Advice with proportions when sculpting likenesses
Hi
Hope this is the right place for this, but I was wondering if other character artists out there could give me some advice when sculpting likenesses? Specifically, for getting the proportions?
Often, I’ll use Photoshop and do paintovers on reference images, use measuring/guidelines, and overlay a WIP mesh and scale it over a reference. However, like with a likeness I’m working on at the moment, that hasn’t been helpful. The subject I’m working on doesn’t have many high quality reference images as they were taken years ago, their head is always angled, and the focal length differs from shot to shot.
Normally, I always work in orthographic view. I think a few of the references I have use perspective, though I couldn’t say what the focal length is. I think my likeness is getting there (slowly), and I’ve added a lot of the main landmarks, forms, and underlying muscle, but I’m still not sure if the proportions are quite there. They may only need altering a little bit now. I have been training my eyes more than anything, so I’ve been eyeballing it as much as possible. I’ve also tried zooming out on the mesh and looking at it as well as standing up and looking at it from a distance. I’ve also tried using the Transpose Line in ZBrush to measure the distance between landmarks in combination with the Ruler in Photoshop, though because of the size of the images/focal length etc. the measurements are always smaller.
Sooo, yeah…Was just wondering if anyone had any advice or tricks for getting the right proportions for a likeness/portrait? It’s usually for things like the positioning of the eyes, nose and mouth in relation to each other.
Thanks
Replies
You are doing the right things. There is no magic trick. Just keep going and train your eye. If you have done hundreds of heads you need less measuring. But doing likeness is never easy.
And do some rendering outside of zbrush. The viewport in zbrush is lying. 😁
Okay, awesome. :) I guess the best place, apart from UE4/5, would be Marmoset then. I usually feel fairly confident with doing likenesses, but this latest one has been challenging. Really struggled yesterday, but I feel it's getting there now. And the corrections I've been adding have all been through observation, so... 🤞
I wouldn't mind some feedback off various character artists if possible. Maybe you guys will spot something I've overlooked? :) My chosen subject is the late Peter Steele from Type O Negative. The plan is to make the character Jackie Estacado from The Darkness but using Mr. Steele's likeness.
Here's how it's looking in Marmoset. I still have the eyelids to refine a bit and I've not yet added any asymmetry. Once I'm happy with how it's looking, then I'll move onto the asymmetry. :)
Just a quick mention: Not a super expert but how i see this being achieved correctly would be what you are doing but along with, studying the structures of the human head, make basic shapes skull jaw, then the STRANDS of muscles that work the head, the veins if you want to, then FAT deposits, with the control over the skeleton,skin + muscle you can theoretically make any face at anytime with a fantastic base as a guide and in short order, which would be most beneficial. This is what i am finding out anyway from the past 1-5 years experimenting with "being quick & exact".
With the grainiest of salts, All the best. Edit^ all separate subtools btw..
Hmm. Interesting approach. So, do you have the skull as a SubTool then once it’s finished you duplicate it and add muscles to that one, duplicate that and then add skin/fat over the top of that one? So in essence - 3 SubTools (or 4)?
Isn’t it a bit awkward having to change the proportions for each SubTool based on a subject’s proportions? And so everything lines up correctly?
Also - just thought of this and I’m genuinely unsure - but don’t the landmarks of a skull vary from person to person? I mean, some people’s skulls are bigger/smaller than others, so wouldn’t some people’s orbital sockets, mandibles, zygomatic bones etc be slightly higher up or lower down?
Prevent yourself from ever using the Zbrush "shift snap to front view" shortcut, at all costs - Because there is no guarantee that the angle of the head in your model is the same as the angle the head was when the photo was taken. Meaning that if you do snap to front view and match what you see on a good front view photo ... you might be applying the fixes in a slanted way. You can test it yourself by manually angling the head model (without shift-snapping) in a way that matches one of your references, and then shift-snap : most often than not, the "shift-snapped" front view will be quite a few degrees different. Meaning that all fixes that could have been applied that way would have messed up your sculpt further rather than fixing it.
On top of that, Zbrush has very poor perspective representation. I would say that most of the issues that your model shows (elongated shapes, feeling of a character "looking different in front view than in side view", and so on) are actually 99% caused by these Zbrush-specific idiosyncrasies. Now of course there are people out there who manage to bruteforce through that, but I've see first-hand some a world-known modeler (and Zbrush user) being tricked by these very problems. If anything, sculpting in ortho might be the safest.
I'd say do yourself a favor and attempt to sculpt at least the first few levels in Blender or Mudbox. It might seem convoluted and overkill but you'll be surprised at how much easier it will be.
Also : facial features are actually not necessarily the most recognizable thing about someone. You want to include the hair as early as possible ; and also, subtle grayscale values for eyes, eyebrows, lips.
