I need an advice on improving sculpting skills.
Examples:
Reference:
Result:
e8763949189041ab9a8ccbd1d62d5ec6
Exercises or something like that?
For example, a good way to quickly boost 2d drawing skills is gesture drawing. What would be an alternative for 3d?
Software used: blender 3d.
Yeah, practice, practice, and practice. It’ll get easier the more you do it, but that’s not really useful to hear, so let me list a couple of things that come to mind.
It looks like you increased the density of the mesh way too early. This is why you have a lot of wobbly-looking geometry (eyeballs, eyelids, nose) because it gets a lot harder to smooth everything once your mesh is that dense. Concentrate on the shape first before you tackle the details. Only increase the mesh when you have to, not when you want to. This is a very common mistake when you’re first starting out.
The eyeballs are usually a sphere that is not connected to the head. It makes them much easier to animate and even when I do very low-poly art I use a separate mesh simply because it gives me more control without affecting the topology of the face.
The eyes are too big compared to the structure and style of the face (at the least on the bust with the hair; it looks better on the sketch and the sketchfab model, but even there you’re skirting on the edge of being anatomically correct). The space between the eye is usually exactly the size of one eye. Regardless of whether you’re leaning more towards a realistic or anime style, this is sth to keep in mind.
The ear is a lot bigger than most people think. The woman in your reference image has fairly small ears; usually it goes as low as the bottom of your nose and up to your eyebrows. You might want to model the ear on a separate mesh and then
If you’re going for a more anime-styled approach, you want to keep the noses slimmer and less pronounced.
If you’re sculpting with a reference and you don’t have much experience designing characters, try to match the concept art as closely as possible. Looking at the sketch and the head below it, the 3D model’s chin is a lot rounder, the space between the nose and mouth got longer, and the nostrils got a lot more pronounced (but to be fair, the sketch is stylized so that might simply not translate well into 3D, at least not with a proper shader to bridge the gap :p).
But yeah, maybe look at a couple of tutorials to see if there’s something that can help you, but apart from that, I think practice is the only way to really improve.
Thanks for the response.
It is a shame that “the only way to practice” is the answer, but I sorta expected this. :-\
Opinion (disregard if you want)
I don’t really see the ‘need’ to sculpt the model above. There isn’t any detail that needs sculpting to be accomplished later with baked normal maps.
The only reason I could see to sculpt is if the artist feels more comfortable (is faster at producing the art) using sculpting and retopo, instead of straight poly modeling with a smoothing modifier applied.
Recently learned
I recently listened to a suggested workflow while watching a sculpting video (sorry don’t have link). The video/artist suggested to control the viewport/camera distance to the model when sculpting to improve workflow and not to get too detailed before it is needed.
When roughing in shape and form for ‘the right’ silhouette - keep the viewport pulled back away from the model so the entire model is always in view.
After finding the silhouette and the form is locked in, refine as needed, then increase mesh density and zoom the viewport to be 1/4 or only part of the model. Here the artist refined specific areas of the model, and added/increased detail on inner shapes, edges. This detailing pass can be more than one resolution pass, but should be consistent level of resolution throughout the model.
Lastly - up res the mesh to the highest level to add fine detail to the model, mostly surface detail.
This is really kind of a given workflow, but I found it helpful to keep these techniques in mind to help improve form and surface flow of a model, before getting into adding un-needed detail too soon, which can cause the model to get out of hand.
The advices of krraej and theANMATOR2d are pretty spot on.
Personally, I’d not suggest to start with already stylized characters like the first one. Learn from real anatomy references at the beginning.
One good source is this site (contains nudity). It sells 3D scans of humans.
But the example photos are a valuable source for references, because you get the models in orthographic view, with t-pose and as clay render. Makes it ideal to learn from IMO.
I’m not sure about the workflow you’ve used here. I guess it’s Blender’s Dyntopo.
In my opinion, starting characters from a sphere and use Dyntopo (or Dynamesh in ZBrush) is a double-edged sword.
You have a lot more freedom and can experiment more. On the other hand you get bad topology and you’re not be able to go back to lower resolutions to work easily on the overall form.
I prefer to start with a human base-mesh, I’ve created myself and subdivide it when I add more details.
The human form is basically always the same. So you have a good starting point with the right proportions etc., which can be altered if needed.
Just to point at oddities in the anatomy, you’ve got some big chins and some weird jawlines. Chins mostly just protrude forward, and you have them extending sideways a fair ways. A thing about the jawbone is that it stops tapering once you get to the back, so the jaw ends up being square at the back.
It’s not actually that hard to get the resolution lower with blender depending on the dyntopo settings. There’s also the subdivision modifier (I think) that can be useful with sculpting.
