So I started this project a few months ago in order to fully follow a AAA character production from scratch. My computer not being that powerful, I sort of started off at a PS4-like objective in quality, with options to upscale it with a subdivision later to match today’s standards.
For now, I have done the sculpt, the retopology, the UVs and the first iteration on texturing. I am now working on a rig to be able to pose and animate her. On the rails there is hair with xGen with both hair cards and strands, as well as the clothes with Marvelous (and the resculpt, retopo and skinning of the clothes...). I am already around 80 hours of work, rigging not included.
Topology around 30k tris
UV maps split on UDIMS
Close up on the face topology
Close up on the face rendered with 1 subdivision level
Body rendered with 1 subdivision level
Back view of the body with one subdivision level
So far I am very happy with how the topology went out and the texturing is decent, but it still feels a bit blotchy and undetailed. I wonder if it would be worth doing a higher resolution resculpt to add more detail to the face and have slightly crisper normals, but that might be overkill for something that is supposed to be a game asset at the end?
In terms of art direction, I really wanted to stick as close as humanly possible to the physical traits of the Major as depicted in Mamoru Oshii’s film. Her physical features are a blend of Olympics athletes and Brooke Shields, whom I found to have very similar traits. For the carnation I went closer to Carrie Anne Moss and Juliette Binoche. During the entire process I had the original concept arts and screenshots of the film to maintain continuity in reference.
I guess I could use some feedback regarding the texturing in particular. I feel like it still can be improved, but is it really worth it?
You’re welcome to comment or make any other remark! I am fairly new to this type of 3D modelling and it is the first time I am using Maya as well, as Blender has pretty severe limitations in optimizing texture space.
Replies
The references:
And a render for good measure:
Now obviously with the ray traced render of marmoset I can tell that the texture feel vastly insufficient for a close up like this. Maybe the lashes would deserve to be denser or thicker too. Otherwise I am reasonably happy with the groom and am in good hopes it will export well in Unity.
If not done 'right' I think it tends to look freaky and unsettling...
I'm not sure adding those huge eyes onto Carrie-Anne Moss' face works so well without at least giving her some degree of stylization to help blend them together...
So i guess my advice would be to push for more stylized facial features that help work the size of the eyes better with the rest of the face.
Right now, there appears to be a disconnect between the realistic face and the huge eyes.
I wish I could be more helpful and give more technical suggestions on how this could be achieved better, but i can only give my critique on how it looks to me.
I think the 'Alita' movie did a good job of achieving this kind of thing and managed to pull it off well.
Again, I think it's a really tough challenge and look forward to seeing how you progress
I talked about it with friends and we compared the render with the model in the viewport, one of them noted that the iris might be too big while not having the proper falloff with the cornea and that’s what gives off the anime/artificial vibe. Also the camera on the last render render is so close, the lens does distort the features of the face, making it appear both skinnier and with a bigger nose and eyes than what it is.
Whereas without the shading and lens distorsion both the bone structure of the face and the eyes mesh look physically accurate (for someone with pretty big eyes, but still in the realm of okay).
So the details cited above are definitely something that is going to change. I also might inflate the eyelids a little bit as they now appear too thin to look real, and I think this also will contribute to reduce the perceived size of the eyes and make it more natural.
I spent the afternoon checking on casting websites to find that type of feature and yes, there definitely are women with eyes so big they’re a bit scary, and that uncanniness is something I am looking for, albeit finding the exact balance is difficult. I’ll iterate more with a collection of pictures of these, it’s a rare feature, like less than 1% of the actresses and models I found had it, but it exists.
Just to clarify, I didn’t base her face off Carrie Anne Moss, that was just for the eyebrows and the skin carnation. And again I think I’m not there yet, the skin is still to splotchy and lacks the type of illumination that I’m looking after.
Thanks again for the feedbacks! It is useful to iterate this way
During retopology, for both pieces of clothing I tried to preserve the body’s original topology as much as I could so the mesh still will fold well when animated; projecting the original mesh then arranging the extremities so it subdivides correctly on zbrush. This also should allow to preserve the UVs on the catsuit so it can bake some of the skin color to give the illusion of translucency during texturing. (If everything goes according to plan). Ideally as well, the meshes will be welded together with a single material and no risk of overlap/clipping when posing.
I’m now happy with the overall shape, it does need some more clean up to clear the self-intersections. I’ll render it in Marmoset later.
References for the hair cut: