Dyntopo, are skinny faces causing problems?

[anon92502066]

@zeauro

I think that there is no argument valid that can justify to use a Collapse method for a brush which have for main feature to pinch faces.
“I want smaller close faces but don’t create smaller faces.”

Like I said , you can debate all day and some one at some point will have a use for it. As long as its consistent with the rest of the system nobody will remove / limit collapse for crease.

Yes but you also have these abilities with other brushes. If you don’t use Crease to use its pinch slider, you can use default Draw brush like that.

That doesn’t mean draw can be efficiently used for creasing, most brushes have auto-smooth attribute which smooths the stroke that doesn’t mean we don’t need a Smooth brush.

It is not because you can do it that you are forced to do that.
But the setting will not disappear of UI because it becomes a per brush setting. Are you saying that you never modify brush settings according to workflow ?
If you don’t want to create a custom brush you can still adapt the setting at the moment when you need it.
Sculpt Plane is a brush setting.
I would probably adjust a plane offset more often than change subD refine method if it was a per brush setting.

I keep a set of base brushes which I use both for organic and hard surface sculpting. I seldom require the need to change an attribute other than size ,strength. However I dynamically change the refine type based on the topology.

The UI might stay in place and it might be as simple as keeping couple of floating point variables and couple of enums in a brush object. But that doesn’t mean its efficient either. Because dynatopo methods like relative detail force the user to be aware of the topology. Otherwise your vertex count can bloat. Which in turn chokes the 3dviewport and brush responsiveness.

Per brush subd settings work best if brushes use subdivide as the only refine type and collapse operation is done by simplify brush. Because then you only need to keep in mind 3 variables (Size, Strength, Subdivision detail).

But Blender has 2 different methods (Relative, constant) and 3 types of refining (Subd -SubDC - Collapse). Keeping tabs on each subd attribute on top of each brush’s own attributes for each brush is not efficient.

I am just talking about making one setting per brush in order to avoïd to check SubD Refine Method each time you access default crease.

I am not requesting for hundreds of brushes in startup.blend.
Brushes management is the responsibility of users.
Mature users are not stupid at the point to put hundreds of brushes that they don’t use in their startup.blend.

You might not want that , that doesn’t mean other people won’t do it. The current brush palette is hard to manage as is.

If you design a system which supports better brush management or a simpler subD method I won’t be against it.Current state just isn’t efficient in my opinion.

Anyway if people are in favor of this , then at least it should be optional.

[zeauro]

@anon92502066

That doesn’t mean draw can be efficiently used for creasing, most brushes have auto-smooth attribute which smooths the stroke that doesn’t mean we don’t need a Smooth brush.

My point is that if you don’t use pinch value of crease brush, you don’t use it for creasing. So you can use draw brush.
Don’t say that you never change your brush.

@anon92502066

I keep a set of base brushes which I use both for organic and hard surface sculpting. I seldom require the need to change an attribute other than size ,strength. However I dynamically change the refine type based on the topology.

So , you agree. If you don’t have to change it, you gain time and hhings are simpler.

@anon92502066

The UI might stay in place and it might be as simple as keeping couple of floating point variables and couple of enums in a brush object. But that doesn’t mean its efficient either. Because dynatopo methods like relative detail force the user to be aware of the topology. Otherwise your vertex count can bloat. Which in turn chokes the 3dviewport and brush responsiveness.

You can easily have an idea of topology by disabling smooth shading.
Workflow shaders project should help to have a non-distracting wire overlay.

@anon92502066

Per brush subd settings work best if brushes use subdivide as the only refine type and collapse operation is done by simplify brush. Because then you only need to keep in mind 3 variables (Size, Strength, Subdivision detail).

No. It is incredibly slower than if constantly give a second pass with simplify brush.
For most brushes, Subdivide and Collapse is a benediction. But not for Crease.

@anon92502066

But Blender has 2 different methods (Relative, constant) and 3 types of refining (Subd -SubDC - Collapse). Keeping tabs on each subd attribute on top of each brush’s own attributes for each brush is not efficient.

It is not because current UI is grouping both options that ungrouping means a dramatic regression.
It is an improvement.
Let the switch Relative/Constant that is frequently used as an highlighted setting and put at a place where it can not be confused with the one that is almost not used.

@anon92502066

You might not want that , that doesn’t mean other people won’t do it. The current brush palette is hard to manage as is.

Current brush palette is the most basic brush palette that can be done. One brush per Sculpt Tool.
Dingto deleted Polish brush because it did not fit in.

@anon92502066

If you design a system which supports better brush management or a simpler subD method I won’t be against it.Current state just isn’t efficient in my opinion.

Blender can append brushes from different blends.
With project about new asset browser, it will be easy to keep exotic brush libraries as separate .blends and append them only when they are needed.

@anon92502066

Anyway if people are in favor of this , then at least it should be optional.

My position was never to impose detail refine method. Just to make the option adjustable per brush.

[Hadriscus]

I used dyntopo recently after a long time and was impressed by how good it is. However when I reach around 500k tris the undo takes fifteen seconds… I know development is not focused on sculpting these days but since ideasman committed a few things I wanted to ask whether this situation was improvable ?

Thanks,

Hadrien

[michalis]

@Hadriscus
The undo issue. Psyfi could have a look if he had some free time.
He had some ideas about it.
There is a more nasty issue though. It is the autosave in prefs.
Please de activate it and see.
After all, it is a bug. Autosave doesn’t save anything when in dyntopo mode, still it freezes some seconds every 5 mins by default settings.
The more dense the dyntopo the worse it becomes.
Like the undo issue.
New (reworked) behavior of brushes, the new clay brush especially (unlock/on the left, the strength), is amazing. Thanks to Campbell (ideasman)

[Hadriscus]

Thanks michalis but it is not autosave, always been off on my side.

[Ace_Dragon]

By now, the only big bottleneck left with Dyntopo sculpting is the undo.

I wonder if it can be made faster though by having the undo operation only apply to the PBVH nodes that have been touched during the execution of the stroke instead of taking the whole mesh into account. It would be good to have the undo system become a smart system in this case (if you only touch 5 percent of the polygons, then the remaining 95 percent would ideally just get ignored).

Also, something related, I thought I once saw the start of a branch that would use a similar PBVH tree system to accelerate vertex painting, it would be nice to see that revived and committed to master.

[michalis]

@Ace
About PBVH (it sounds like a virus LOL)
In dyntopo, CTRL+ALT+D, enter value 14. The debugger.
I’m a little confused.
Test 1:
enable dyntopo
use the grab brush. As expected it doesn’t alter topology as it’s not a dyntopo brush (it shouldn’t be, hook is for the job). See these cubes they just change dimensions. Fine. Now click on optimize button in dyntopo panel. hmmm, interesting.
Test 2
Using a dyntopo brush (claystrips etc). Structure of cubes changes as I add topology, fine so far. Click “optimize”, hmmm interesting, the whole structure changes a lot.
Add some more geometry by brushing. Do the same, ask for undo. Though undo happens alright, the structure of the cubes isn’t the same. A new structure.

Difficult for conclusions by me. However, is this what we’re talking about?

[michalis]

More conclusions on debugger.
Optimization button takes some time to recalculate the whole mesh (rebuild PBVH). Especially on heavy dyntopo meshes >500k, at >1M what do you expect?
Fine so far.
Now when asking for undo, optimization also happens. This is why it takes so much time. I’m not sure if this right.