Developer Meeting Notes: April 27, 2014
It’s a bit of a slow week; the biggest news is that Blender is moving to Python 3.4.
Ton Roosendaal writes:Hi all,
Here are the notes from today’s meeting in irc.freenode.net #blendercoders.
1) Blender 2.71 targets & planning
- Planning and targets overview.
- Campbell Barton: we move to Python 3.4 this week. The platform maintainers have been notified for it. (windows, osx)
- The planning was to move to BCon3 today – but that means all branches and release targets should be applied to git master. Paint tools, Cycles Bake and Freestyle texture brush is still missing.
- After careful deliberations, meeting agreed on moving the release planning a week.
2) Google Summer of Code and other projects
- Reminder for students and mentors to get their proposal in our wiki!
- Overview page.
Thanks,
-Ton-
As a 3D sculptor, Blender is simply excellent. Dyntopo and Multi-res have come a long way, and challenge the leading industry standards. However, when it comes time to add color, we still have to go through the process of retopo beforehand, and we don’t have symmetrical painting either. We have pretty good retopo tools, but it’s still time consuming and dampens creativity. The most ideal solution would be to improve the Vertex Painting (and I can see that as one of the planned dev targets) because we could instantly go from dyntopo to color, without any steps in-between.
But the problem won’t just be fixed by adding PBVH optimization (like the docs say). Currently vertex paint effects face-verts which, in terms of model painting, is pointless and slow (both data manipulation and rendering). We need one-vert-one-color, or at least that option. Moreover, we need a “uniform vertex density” modifier, which sub-divides a triangulated mesh based on polygon-surface area (to generally uniform the vertex density for poly-painting) so we don’t need to manually brush in vertex detail prior to painting.
Just throwing that out there in hopes the developers who work on this will see it. I’ve played around with adding these features myself, but I’m not a C expert, and I don’t have time to develop all of what is required at the moment (it needs a new render path for one-vert-one-color rendering, etc). Either way, thanks for everyone’s work. I’m glad to see this is a planned feature in one form or another as it’s the biggest area of annoyance for my personal Blender use today. :-)
I wouldn’ say Polypainting is the future when it comes to texturing a model with Blender, Painting tools offered by the texture Paint Mode are quite more powerful and flexible ! PolyPainting with a mesh sculpted in Dyntopo doesn’t really make sense. :)
I think Blender is heading in the right direction with 3D painting tools and I’m glad to see some development for these tools !
Well texture painting tools are very important too. You need good texture painting tools for many things, not the least of which is low-poly modeling. The main limiting factor for 3D sculptors however, is that you can’t efficiently unwrap/texture-paint a sculpture with thousands or millions of vertices. You need to retopo first, which dampens your “creative flow” a bit (and means you need to finalize the geometry before playing with color).
I completely disagree that it doesn’t make sense to poly-painting with dyntopo. It’s ideal when you want to quickly experiment with color along with your shape and still have the freedom to change everything. Perhaps more importantly, poly-painting makes a ton of sense with high-detail sculpting. Often you get a lot of millage from simple alpha brushes (like vein textures) which can add slight surface bumps as along with color. I’ve used this technique in Zbrush often and it works very well. Unfortunately Zbrush is the only software which allows you to do it ATM.
So i’m not against texture painting tools or anything, and would like to see those tools developed more too. We still can’t do symmetrical painting there either, which is a problem for any asymmetrical model. But for sculptors, texture painting is not as ideal as poly-painting, and currently the poly-painting in Blender isn’t really a realistic option at all.
Is there a build with the Split Normals feature in it yet?
Yeah this is something that is preventing a lot of indie devs getting to the next level in game art, Splitting and transfering mesh normals are very important, but as far as I know it’s not in blender yet. I’m guessing because its still primarily focused on rendering rather than realtime.
Dynamic topology has already massively improved- a few releases back, enabling dynTopo used to destroy existing edge flows and create a new triangulated mesh, but now, it triangulates the existing mesh! I am absolutely loving that feature.
To paint on dynTopo mesh, could the mesh be subdivided only for vertex colors somehow?
I am not sure but game engines might be using some features like this to paint terrains.
If that could be done, we can have a medium-resolution mesh with good resolution paint detail which can be transferred to a uv map later.
This can also be a problem though, for instance say the original geometry was created quite quickly and contains a lot of poles or areas with very little geometry like if when creating the nose all you did was pull out a vertices for speed, when it comes to activating the dynotopo feature all it does is triangulates he original mesh instead of creating a new mesh in the shape of the original one especially when you cant rely on the remesh modifier to do a good job on organic models.
Due to the way that the Dynotopo feature works, if your model contains poles, when it triangulates the mesh, it still contains poles which creates a problem when it comes to sculpting the model, because no matter how much you smooth the model, it will always contain the pole and you wont be able to get rid of it, but the way it worked before when it created a new triangulated mesh, it didn’t have that problem so sculpting was so much more fluid and wasn’t limited by the original geometry.
I just want to take a quick second here to thank the devs for some significant improvements in the code that make a lot of difference in version 2.70 of Blender’s modeling toolset. One subtle improvement I have noticed is an improved Bevel tessellation scheme. Before 2.70 it was necessary to re-align edges running crosswise through the edges that have been beveled. With the updated coding, everything stays aligned by default, and it’s a powerful improvement that saves a boatload of vertex fixing after the fact.