Developer Meeting Notes, April 15, 2012
The main news this week: a 2.63 test build is now available, and Daniel Genrich will work as a Blender Development Fund developer on smoke simulations! Awesome.
Hi all,
Here’s the notes from today’s IRC meeting.
1) Blender 2.63 release
- New test build is online since yesterday.
- A long review of status was being discussed, especially related to stability (loads of bug reports) and BMesh. A proposal to make an official Release Candidate this week didn’t get enough support.
- Accepted proposal: we wait for the main bug fixers Sergey/Campbell/Brecht to give in majority green signals, then we make an RC build immediate. This might happen within a week, we’ll check next sunday again.
- Release log needs work, for BMesh especially!
- Thomas Dinges updated the project schedule.
2) other projects
- Daniel Genrich accepted a Development Fund job on improving smoke simulation, especially for animated objects inside domains.
- ‘Tiles’ composite project is a target for Blender 2.64 (still :).
- Sergey Sharybin checks on greenscreen keying (nodes) improvements, Bob Holcomb will check on this too.
3) Google Summer of Code
- Deadline for reviews is this friday. Monday after the announcement will happen.
Thanks,
-Ton-
Great work guys. I have one question. How come despite all the works and improvements, Blender sculpting does not seem as smooth or as fast or handle more polygons despite 64-bit over the 32 bit Sculptris or 32-bit ZBrush. Is it the algorithm?
sculptris and zbrush were developed from the start to be sculpting programs to be able to handle high poly counts, nothing more. Sculpting in blender was developed as, well, and afterthought. it’s a modeling, animation, rendering, texturing, simulation, game making, etc program.
In essence a jack of all trades, master of none. Which is what it’s supposed to be. Other commercial programs (such as Maya) have sculpting modes as well but they too are no where near as good as zbrush or sculptris. I doubt blender or maya or softimage or Houdini or (insert preferred package here) will ever be as good as zbrush, sculptris, mudbox or other programs that pop up specifically designed to be sculpting programs. They are just built for two separate purposes.
don’t forget you can multiply blender sculpting smoothness if you disable the double sided, and you enable VBO, it is very important!
“don’t forget you can multiply blender sculpting smoothness if you disable the double sided, and you enable VBO, it is very important!”Well is it possible to have a bit more info about VBO? and also the effect of desabling the double sided state?
Thanks in advance
yves
Are you running Blender on Windows? It’s well-known that it runs better on Linux.
Make Sure you are using the multires modifier, not just subdividing, multires is built specifically for sculpting, also make sure you have a good base mesh and not just start with a cube
Hi, i’m a max and maya user and i can tell you that the sculpting in those is awful compared to blender.Blender actually has good
modeling brushes compared to the big apps.
Blender is definitely better than Maya. I would even say that I prefer it to Mudbox. Maybe it cannot display as many polygons (it can still display enough to make good sculpture), but the feeling of the brushes is much better to me. And you can use remesh modifier which turns Blender in a sort of Voxel Sculpting tool, and that’s really great.
I totally agree that Blender’s sculpting is in a whole other league than maya and max. It really works terrific. I just wish it could handel higher polygon counts so that it could be used to sculpt in surface details.
But for general sculpting it is terrific. Somebody is developing masking for Blender’s sculpting which will take it to a whole new level.
I agree that Blender’s sculpting is terrific. One can create high quality sculpts with it easily on par with Zbrush or Mudbox. Blender can handle more polygons than Sculptris. Moreover, Unlimited Clay–if continued by Farsthary–will provide similar, dynamic tessellation capability.
As for the polygon limit, my budget ($750ish) home-built PC can sculpt butter smooth with tens of millions of polys. You will never need more than that for a sculpt of even the highest level of detail.
