thanks a lot, iceking, for sharing your workflow. i see the vertexpaint as a big plus of the floodfill now. yes, it will get much more uniform in quality. good point!
getting a stable mesh that is despite the detail still rather lowpoly was never a problem to me without these features, but in next sculpt i try your workflow, it is always interesting to experience how other do. thanks!
The constant detail is great to sculpt and I love it.
I create my shape with constant details, when Iâm ok, I fill flood and after I use contant details to sculpt my parts, I change de value to add more details.
Maybe if fill flood had a smoothing on/off option it could have more use ?
Flood fill adds new possible workarounds.
-
As iceking mentioned, Flood fill and work using subdivide edges. This is possible now.
when using subdivide collapse only, you may not find it that useful. -
Floodfill is a triangulated remesher. IMO, the only remsher we can trust in blender. It is the equivalent of the zbrush dynamesh by any aspect. Dynamesh creates quads you may say. The topology of these quads is of no importance, the behavior is exactly the same as in triangles-dyntopo floodfill.
-
@iceking, (the quasi retopo way, close to your above tips)
a. duplicate the dyntopo sculpt.
b. decimate it, by using floodfill at ~10 k - 20k
c. edit mode. Make seams by selecting edge, ctrl+click (the shortest path). You can have a decent unwrapping in a few minutes this way.
Add lights to the scene so have a nice render. three lights + environment. a base plane to have some reflected light-color, all this stuff, right?
Here we go! The new way.
use the cycles bake branch and directly bake (to active) a combined texture. Enter BI/GLSL, assign a material,add the texture as color=1 , emitter=1. A little value on normal (bumps)
Add a light in the scene to control glossy and youâre ready for a real time animation with cycles quality.
@Doris, and all zbrush users, z remesher is the best fast way.
Something funny. After cycles baking, import the obj and the texture to zbrush. Use the shadeless matcap. LOL Cycles render in zbrush⌠really funny. -
Performance
You may like to start sculpting dyntopo on a cube or a base. After a few hook-move brushing, Flood fill the mesh at ~100k - 200k
Start relative dyntopo as you wish. IMO, performance is much better.
I tried flood fill at ~ 500K. Performance was excellent when adding more details on this.
A more natural feel, canât describe it better.
ah, thank you, ,now i start to see possibilities⌠yes, i had subdivide collase on, and could not see good benefit⌠and using it as a remesher did not occur to me. will try to see how it works⌠performance, i had no problems but i check this too. thanks a lotâŚ
@doris,
yeah, performance wasnât an issue, for me.
It feels more right. How can I describe it? Just try. On your computer, I would say⌠go for a 500k evenly distributed mesh (after blocking the basics, the forms) and start working on details. Crease and pinch brushes work better under subdivide edges mode.
performance and workflow efficiency matters to me cos i got only 2Gb ram, hehe. gonna get the pen first before i up my specs.
@Sanctuary: autosmooth can be advantageous, zbrushers from the ancient times without these fancy remeshers and using only good old subd to build sculpt, - will often have in their tuts = dont smooth it all out, once you subdivide a level all that will be âsmoothedâ anyway, (this is primarily for organic models); what they meant by that is the building of anatomy or muscle forms/shapes, these will meld naturally by using the subd smoothing reaction when increasing its level.
i have tried this before in dyntopo, no constant detail in those days yet, i replaced the âremeshâ button in sculpt UI with âsmoothâ button which instead of auto-applying a remesh modifier it uses the smooth modifier instead, (with undo options etc), to replicate the process i mentioned in the previous paragraph, needless to say = it freaking works.
so yeah, autosmooth i think would be feasible. just not on/off, please, but rather a slider value of 0-1.
@Iceking, sanctuary.
An optional auto smoothing, yes.
Tip 1:
Auto smoothing, or the old subd (catmul clark) (multires, zbrush etc). Better avoid if you sculpt rocks.