For sculpting in Blender you usually use the Multiresolution modifier. But it’s not supported with Dyntopo. Same goes for the Subdivision modifier.
You can change the resolution of Dyntopo. That’s right.
But you can’t go back to a lower res mesh, due to the nature of how Dyntopo works. It creates new mesh topology on the fly and does not simply subdivide a mesh.
The reason why I sculpt and not model is because dealing with large number of points ticks me off, and sculpt-basing workflow is more intuitive. So it is not a matter of being comfortable (as you said), and not about having more detail.
I tried subdivision surfaces and poly modeing, they basically are much slower due to constant need to deal with topology.
When I sculpt (which isn’t incredibly often) I usually make a model with very basic detail and export it to sculptris. It really helps to make sure that the foundation of the work is good, and that if for some reason I mess up and can’t go back, I’m not going to have to start with a sphere again and go through all of the blocking out.
I think the most important things with learning sculpting, pretty much as krraej said, is to really nail down the idea of working at uniform levels of detail at each step and not jumping ahead too fast. Especially if you’re doing something with a smooth curve like a jaw or buttocks, it’s very easy to do if you do it right the first time, and very hard to correct once it’s hit a high level of detail.
If I were you, I’d practice by trying to achieve something like what you posted but with much less density, something like a game-ready topology with maybe one or two subdivisions.
are you apart of any of the more artists oriented communities like polycount, there are a lot of really good sculptors there that could critique your work
I sorta already tried asking similar question on polycount and received no responses.
Is there a specific polycount section you’d recommend for this? Overall it sounds like you’re asking me to go through WIP and work on a single model for a long time. Is that correct?
Few people mentioned low-detail base…
Would something like this work for this purpose?
That’s too low detail.
You need a “low detail” that is generally on the low poly side, but has enough polys so that head shape is clearly defined.
Which means if there are any anatomical issues, they should be obvious at this stage and should be fixed at this stage.
It’s meant to make you focus on the base shape, which is like 90% of getting a head right. Once you start recognising what is wrong and start to fix it, you can go ahead and sculpt any way you want.
(btw, the most glaring issue in your models I think are the jaws, on both examples they look super weird)
(also, disclaimer, I kinda suck at making heads too, but with great effort I managed to get from “OMG WHAT IS WRONG WITH YOUR MODELS” to “those aren’t too great” and I think most of my problems where from getting the base shape wrong and the only solution to improve was, make a head, try to identify what is wrong (which is harder than it sounds, try getting someone else’s opinion as well, it may open your eyes to what is wrong), try to fix it and repeat)
I agree with AcidArrow. It’s a bit too low poly.
But there is a bigger issue here and that is your topology. You have a lot triangles. Especially around the mouth.
For base-meshes that are supposed for sculpting you need to avoid triangles and poles as much as possible. Otherwise you’ll get weird issues especially with smoothing at higher levels.
Also try to keep the quads as even as possible.
For reference, here is a female base-mesh I did in Blender.
The head, hands and feet have a denser mesh, since it need more details.
I know proper topology (been looking into face loops before I settled on sculpting), the thing is I use dyntopo sculpt, so the moment I touch this with a brush, any topology will be destroyed and it will turn into poly soup. Basically, I’m only looking at getting shape right, and making proper retopology wouldn’t be an issue.
Could you elaborate on that part? Because it is something I do not “see” or “notice” myself, meaning if you point the problem out it’ll probably help a lot.
I recommend this CGCookie tutorial - it’s archived so I think it’s downloadable on a free account. Anyway I found it very useful.
I’m aware of that.
But I thought you want to try the multires-modfier approach with a solid base-mesh, rather than dyntopo.
You said it yourself. What you get with Dyntopo is a mess, and you can’t go back to lower res later on.
Study the underlying musculature flow and make your positive and negative deformation strokes flow along those. It helps to have the cage mesh flow along muscle lengths as well. The deformations look more natural. Plant the root of the hair polygons below the scalp. Extra polys are needed around nostrils, eyes and mouth to be able to get those subtle details that indicate the topology of a face.
On the more “anime” like character, while it looks like it should have a pointy like chin, you made it kinda flat and too “bulky”. Anime chins are usually more effeminate.
On the second character, you again, have a very strong and “flattened” chin and the jaw in general is a bit weird. Like : The corner of the jaw line (below the ear), is very important in giving the overall shape, it’s not very well defined in your model. You should be able to see a corner when looking it at the front. Then it folds inwards a bit, making her look older.
Also around her chin, I can’t see the “bone” structure. The chin itself feels very strong and bony, but it folds inwards too much as you go to the sides.
I guess in the end I should just suck it up, find a good looking sketchfab model and keep trying to replicate it till it works,.
Thanks for the responses.
