The big Blender Sculpt Mode thread

I keep seeing this, the comment about realistic versus stylized art. I have to say that this is so off putting to artists- your computer programs do nothing but manage pixels, and realism is subjective honestly.

Pick up a paint brush, and even with all your best effort to render - at some proximity, the illusion is broken and is revealed to be an imitation. Realism is only what is real, and might as well be a golden ring Tantalus holds because we cannot ever achieve it - we can only fake it.

Those that pursue stylized art are free to go wherever they dream, and it takes all the same tools to do so but it takes a lot more imagination to get there.

This idea that a tool in Blender is somehow limited to stylized work - pfft, that is hogwash. I was using all these tools for ‘realistic’ imitations all the while, using remeshed sculpts with texture painted maps fed into Principled Shader trees to imitate real food items for photo shots for packaging. Like I said, get close enough and you will discover the lie, but far enough back and boom, the customer is happy.

Add tools to Blender, and artists will find ways to use them and then share what they did - that’s how it all works. I’m scared about dropping the texture paint workflow in favor of painting while sculpting, but I’m not the developer that can imagine the workflow so I’m just going to ride it out.

21 Likes

All it needs would be to look into the whole industry standard character sculpting/creation pipeline. Blender doesn’t need to invent a “special” workflow or any kind of “special” tools like Pablo had in mind.

Why trick Blender into a great program for stylized work if it could be like ZBrush and great for (basically) every sculpting and character creation needs.

8 Likes

I’ve always read the “realism” argument as a dog whistle for “has to be as fast and good as ZBrush.”

3 Likes

Maybe the I’m missing something but isn’t the realism stuff a reference to being able to sculpt items such as pores and scales in Blender?

Though i thought stylised stuff also has pores and scales in some models.

4 Likes

It shouldn’t be about realistic or stylized (or whatever) art is should be about the tools, performance and functions professionals need.

For example the current state of shape keys isn’t usable with lots of polygons. Useless for people who push lots of polys and need the abiltity to have a “sculpt layer” system.

edit: Another example would be Quadriflow which is pretty much unusable for any kind of work.

2 Likes

EXACTLY. And it’s kinda insane that people seem to have this epiphany every hundred or so posts in this thread as if it’s the first time it’s been said. Every time.

This ain’t really a matter of sculpting for realism vs stylized art. It’s a matter of sculpting super detailed highly representative art pitted against sculpting for abstracted low detailed art. Everything can be “stylized.”

Pores can be stylized. Maybe you need denser polygons to make brush strokey sculptey patterns on a stylized face extra sharp. Look at Dishonored! Those faced are EXTREMELY stylized, and I know my workstation would actually chug if I had to sculpt comparable models in blender. I’ve got 32gb of 1800MHZ ram, a Ryzen 7 3800x, and a 3090. Blender STILL chokes on a mere bust with 5 levels of subd. Imagine a whole body. I think Kent Trammell’s upcoming tutorial may serve to highlight Blender’s weakness in this area tbh. No doubt there’s gonna be a few workarounds to the performance limitations, not to mention he’s also just sculpting a bust, not a full character.

I think not making performance the first and foremost important priority was a massive mistake. I’m happy for the QOL tools we got with the face sets and cool pose tools. I use em regularly! It’s just that QOL tools will only improve the QOL so much if the base Q of the base L is low to begin with.

I mean, shit. We’re all artists, right? We gotta acknowledge that stylization and realism are marketing terms for consumers. We have to know better. Not a single damn person in this thread should be talking about realism vs stylized as a dichotomy. What we create ranges from specific to broad. Representational to abstract. All things in between. The point is asking for tools specifically for modeling “realism” is stupid as hell because that’s not something that can even be objectively defined. And… It’s kinda obvious that everyone should agree that being able to push around hundreds of millions of polygons would be a net positive for everyone regardless of their purposes, right? Some people want to put pores on their huge-eyed pinup girls of questionable looking ages! Some people wanna sculpt a super abstract character or thing that’s made for 3d printing and therefore needs to have the threads and texture of cloth sculpted right in! Right now, performance and the fussyness of managing brush textures makes that all damn near impossible. That limitation has nothing to do with striving for realism.

And Obsurveyor, I don’t think we NEED a dogwhistle for “I want blender sculpting to be as fast as zbrush.” I think referring to that as “realism” is just downright ignorance and naivite. If folks want Blender to perform like zbrush, they should straight up ask for that.

But I don’t think we really need blender to perform like Zbrush. Zbrush’s performance allows for absurdly dense assets that are more than even some feature films need. I think a more sensible expectation of Blender’s sculpting mode would be, like, Mudbox. Mudbox can’t do as much as Zbrush, but it’s an INCREDIBLY user friendly and flexible sculpting tool. It’s a shame Autodesk Old Yeller’d it, because it was the best designed of their products imo. But yeah, Mudbox has sculpt layers that are right out in front of ya with intensity sliders and mute buttons, a sensible outliner that’s actually a lot like Blender’s current one, super easy brush texture management where you can just drag in hundreds of textures at once and it brings em all in, and the ability to sculpt SUPER dense meshes. Not as dense as zbrush, but still denser than Blender by orders of magnitude. If Blender had to absolutely be comparable to any dedicated sculpting suite out there, it should be Mudbox. Imo.

