I wanted to try SubD and see what I could learn from it. Never touched Max nor done more than a Coke bottle in Maya. FWIW, I really rate Blender since the 2.8 update and this has made me want to learn some polymodelling, so I am still wondering if it’s the better to place to cut my teeth.
I thought doing the Panton Chair would be something that’s both doable, but also challenging enough. I learnt quite a bit from it, coming from someone who doesn’t really know what to look for. Process was mainly by guesswork, but:
Made a Cylinder. Saw the pole. Realised it was going nowhere.
Made a Cube, realised it was a better start.
I then ended up with this topolgy by some accident, where the red ngon on the corner gave me kind of the right flow. I persevered with the green corner for the back radius for a while, although knew I was kidding myself.
At that point, my idea was to basically kick out the topology in the bends and kind of join up straight, and to then sculpt by hand. But I thought I would see how QuadRemesh fared.
QuadRemesh did basically all the hard work and thinking for me which was handy. No idea if the flow etc. is considered ‘good’ but I had the control I wanted.
Most handy tools:
Auto reset Gumball option (didn’t realise it was an option)
SoftTransform - A bit weird since you dnon’t really see the influence area
SetPt worked nicely for straightening up
Gumball Align to View for tidying was essential
Reflect, obviously
Slide for edges and vertices, and just with Gumball and near snap
3DFace for handbuilding one of the corners
If anybody wants my musings / wishes from doing it, I can list them too
Thanks Kyle! Yeah your way of doing it has hit it on the head. That sort of starting point makes it much easier. I tried to model ALOT of detail very quickly and it burned me a bit (eg below, during). What I’m excited by in SubD is using all the known tools to, as you did, build a cage. Not just primitives.
Below was my attempt at refined topology (right). But honestly I feel like there were plenty of ways. I’d love to know what sorts are good for what purpose. For me, I would probably be thinking about where I would be likely to make more edits., and build topology in a way that means that bit could be easily edited. I guess it’s not rocket science though. And, for example, to centreline or not to centreline.
Okay, so, I think I’ve got all my conclusions from it for now.
An easy win here, I think. Single SubD Face in the SubD Toolbar: Right click tooltip could be for Mode=MultipleFaces. The current single face would then need to be updated, so that the Mode=SingleFace is actaully sticky. At the moment, if you’ve most recently done MultipleFaces then that sticks, even though you’ve clicked on the ‘Single SubD Face’ button. I reckon that’d be better for the default toolbar.
When dragging out the SubD primitives, would it be a lot to ask to see the cage object? So for example, if dragging out a SubD cylinder, you just see a generic cylinder as though it was a polysurface. Can’t help feeling something is not the right way round there. Does it seem logical you would only get one hit at visualising how much base geometry you will have to work with, before even defining the radius/height. Probably just be me.
Could PlaneThroughPt work for certain bits of SubD? Not sure how it would work out in practice as far as whether it uses the sharp/smooth point/edge/face.
Would it be crazy to ask for CurvatureGraph on Edge Loops? I couldn’t always guage if there were inflexions from the cage alone, or pinpoint from emap. Currently worked around with DupEdge+History. Then I guess you would lock those child objects so you dont select over them.
Would it also be crazy to ask for a Display Mode override to whether sharp/smooth editing of SubD is shown? So you then have editing in a top/side view but still see the smooth mode? I don’t know. I guess some of it not just being this OR that. Maybe it is crazy and would break stuff.
How can we go about lining up points of a cage, relative to a CPlane? This curve is across the centre, so I would want them to kind of shoot up in Z only. Could give a bit more control further down the line modelling for aligning stuff in top view even. I can’t find an option yet. BTW Align: ToPlane is really nice for straightening up Points.
SubD COntrol Point or SubD Vertex ? Sometimes I am seeing a choice. I would say that, out of the two, the Control point is more handy in that it kind of indicates which face/s it belongs to. Say points are close to stacking, and you want to know which one you are about to move.
I am seeing some strange display behaviour which looks like behind the scenes vertex order stuff rather than my aged GPU.
Typically, it’s more fluid when you don’t lose your selection after moving something a bit. For things that straddle the reflection plane, I seem to lose the selection after transforming. Anything not straddling the plane is fine. There is some similar stuff to that which bugs me in relation to Undo - sometimes you lose your selection, sometimes you don’t. I think it’s do with subobject selecting.
Other Qs:
Does Soft Transform need an indicator? Feel like that could be a can of worms though
Check the below out >> Blender proportional edit WIP https://developer.blender.org/D7583 - maybe even just considering the idea of the softly transformed stuff redrawing live. The same way you see the results of what you’re doing in the SMooth dialog box. Although of course I get it will be a resource drain.
I think SoftTransform works ON TOP of Smooth. So for Smooth it will add an additional level of fall off on top. Maybe it’s desirable, maybe not. I nearly didn’t notice I’d got some extra smoothing. In my eyes the point of the Smooth is only focusing on that stuff you’ve explicitly picked.