Back in the days of multires, zbrush sculpting, the simple option of subd was the way to go. To keep the low poly triangulated (or quad) feel and smooth manually what is needed only.
Tip 2:
Having finish the basics, nice clear forms, shapes, we may start adding pores, wrinkles, all this kind of stuff. These tend to smooth these clean forms. So, avoid auto smoothing as possible. It will happen anyway.
@Psy fi
Changes on the Detail Size (in px)
Relative detail
Now I have to use 2 px 3 px, 1 px. Instead of 8 -15 px.
This is quite wrong. A 30-40 px creates 2-3 triangles only.
(on a few mins ago build)
Revision e659cfd by Antony Riakiotakis
1 day and 1 hour ago
Change dyntopo detail size calculation slightly so the result is close
to the radius of the widget when setting the detail.
@michalis: i think this is it youâre talking about.
i kinda like it, really low doesnt seem like an overkill now. feels safer. even relative detail 1 is smiling now. saying âim your friend, use meâ.
edit:
also seems to be faster, (users with higher spec will not notice this), my iceking sculpt is 500k tri and when i sculpt that in dyntopo mode on (relative) before thereâs a bit of a lag, now it feels a bit more breezy, hhmmmâŚ
OK,
Now I have a slider, but I can use 1,2 3,4 5, and thatâs all.
@michalis: what do you mean?
i recreated the holes by doris, those arenât holes, those are overlapping faces and sometimes normals not updating fast enough (go out of dynto and go back in again, most of the âholesâ are gone), smooth them out and the âholesâ are gone.
Sorry the Relief Maker fine-tune is not finished and I have a date Now 
I am making good progress with the screencast recorder ( Alt-F3)
âŚcause I have to show you guys how this is supposed to work.
Meanwhile a fruity variation of the clown:
I agree with Michalis after just testing a new version i really donât like that new detail scaling for the Relative setting, where i could sculpt nice at 20, now i need to go down to 5 to get the same kind of details.
Everything above is just making too big triangle (unless youâre really, really close to the surface) to be really helpful in building details , so it looks like 5 is the new 20 or something around that.
looks like 30(starting value) is sth like 10 now and if we will get the shift D to jump by 1 not 10 i can live with it (still i liked it the old way). And I agree with iceking - performance is better now but that may be becouse of other changes we dont know yet.
And since i didnât post anything here for ages⌠here is my mutant that i was working few days back
Right Sanctuary
It also brakes compatibility with previous builds.
I opened my test on a few days back build and density is still at 4 px.
Not really important though. Just keep in mind when testing some other things, like performance. LOL
OSX: Remove the apple/omp workaround for now, to better judge compiler differenc
I read this on new commitments.
Well, now, blender suffers from the same lagging as the official 2.7 OSX build.
Now the issue came into master.
OK, it is temporary. I hope so.
Now, dyntopo performs as the first experimental dyntopo builds 2 years ago. On 200 k lagging.
However, isnât the rotopo, turn it off, same lagging. Turn off threaded sculpt (this is all about), it doesnât fix the problem.
So, a warning to Doris and other OSX sculptors. Avoid new builbots.
It is temporary. I wonderâŚ
thanks man XD
The rationale here is to have this be closer to the radius of the circle when using shift-D. I agree this leaves little room for detail
refinement though so changing this to float rather than integer with focus on low values should solve this.
Today on the meeting notes, we had a discussion.
Points are:
- Some mac systems do not suffer from this âŚ
- so we need help from those who do suffer from the issue so please do report the issue (take care to report your processor, graphics card, memory and version of blender used).
- Solution will most likely take the form of us doing some custom builds and asking people to test them, most likely from the bug report page.
- Also in addition to that we are preparing a commit that will allow custom setting of openmp threads.
We do care about it, enough to do a call through the developer meeting:
http://lists.blender.org/pipermail/bf-committers/2014-March/043215.html