The BIGGEST downside to Blender sculpt mode compared to others is the “grab brush”. The importance of the move/grab brush in digital sculpting cannot be overstated. Blender’s does not work well to modify proportions at medium to high poly levels. It is difficult to deform sculpts without create creases/valleys with each brush stroke. Zbrush, Mudbox, and Sculptris seem to use a different algorithm that more effectively blends with geometry outside the brush radius. It’s difficult to describe. I’d suggest anyone who’s interested to grab a trial of Zbrush or Mudbox or download Sculptris for free and experience the different behavior between the move brush in those apps compared to Blender. This is by no means an exaggeration. The lack of a proper move brush makes sculpting in Blender a great deal more difficult and tedious.
What is your computer’s specs Iso? I would love to reach tens of millions of polygons. At the moment on my pc the lagging starts at 1 million. I would have though that my gtx 580 would help but it does not.
Why dont you just use the grab/move on your lower subdiv levels to fix the problem? Why do you need it on higher levels?
AMD FX6100 CPU
HD 6850 GPU
16 GB Ram
I just did a quick test with a constant brush diameter about 1/3 the total diameter of my sphere object. I can get up to about 25 million before experiencing hindering amounts of lag. This can be solved simply by decreasing the brush radius. At such levels you may only need a large brush radius for broad texturing. In such case, anchored stroke seems to be very reliable. If we every get a texture overlay tool as seen in Nicholas Bishop’s vertex paint work, it would be possible to apply broad areas of texture with a smaller, less lag inducing brush radius.
As for my settings, I have VBOs on. Doubled sided mesh has little apparent impact on performance. I’m only using two of the three opengl lights, which may make no difference at all. Use solid viewport shading. Textured viewport shading with a matcap (angular map set to normal coordinate space) will greatly limit sculpting performance.
I’ve experience problems where sculpting with even a small brush radius at around 1.5 million polys becomes laggy. It turned out to the the mesh, though I don’t understand why. I started again from a cube, and did not encounter the same lag issues until around 20 – 30 million polys.
Nicholas Bishop wrote some partial visibility code. Builds for that are currently available on graphicall.org. This can make a huge difference. For instance, my laptop with intel integrated graphics (gma hd i think?) cannot sculpt effectively above 300k polys. However, with the partial visibility tool, I could isolate a section of my model and sculpt with the smoothness of a model around 100k polys, where the total poly count of my object can well exceed 2 million polys.
I suspect the problems you are suffering come from your base mesh. I don’t know why I’ve had poor performance with some meshes compared to others. If you still experience horrendous lag, try projecting new topology to your sculpt. You need only do a very rough retopology, as in basic loops and regions of high detail but nothing terribly accurate. Use face snapping onto the surface of you existing sculpt. Next, add a multires modifier to your new topology. Add a shrinkwrap modifier and play around with the settings to get things just right. Personally, I’ve had best results with the “project” method. Now, subdivide your mesh via the previously added multires modifier until your new basemesh has enough polys to look like your original sculpt. Apply the shrinkwrap modifier. Thus, you have a retopologized sculpt with multiple levels of subdivision. This will help you add detail in certain areas. It may also be a way of fixing a mesh that, as I’ve said, for some reason I can’t explain causes excessive lag.
The grab brush functions the same on all resolutions. Sculpting on low resolutions only makes its flaws less noticeable. Nonetheless, these flaws become a real problem around 10k polys, which is more to the low end than medium. If you have a model at a resolution of 10k polys or less, it can be difficult to gauge proportions of finer elements likes eyes. The problem rigidity of the grab brush cannot be solved by simply lowering the resolution.
I hope this helps you. And I am not, by an means, saying that Blender’s sculpt mode is an equivalent to Zbrush or Mudbox. I am saying that it is very capable, perhaps under appreciated. In the hands of a capable artist, it can produce results every bit as impressive as the industry leading sculpting apps. Some aspects, however, make it more difficult to achieve that end. I don’t expect Blender to ever replace Zbrush. That’s why I own Zbrush :)
Thanks Iso.