Last but not least, anatomy knowledge in general is very important : As accurate as you might try to be from the outside in, the best likeness sculptors are the ones who know how the head is actually built. That is to say, a 100% "technical modeling" approach likely wont be enough as there are just some many things to be aware of with this subject matter.
Again - some very insightful tips here. I never even considered that about the snapping, which makes sense. That's definitely something I'd have to do if I was using the Transparency slider. It would be quite time consuming if I was overlaying in Photoshop and trying to get the right angle/position. I guess I'd have to use the Transparency slider with a reference in PureRef first, get the right angle/position and then render it. And also save the custom camera position.
In the case of my model though, I've been primarily using orthographic. That's what I prefer doing when sculpting. I may have used Perspective once or twice whilst overlaying a couple of specific references that definitely have it. At some point, I really need to re-learn Blender. That's the first modelling package I used, but since using 3ds Max and Maya, I've totally forgotten how to use it. 😖 I understand that Blender is quite good for sculpting. I want to give it a try at sometime. Likewise, I do want to make an entire face/head from the inside out. The current projects though have a time-limit, so I don't really have the time to go all out, but instead, work from the outside in and add the landmarks.
I also made a blockout for the hairstyle, but it's very VERY basic at the moment. A blob, essentially:
Also, I started off with a base mesh, so a few landmarks and the anatomy were already included. Though I did re-add/enforce the body anatomy.
I haven't read the whole thread so forgive me if this has been mentioned. I saw a youtube vid comparing zB and Blender and the presenter mentioned altering the camera focus length in Blender to match the model as closely as possible to an image overlay, then making alterations from there.
The video I am refering to is : https://youtu.be/0ZvyqOAXMZI and the section starts around 6:00 mins.
Hi there again - well, by "inside out" I don't mean to say that one should do a layered anatomy study in sculpt or anything of the sort. What I am saying is, if you were to work on your portraiture skills in 2d for a while (absorbing what makes a face ... a face), then your ability to make likeness sculpts will grow significantly. Because you'll learn subtle little things about the different kind of noses, eyelids, bone structure, and so on. And doing it in 2d is about hundreds of time faster than doing the equivalent in sculpt.
As for the shift-snap issue : this applies as much to perspective as to ortho. As a matter of fact, if a sculpting program other than Zbrush had this feature, it would be just as problematic even with correct perspective.
Unfortunately the one single, biggest dowside of sculpting in Blender is that it still doesn't have a proper planar brush (Zbrush doesn't have one either, for that matter) - meaning that one has to use some annoying workarounds (like masking and smoothing) to create some true flat planar breaks. Nothing impossible of course - it's just much, much easier in Mudbox for instance. I'd assume that 3DC probably has a decent one too.
On a technical level I would say that you still need to find a way to apply painted values to the face as you work on it, as there is zero reason to work on such things linearly (sculpt then texture). The more you work on all elements at the same time, the more efficiently you'll reach the end result.
You could make the blob hair fully unlit black, as it will create just enough suspension of disbelief and won't distract you when working on the face.
Ahh, yes. Outgang/Laura Gallagher. I’m familiar with her videos; they’ve been super useful. :) I’ve still yet to watch that video, mind.
Hmm. I guess that can be a little side project for in the future then. I’ll have to find some portraiture courses and do a bit of sketching. :)
I wonder if anyone out there has made a custom planar brush for doing such a thing.
Regarding the painted values - how do you mean exactly?
For unlit black hair - can that be achieved in ZBrush or would that have to be done in a renderer? The only thing I can think of in ZBrush is applying the Flat Colour material and putting it to black. 🤔
I classical sculpture things are build from one or few reference points , for portraiture that point is intertragic notch [its in teh middle of the face but its as good as any ,any other point could serve the same purpose as long as it stays stationary ] You use that initial point to compare with any other to establish proper relations [angles and distances]
For likenesses thirds division of the face is very useful , in ideal case they are equal but in most ppl one is longer
You could use grid like system to compare position of smaller forms comparing with already establish points
Side note : likeness is hard and to have a better change of success I suggest highly to master generic head structure and skull first , walking before running
On painted values : I just mean that you have to find some way (out of the many available, from regular UVs to vertex colors ...) for your sculpt to display local values on eyes, eyebrows, lips, and so on. Just because many people seem to start from a blank monochrome sculpt doesn't mean that it has to be that way ! And I refuse to believe that you don't know how to do that :)
Oh, I see. You mean to sort of block-in the right kinds of colours and colour zones for each area to see how it's looking?
@pior Looks like your suggestion regarding positioning the mesh to match the reference more closely has paid off. I've done that against a few front and side images and now it's looking more like him. :) It's quite fiddly though, trying to align things up and getting the right angle.