Syntax and nomenclature stuff (uh oh) I guess the below kind of works, I get there’s a bit of limbo at the moment where I feel lucky trying a mesh command and sometimes it works on a SubD mesh face. Seems to me that ExtractFace would be used in place of ExtractMeshFace, and take Mesh / SubD. I know I know, it’s still early. ExtractControlPolygon for instance doesn’t give the expected result, in my eyes.
DeleteFaces - Mesh Faces and SubD Faces ExtractMeshFace - Mesh Faces Only Extract Srf - Surfaces and SubD Faces (?)
Some other takeaways:
Traditional Smooth command is really handy
3DFace > AddTo option is great
Didn’t mention, but T-Splines was before my time. So really, in the grand scheme, my judgements are probably quite ill informed.
You’re welcome. My views are far from gospel, think of me as trying to look out for the beginners I feel like anything that needs hunting down should be avoidable.
Are there reports made for 8. Vertex Order Display glitch (if you can repeat the steps) and aligning points to a curve in a Cplane?
For symmetrical stuff I am really finding Smooth useful. Never did very much for me with surfaces, but it’s great for averaging out even just a vert at a time or a row, average out a vert only in X or Y, etc. Which reminds me I should maybe test smoothing in a CPlane XY - CPlane co-ordinates yep, that’s really cool
Huh… so on the spheres and torus, this works. So yeah, that across the board. Maybe it’ll look cluttered and more will oppose than favour it though. I mean, the circle dragging widget on the SubD Cylinder is basically pointless anyway, as the picked point is for a cage point (So creating a 20mm diameter cylinder creates a 20mm diameter cage object, but a 15mm diametersmoothed object).
Thanks for noticing that - that’s good. My previous bind of trying to line stuff up on a non-world plane can be fixed up with that. Pull was handy there as well.
Can you see 8. as a bug on your end? I’m seeing this kind of temporary thing a lot. I don’t have another GPU to compare to. sysinfo_jhut_wip_24.05.2020.txt (2.4 KB)
Yeah it does it very regularly - as I say the step is when having a face selected, and doing a toggle.
I’m looking into this now but I can’t find any available updates. It’s an Nvidia 660M. Nothing is screaming at me - Rhino WIP still gives me the notification tip but doesn’t have anything better or more update to suggest. The best I can do is get the 660 (not for notebooks). Anything that technical folks can recommend?
I really don’t see anything later than 11 / 04 / 2019 - 425.31
Yeah I’ll do that. A file example and video is above if you want to try and reproduce.
Edit. Nothing more up to date available.
I mean, I’ve kinda just become used to the graphics display being crap. Although I’m sure there’s plenty of hard work behind the scenes on rendering, I never see any of the fruits of these labours… Raytraced mode can’t even manage to generate a grouns shadow, and hitting Render gives me a black screen. Guess it’s a good thing I don’t want to use Rhino for that right.
@theoutside Here’s a quick one. Should raytraced viewport do some kind of ressing up? (ala KeyShot) again I’ve kind of just grown accustomed to the Raytraced viewport being non-functional and avoided, regardless of it’s reported use of Cycles.
It should be on the horse and cart anytime now… that should make things clearer. It’s coming up to 8 years old, and it was a mid-level card at the time, so should it really completely not be able to run Raytaced properly at all?
Let me know if there are any tests I can help with/show if things should be working better on this kind of card… I don’t really have any testing utilities for this - certainly I have no hardware fault notifications.
8 years is forever in computer time…do yourself a favor and give it a proper viking burial. If you can’t up date the video card driver any further, it’s a boat anchor now. I’d get a new machine, and demote this one to a backup.
Go into settings and set Cycles to use CPU.
It will be faster than the 660 anyways.
CPU works just fine!
(but new gpus with thousands of cores are faster than a cpu of course.)
Works a treat - maybe that Render panel will make it’s long awaited return to the table as a docked toolbar. Really overexposed and not using any features BUT at least I made use of the display mode right?
No, I have the 750m in a mac (running windows) and there too the cpu is faster.
The 960 and above would have been an upgrade. But no shame in using the cpu, just get yourself a larger cup of coffe
My new build will use an RTX 2060. But I heard something about the Ampere series really being a massive leap in performance, and might not be a huge jump in price for an equivalent ‘level’ ( a 3060, so to speak ).
According to Guru3D the ampere 3060 will have more cuda cores than a 2070 (and less than a 2080), so it will be a big upgrade, about 50% more theoretical power over the 2060 with it’s 1920 cores.
But that’s how it’s always has been, so can you wait? If you need it now then get the 2060 or a used 1070, and swap it out for the 3060 when you need more power. (When I say “need” I mean do you have paid jobs that require that you render in 1 hour instead of 1.5, or 10 minutes instead of 15? If so you are making money and should not wait. If it is 1 minute instead of 1.5 then you shouldn’t care too much. )
I feel like anything that needs hunting down should be avoidable.