That is nuts, I expected your machine to be a beast as I have a q8300, 8 GB Ram, GTX 580 which is not thaat baad… I will have to look into it. Might be ram size and speed
I know Blender can’t replace Zbrush. Zbrush is a very unique and special tool with all the insane things it can do. Pixologic really do some crazy groundbaking things with every release. I will never part with Zbrush.
That doesn’t mean I don’t have a special place in my hart for Blender and specifically Blender sculpting. It has it’s uses. And I way prefer it to sculptris.
you kiddin homie Zbrush kicks that open source crap blender in da ass
Obvious Troll is obvious.
hahaa! when we get unlimited clay you will just shut up, hahaa!
Can your ZBrush do this?
http://www.miikahweb.com/en/blog/category/Development
So close yet so far. Let’s go community. Start testing so we can have this awesome release soon.
Bmesh needs a LOT more work before it can be released for serious work (imho), I do work with Blender in the ad-agency I work for, and Blender is wonderful as always, but the tests I´ve done on the side with Bmesh, indicates that it can not be used for professional work and should remain in the testing phase until it can cope with modifiers like Bevel and SubDiv-(SubSurf). Simple subsurf works, but too many “ngons” and bye bye smoothness and hello tris & pentagon stars… After a while, you´ll get holes in the mesh all over.
Other than that, I´m looking forward to Bmesh as a kid in a candy store, but I appreciate the incredibly hard work that must go to make Bmesh useful for work, I take it that coding this must be some kind of hero work or coders-nightmare.
That subsurfs don’t work when you have many ngons is not really a bug. You just shouldn’t do that. Just like you shouldn’t use tris when you use subsurfs. You can use one here and there to solve some topological issues, but other than that, you should just use quads when you plan on using subsurfs.
I didn’t get the point of Bmesh either, until I watched Jonathan Williamson’s tutorial on it. He says that on a smooth mesh, you will probably want to use ngons only as intermediate solutions, to make changes to your topology, without having to rip up your entire mesh. He says there probably shouldn’t be any ngons in your final high-poly mesh. An exception he mentions are architectural and engineering projects, where you have many planar structures. There, the ngons can safely be left in place.
I disagree that NGons are only good for planar objects. NGons can also work on characters. As with the mesh in general you should be aware of how a NGon affects surface curvature with subsurf.
I use this modeling a lot and do not have any issue or limitations – it rather finally pushed Blender a step further.
I think this idea about NGons being bad or only useable for re topology should be re-evaluated. It is a modeling tool which also has to be employed correctly – as everything else.
That is very interesting to hear. Do you have a model somewhere to show this. How does ngon mesh behave under deformation?
while bevel is broken but it is been fixed but Subsurf is working correctly. When you use catmull/clark subsurf ngons and triangles are tessellated into quads. that tessellation will result in poles been created and that is where pinching tends to occur. If your surface is not really planar or will deform in some ways than its best not to leave ngons or triangles in place.
You can use select by number of verts to find ngons and triangles and clean up your mesh.
Very true. I would like to add my two cents that may help dispel some of the bad, undeserved reputation poles have. Poles are not bad; they are essential to controlling the flow of face loops–edge loops is a bit of a misnomer, as it only refers to the boarder between parallel face loops. The entire point to using Subsurf is to control high levels of detail with a low resolution “cage” mesh. The only way to achieve that is with loops. The only way to direct loops is with poles.
Furthermore, ngons are converted into triangles prior to subdivision. All triangles are divided in the same manner as quads. There is no separate algorithm to handle tris. Triangles effectively subdivide into three quads. Poles result that guide face loops in crossing paths. Imagine a “T” divided in half along the vertical axis, then moved horizontally such that the right side of the vertical column of the “T” is now on the left side and vice versa for the left column. This creates a change in the flow of geometry one may call “pinching”. Really it is nothing more than unintended face loops. An identical result could be produced by having quads flow in the same manner in the cage mesh.
My point is poles and triangles (tessellated ngons) are not bad. Poles are essential. Triangles are not intrinsically bad. They are just less predictable than quads. Therefore, it is advantageous to–and, honestly, inexcusable not to–model your subdivision cage in ALL quads.
