s going to take new users awhile to learn good modeling techniques with ngons. The ngons are pretty irrelevant here – It just takes awhile to learn how to model with good topology is all. :)
What xrg said.
You should use the Knife tool, you will directly see the improvement with the possibility to have n-gons. Also, for objects that you won’t subdivide, or animate, you can just keep ngons without any problem. Because in the end, it’s still rendered and displayed as triangles.
Bmesh works great while modelling. But thinks get rough when you apply a surface modifier
what is exactly wrong with the sub surface modifier?
ngons are awesome :D
But why removing Fgons ? Were they not compatible with Ngons ? They were useful and non destructive on the topology.