/3/ - 3DCG

>>389003
Try it and post results.

>>3890821
I tried. Excuse my shitty modelling skills.

>>389093
final results look nice but the form of the eye isnt exactly like the 2D drawing

Creasing seems to work well enough, although it took me a few minutes to realize that you need to crease the edge coming out of the eye corners as well.

I suppose creasing would be the best solution, but maybe I'm wrong.

>>389094
While I'd love to make that character, I'm just trying to make a face that works and I was still in the process of shaping the face and topo. I just slapped that together to test this out.

If you think about it it doesn't make any sense to have the eye as a separate object for anime, in drawn anime the eye is - drawn - not modeled.
So much of the anime look and feel is eye centric so doing eyes as texture makes much more sense.

Way I would do it is to have the iris of the eye on a sepparate UV set and animate the placement of that for eye movement and use texture swapping for the
eyeshape and mask.

Now even on realistic looking characters that area you're highlighting is often a issue so you need to deal with it in mesh too if you go down that route.
Way you do this is to fold the eyelid back so there is a rounded edge before the eyeball instead of letting it sharply intersect.
Another useful technique is to add a separate band of polygons to sit on top of both the eyeball and the eyelid and mask it of softly with an alpha map
to suggest the presence of a fluid layer catching a highlight where the surface tension causes a bent surface prone to cause them, commonly referred to as the tear line.

>>389096
Subdivide the nice base and then adjust the edges of the subdivide result so that it's a sharper line and corner. Then when you do a smooth, it will hold together. The base will still be low poly enough for easy animating.

>>389125
is there a method in maya for interpolating the texture swapping of the eyes so that it isn't instantaneous or fade-ins?

>>389151
I don't see a particularly good reason to make the eyes part of the face mesh as was suggested. Obviously the eyes are textured but if you look at all the decent anime characters, the eyes are separate as in any characters. The main difference is that anime eyes are almost flat, instead of round, and there may be multiple layers of geometry for the different graphical aspects of the eyes. For example, you might want a physical specular highlight that is separate from the rest of the eye textures and can be positioned freely.

>>389160
Nothing stopping you from masking off the area and running alternative specular and normal sets on a decal sitting on top of a surface.
Having a bent normal map on a flat eye is probably the best way to go for successfully replicate an impossible shape like a 2D style anime face in 3D.

>>389125
>in drawn anime the eye is - drawn
I'm not sure how you're making a point in this. You could say the same about any part of 2d.

Regardless, the best solution for 2d is using floating facial parts and inverted normals. Using texture swapping may work for games but I don't think it's a satisfactory solution for quality work compared to inverted normals and floating parts with flat shading and single-sided faces.

>>389175
From the sound of it, I think inverting normals is probably easier and better, but I'm not sure precisely what you're talking about.

>>389231
Inverting normals to what end?

>>389236

bent not inverted, and to give the eyes shading of course. I started a little side project to perhaps try it out but I ran out of time to set up the eye system I was talking about, see WIP thread.

>>389236
>>389242
No, I do mean inverted. Maybe I'm wrong, but I don't know of another way to easily produce effects like the girl's mouth in the OP post, namely the common anime practice of putting the mouth at a place that would make it much farther back on the Z axis(not blender's) than it normally would be.

Look up a model called Lat Miku and if you look at what he does you'll see what I mean. You could say his model is sloppy or bad but the technique itself is good. It allows you free placement of individual facial parts(since they're floating), even at places that would be physically impossible to reproduce without complicated morphing or other techniques(without deforming the apparent shape of the face).

The only problem is that it only works with flat shading. I think. There might be solutions to that but nothing I've personally tried.

I didn't even need extra vert or special topo to do this. Just model with smooth preview turned on so oyu know where to move verts to compensate for the smoothing at the end.

>>389244
I'd note that if you had an effect of some sort which causes vertex morphs to be change based on camera angle that would work just as well. For example, a morph that activates as the camera rotates to the side of the model which brings the mouth further into the head and moves other parts forward to compensate. I still believe inverted normals(or inverted faces if you want to call it that) is a simpler solution, however.