Uhm BMesh and surbsurf are not working? Hello? I am product designer and finally Blender can do the same as T-Splines. I am not sure what you tested BMesh on but it works great and not really slows down or makes subsurf unusable.
I have just done a short test, and I have not yet tested the new features, but I am happy to see that UV mapping display that was broken on some builds when in Edit mode on meshes with ShapeKeys, works again fine ! Thank you very much !
Hello to all. Firslty thanks to Ton and all the Devs that are working so hard. Blender is really becoming a monster :).
I know that many people are waiting for Bmesh. I saw some videos of how to use it in your worflow but I think I’ll have to use it to really grasp its potential.
But I would like to know if there are future plans for Cycles? I’ll be frank that for me Cycles is one of the best new feature that have been added to Blender. I was using Yafaray before Cycles, but with Cycles ease of use and the result I have got its just awesome. I have used it until now with CPU but I’ll get my new graphic card this week.
I would just like to know if features like volumetric lighting and rendering support like using GPU and CPU at the same time are being considered. Also at the beginning of Cycles, some filters were available to enhance images, is that gonna come again? I know that many are going to say that Cycles now can work with the composite nodes for those effects. Even if I have been able to understand some, I think that like me there are a few that are not as good at it as many others.
Nice day to you all.
Read this interview: http://www.blenderdiplom.com/index.php/interviews/item/84-interview-brecht-van-lommel-on-cycles
Cycles will NEVER compete with commercial renderer or yafaray (SAD). The only cycles developer is Brecht, he’s working 50% on cycles (I think less than 50%) and his purpose for cycles is not to create a ” Hollywood” renderer but just a modern blender internal easy to use. So don’t have too much expectation on cycles. I think cycles can replace BI (all functionality) in 5 or more years at this rate, but will never compete with other renderer in performance and features.
Although it is true that Brecht only spends half his time working on Cycles and Cycles is not intended for big studios. You are completely underestimating the potential and the power it already has. If you look at the rate of development I am pretty sure it will be full featured in 2 years.
All the features I am waiting for (volume, motion blur, sss, hair) is scheduled for this year( according to the roadmap http://code.blender.org/index.php/2012/01/cycles-roadmap/).
Cycles has a very bright future.
I’m not understimating potential. I just think that BF and Brecht will not develop cycles as a competitive renderer. Even small studios needs renderers better than the actual cycles + volume,motion blur,sss and hair. Cycles is just a good renderer for amateurs or very small studios who need to make objects shining. Read this post to get an idea of what cycles need to be taken into account by a small studio. http://blenderartists.org/forum/showthread.php?243645-Cycles-HDR-texture-support&p=2052170&viewfull=1#post2052170Cycles is great even now, it has big potential, Brecht is awesome and kind, BF SUPERGREAT but don’t think it will change your life because “becoming a free Vray, MentalRay,Pixar Renderman,Arnold…yafaray” competitor is not in the BF/Brecht purposes, and cycles is not a BF priority just because it needs a lot of money/people and I think they prefer to fund many projects instead of 1 big project. But this is just my opinion so if BF will act otherwise…well… I will be superhappy like all blender community.
“who need to make objects shining.”… come on.
I don’t know. Give Cycles to capable enough hands, and we can see some pretty amazing stuff produced with Cycles. True, Cycles is not as powerful technical-wise as say, Octane Render or LuxRender, but it’s growing and it’s already got some potential to produce more than just make objects shining. I’ve seen people do amazing things with just the Blender Internal renderer (Nature Academy, anyone?), and as time goes by, I’m seeing more and more talented usage of Cycles that makes me wonder what’s truly possible to do with it, in the right hands. Give us a limit, and we’ll stuff it with awesomeness.
This is like trying to guess the future. You can always pay for you personal developers to get the tools you need. That’s the potencial of libre software. As a small or medium studio one could seriously consider to pay for a full featured developer and at the same time contribute to the whole world. The mind change have to come from everyone in a first place.