3 Likes

Preach! I am old and love that after effects achieves great things with its fake 3D workspace. Yep it makes somethings a bit too inception (comps within comps within comps)… but I digress.

I must say that the “average” zbrush user (actually any sculptor) would do better concentrating on good shapes, primary and secondary forms rather than the tertiary details. Like Jeff Goldblum…. “Just because you could doesn’t mean you should”.

That said there are lots of good things ™ that have happened in sculpt. I’m not sure what the future holds but hope joeedh can get his dyntopo changes to main, and I realise as I type this I’m not even sure of the current state of multi res. Does it still have horrible random spike issues and discontinuities?

4 Likes

It does. Way less now than ever, but it’s still… borderline unusable…

Reading your comments I guess you’d be better off using ZBrush. Why wait for Blender to become something that already exists and suits your needs? :slightly_smiling_face:

I own ZB since version 4R7 (love the continually free upgrades), and use it next to Blender, making use of the great GoZ for Blender add-on.

Having said this, Sculpt Mode has advanced so much the past few years, mainly thanks to Pablo, that I sometimes skip ZB, also because it’s so convenient having everything at your disposal inside the same tool: a great poly modeling toolkit, very useful sculpting tools, two fine renderers and much more.

My biggest issue with ZB is the lack of a competent GI renderer. The ZB ➔ Keyshot workflow is inconvenient.

12 Likes

Ah. Unfortunate.

I haven’t had any issues with this since 2.9 I think. As long as you don’t make any topology changes on the base mesh (like adding or removing loops) there shouldn’t be any issues with spikes. Before they could appear randomly just by making small changes on lower resolution levels.

The biggest bottle neck is still performance…

3 Likes

Joseph begins the next refactoring project related to sculpt mode.
rB350e783668d6 (blender.org)

One of the major aims is converting the code to C++, which will unlock new opportunities to modernize and optimize the mode. For those wondering, he still has Dyntopo on his plate as well, not an easy job because of the all of the corner cases that can come up.

15 Likes

Really odd, especially considering that literally no other package has these issues. It’s the main building block of sculpting and made me loose all confidence in working with multires in general.

1 Like

The main problem here is that the sculpt mode was designed to support also animation/pose/shape keys tools. Too much systems interleaving between the same data is inefficient and prone to bugs and strange behavior. Other packages doesn’t have the same problems since they don’t try to make the internal systems compatible, like in Blender. That’s why in Maya you can do sculpt, but AFAIK their tools are kinda of terrible (experiences can vary), but their sculpt tools are stable, altough you can’t directly animate over it. In Zbrush you can’t animate, but there’s no surprises when you use their tools.

Personally i think that only a full rethinking and rewrite of the sculpt mode and the entire mesh foundation of Blender can fix most, if not all the problems that plague the software ATM. The problem is that no developer wants to do that, since is too much work, and they dont’/can’t see the problem that artists (or at least, those that comes from more specialized packages) are having right now.

2 Likes

Yes, I followed the conversation for a while with proposals to make it part of object data and so on, but eventually it seemed like it’s just one of those things Blender folks decided to do their own way and everyone can decide for themselves if they can deal with it or not.

Agreed, also I don’t think that’s even realistically possible given that ZBrush uses their own 3d representation algorithms which are entirely CPU bound and allows them a great level of flexibility in how they display 3d data in a very performant way.

It sounds counterintuitive to forego hardware accelerated 3d this way, but the results are undeniable, ZBrush handles insane amounts of polygons with ease.

However, for Blender to cook up something similar for its Sculpt mode is unrealistic, first off Blender is not just a sculpting program (far from it), and secondly not even other commercial sculpting software has been able to come close to ZBrush’s performance.

Like you said, mirroring Mudbox’s capabilities is a realistic goal for Blender, and one I hope they pursue.

3 Likes

I personally wouldn’t use more than 20-30M for a single mesh, but seeing that 3DCoat can handle the same amount as ZBrush if not more, I can see why for example some environment artists would be drawn towards those packages.

2 Likes

Wasn’t that the motivation for the revamp/reorganize/new editor modes that were proposed? If you just want to sculpt and poly-paint with layers, without multires, with only 9 brushes, on a 700-million-poly mesh, there’d be an editor/mode just for that. At least that’s how I interpreted it.

Yes, and we already seen where it led. (granted, the ideas wasn’t really well redacted, but whatever).

Software development is hard.

1 Like

I had to look this up because I didn’t believe you. This is really surprising. They don’t even give a proper recommendation for a GPU on the website. Anything manufactured after 2008 is good enough :open_mouth:

I almost want to try out Zbrush now, but i don’t do enough sculpting to warrant it. Who knows what next year will bring though.

1 Like