Pablo Dobarro’s answer is (quick and roughly translated, i speak spanish so…):
“With dyntopo there’s so much confusion that i don’t even know where to start… that ‘is an optimization that only adds details where is required’ that’s already a lie…and ‘the remesher is 20000 times faster in any situation even if you put 20 million of vertices [also a lie]’… the thing with dyntopo is that is built on top of b-mesh and that’s edit mode behind the scenes, so what i’m saying about deprecating it, is to not invest anymore time of development on it, if we already know that it will never reach to the performance level [of sculpt mode], due to how it’s based [on edit mode]. I would leave it anyways, but the remesh would be multires. That’s what i say ‘better stop bere’ and we look what to do before to continue investing my time and of other developers in something we already know will never reach the performance [of sculpt mode]”.
Another guy ask if “can be faked”. Pablo answers that dyntopo already pretty much does that, the problem of that is because is coded on top of b-mesh, that is way too heavy for it (bmesh supports n-gons, vertex, edges… and for dyntopo only triangles are needed at all).
Also Pablo states that “people will need to realize that a remesher with 40 million of polygons works better than dyntopo with 100 thousand” [TRANSLATOR NOTE: Pablo Vasquez adds “and GPU” i didn’t notice since audio is not exactly the best…]
The other guy remarks what “deprecation” is: The feature is not going anywhere, but it will not be maintained anymore, for Pablo Dobarro “is not worth for a developer to invest 3 weeks fixing code that can’t be improved at all”.
Probably some other people can translate this better, but this is it.
A 500k mesh with detail added only where needed is faster than a 20million polygon mesh with the same density everywhere, that’s why I use dyntopo for illustration and advertising stills…
Last Friday there was a UI meeting. One of the topics was how what to do with this Sculpt Face Groups. Here is the relevant information from that discussion for this feature.
There was a discussion about this patch using a generic data structure (Face Mask) vs a specific data structure (Current implementation). We decided that for this feature we keep it to a sculpt specific feature and not use current the generic Face Masks. The reason behind it is that the use-case of the Face Masks data structure isn’t clear.
In Blender we use the term Mask for mutual exclusive data. We use the term Groups for not mutual exclusive data (eg Vertex Groups). For this project it means that the term Group should be renamed to Mask eg Sculpt Face Mask. This also means renaming the variable and function names, display etc.
So they came to the conclusion that this feature does not need to be supported outside of Sculpt Mode and that it will not be merged into Face Maps. This is pretty good news in my eyes, since the way faces are handled in Sculpt Mode is not actually selecting pieces of a mesh like in Edit Mode. Now there’s a good chance we’ll see it in 2.83 at last.
Is the code design really in that bad of shape (to the point where it would need to be gutted and rewritten anyway)? Dyntopo going away wouldn’t be near as big of a deal if the voxel remesher’s adaptive detail did a much better job at avoiding skinny and/or invalid topology.
It’s not that crazy that Pablo has that kind of power because it’s his by default of being the only developer working on sculpting features.
Pablo currently has a lot to deal with as the only developer in the sculpting branch. It’s understandable if he doesn’t want to deal with Dyntopo. That being said, if the community loves Dyntopo that much, we should gather further funding to at least get one more developer into the branch so they can work on improving Dyntopo and whatever other features are needed that Pablo can’t work on.
If you want something, fund it. But that’s just my two cents.
If face masks are going in, then this bit is interesting.
It can also be preserved when remeshing the sculpt with the voxel remesher.
Does that mean it can be used to ‘lock’ a section of the sculpt in place so further voxel remeshing won’t impact it? That would help with keeping carefully sculpted details intact.
if that’s what he said then he has no idea what dyntopo is actually made for… maybe he isn’t skilled enough to tackle some real problems? hm
If you think you can do a better job at sculpting development, and if you think you know better as to how the code works, then you know where to grab the source and start writing patches.
Is not exactly a lie… dyntopo is more or less, another tool of edit mode. Smart enough to do sculpting with detail only where is required, but is true that operates on top of edit mode, so in escence (from a coder standpoint) is a hack. And so it inherits all virtues and weakness of edit mode.
I do personally understand why Pablo doesn’t like it, since Pablo is both artist and coder, is quite easy to see why he doesn’t like it. Basically is the same reason a house builder doesn’t like nylon head hammers, where a carpenter will like it and need it.
I think it means your polygroup will still exist even after remeshing, like dynamesh… I don’t think it has anything to do with the details of the mesh…
The lie is not on the technical/dev side, he knows better than me how messy the code is or isn’t, or what is a hack. But saying that a 40million polygon mesh is faster than a 100k mesh done with dyntopo just to prove his point is just not true.
Ok, I will not call him a liar (at risk of being accused of attacking him personally) but he’s exagerating a LOT just to convince people to get rid of dyntopo. And based on his comments I’m pretty sure he just use dyntopo with the same density everywhere, which is not the main strenght of the feature.
So that would explain why Dyntopo saw a performance regression with editmode in 2.8.
Either way, editmode tools inside of sculpt-mode sounds like ugly design and would explain why it couldn’t be made robust enough to preserve data like vertex colors and vertex weights. I know it has strengths vs. voxel remeshing, but on the flip-side the voxel remeshing instantly produces a mesh ready for multires and to a lesser extent animation (with corrective smoothing), and a working multires modifier is being worked on as we speak.
I also understand Pablo’s arguments regarding Dyntopo’s technical inferiority. It’s just that I love the realtime triangle generation workflow as a viable alternative to voxel remeshing, with a number of advantages. Blender sculptors should keep the choice between a Dyntopo-like workflow and voxel remeshing, like ZBrush offers both Dynamesh and Sculptris Pro.
But if Pablo’s new voxel sculpting method will be as non-destructive as Dyntopo, offer creative freedom like Dyntopo, doesn’t melt together parts like Dyntopo, allow local polygon refinement, etcetera, you won’t hear me complain if the new method is way faster than Dyntopo.
If dyntopo is essentially edit mode, does that mean performance improvements for edit mode would improve performance for dyntopo? Or is it not that literal?
He only see sculpting from a character sculptors point of view. Try to add some sculpt details on a CAD file without loosing any of the details of the rest of model with voxel remesh. And adding 20 million polys is not an option when you already working with near 100 million polys. Will have to go back to sculptris or sculptgl if dyntopo disappear from blender.
I think that his point is that handling 100k mesh of b-mesh data (for artists, “working with…”) is way heavier than handling 40 million in sculpt mode. Is not obvious, but after hearing this again (and adding some notes to my post) maybe that’s the problem, and that’s why on his eyes sculpt mode obliterates dyntopo.
Probably will be worth to wait for his implementation of voxel remeshing before judging anything.
For CAD models, improvements by Germano and others (weld modifier, automated edge-splits, ect…) and soon the new booleans from Howard will make it far easier to add small details to hard-surface meshes without the need of sculpt mode at all.
Because of the weld modifier, I can do things like etch panels or text into surfaces. No need for brushes.