I like your thinking Chaos.
This is obviously not the easiest route as aposed to just paying for licenses and having the licenses available immediately. It will benefit many though.
Hello thanks for the reply. I have read the link that you have given. It is true that Brecht is only using half of his time to code for Blender and that he admits that Cycles is not meant to become an Holly wood render engine.
But reading the article to the end gave me hope that there are still great project for it. Like making it work with OSL (Open Shading Language), and using preset for materials and make it behave better with real life lighting.
So I am quite confident that Cycles will still progress.
Well I also hope that with the new tools coming that small studio are going to turn to Blender even more and that in overall Blender itself is gonna progress.
I personally still hope for the return of advanced particle functions that went missing after the first 2.5 releases.
Lukas Toene is great! So it’s http://graphicall.org/905 better and will replace all that cool 2.49 particles’ features.
Daniel Genric,in developing smoke,make sure that domain cage can be deactivated in layers and still see the simulation in OpenGL.That will Rock Blender!
Any news regarding image sequence textures? I want it bad… :)
The come back of Absolute Shape Keys is a great thing ! It comes right when I need it the most !
Congratulations Daniel!
Hi, can someone please explain what the ‘Tiles’ composite project is? Thx
Nevermind, just found this.
http://ocl.atmind.nl/doku.php?id=design:proposal:compositor-redesign
you should try getting a build. It’s really cool!
Just checked out the new testbuild. I think bmesh is almost definitely ready for production-type work.
Great stuff from the developers!
Hi
This thread provides links to test builds. But there are no announcements on the home page (right hand column).
An announcement will attract the attention of more beta testers…
Just my two cents.
Haven’t been able to start the last 2 test versions. I have:
Model Name: iMac
Model Identifier: iMac9,1
Processor Name: Intel Core 2 Duo
Processor Speed: 2.66 GHz
Number Of Processors: 1
Total Number Of Cores: 2
Will my machine be supported when the official release comes out?
Thanks,
Pam
What do you mean when you say it doesn’t start. If you try to start blender from a terminal window, does it gove any errors? You should file a bug report asap.
I was thinking that perhaps there were no versions for my computer. Here’s the Problem Report:
Process: blender [424]
Path: /Users/phampton/Downloads/blender-2.63-testbuild2-OSX_10.5_i386/blender.app/Contents/MacOS/blender
Identifier: org.blenderfoundation.blender
Version: ??? (???)
Code Type: X86 (Native)
Parent Process: launchd [102]
Interval Since Last Report: 276241 sec
Crashes Since Last Report: 25
Per-App Interval Since Last Report: 0 sec
Per-App Crashes Since Last Report: 21
Date/Time: 2012-04-17 09:47:23.583 -0700
OS Version: Mac OS X 10.5.8 (9L31a)
Report Version: 6
Anonymous UUID: 19FC7D76-842A-47AD-8864-FB0707810738
Exception Type: EXC_BREAKPOINT (SIGTRAP)
Exception Codes: 0x0000000000000002, 0x0000000000000000
Crashed Thread: 0
Dyld Error Message:
Symbol not found: __ZNKSt9bad_alloc4whatEv
Referenced from: /Users/phampton/Downloads/blender-2.63-testbuild2-OSX_10.5_i386/blender.app/Contents/MacOS/blender
Expected in: /usr/lib/libstdc++.6.dylib
I’m not savvy enough to know what this means.
Thanks for reading,
Pam
bmesh big problem now is that it counts hiden meshes in same object data and cuts it if thats intersects in camera view with visible geometry. Thanks for good work. looking forward to that
Man, if you know about a bug you ought to report it here –
http://projects.blender.org/tracker/ it’s pretty simple and helps Developers!
ton, cycles still has no movie clip texture, then, still unable to reconstruction of camera tracking in cycles
(Subscription test… either Disqus didn’t forward anything to me from this thread today, or my email is not responding to the subscribe)