>>389246
Sorry, I can't tell if you're misunderstanding or I am, but the left side is the goal, not the right.

>>389244
Or here, I'll post a picture.

Inverted normals are what allows for this effect. Maybe there's another way to do it I don't know about, but it seems like the best technique to me. You can see that the mouth appears to be farther in than it's actually modelled, and it's viewable from all sides. These are pretty key concepts in the anime and manga industry.

For example, if you look at the OP picture, you can see that her mouth is wayyyy farther back than should be physically possible. Using inverted normals it's not too hard to accomplish, at least in theory. Just move the mouth back.

>>389250
Here's a wireframe of that face. It's in tris because it's a downloaded MMD model.

It can also be used for things like this. If you look at the red circles you can see that that piece of hair is visible from "inside"(relative to her face direction), but looking from "outside" it isn't, which makes it look like she has more hair than she really does while still allowing you to see her face well from the side.

>>389250
>>389252
>>389253
Alright inverted normal = visible from the "other" side, so it's invisible from the typical "original" side. Meaning that basically the area that would normally hold the mouth is invisible... so as to allow the deep-set mouth to be rendered on top of the face. Correct? But the mouth itself would have to have default normals...
I'll just download the model thanks for the tip

>>389250
Is there an obj of this model? I can't be fucked with MMD.

>>389255
Alright I just downloaded her and found that the outline for the face is actually baked into the model, whereas the rest is done by the shader as usual. I suppose most of the face has inverted normals so as to be able to render the mouth "on top" as you will. When you activate outline shader on the face it does look like that's the case. Also the outlines are baked on those hair strands you pointed out because again the outline shader fucks them up due to normals being inverted to begin with.

>>389255
That's correct. Sorry, I'm really not sure what terminology to use. I'm no professional.

>>389261
Yeah, the facial edges are a separate model. And yes, as far as I know every part of the face itself has inverted normals and then the facial parts themselves have default normals, and it also has a section at the back for shadow under/around the face.

I'm sure this technique is capable of a lot but I never see it used. It has one big problem in that you can't use calculated facial shadow or shaders because the normals are inverted and the facial parts don't align with the mesh of the face itself.

The modeller leaves his model with a flat color for a texture, but I think if you use a partially transparent duplicate model of the face you might be able to paint detail on the face. Or maybe even calculate shaders or shadow on that "dummy" face, but that's way beyond my expertise.

At any rate I don't know of any other methods of getting these kind of effects. However, I've only seen them on commercial models once or twice, for some reason.

>>389259
Assuming you're not >>389261 and you still want it, I can convert it for you if you'd like.

>>389264
I'm not him, but I'm also about to model some animay grils. It'd be really appreciated if you could convert it!

>>389264
Well one thing you can do in regular 3D is simply render the mouth on top and leave the normals as is... you would at least need a custom shader that writes Z buffer to do that in MMD or real-time, but as for Maya you can render your parts in whatever order you want and stack those layers for the final render.
The same thing can be done with eyebrows if one were to desire the somewhat common style of eyebrows being drawn on top of everything including hair.

>>389267
Hmm, that is an interesting idea. But in order to get the effect of having the face set back on the Z axis like in >>389250
you'd still have to model the face differently than normal, right? It is a good halfway point though because it would allow you to view the mouth from the side, or so it sounds like.

>>389265
You don't need textures, right?

>>389268
Yeah, you could just make a closed mouthless face, then a separate mouth model inside the face as with the Lat model. There's probably more stuff you need to do to make it look homogeneous for different situations... In side view, the mouth is typically part of the face as always... But the same holds true for the nose which is more "physical" when viewed from the side. It's not that difficult to set up camera-based changes.

>>389268
Not if they're a hassle. Geo should suffice but if you could textures would be nice too.

I'm just curious to see how the mouth is done above everything else. I can't visualise what you mean when you guys are talking about it.

>>389272
Yeah I just checked out the blend shapes on Lat's mouth and that's pretty much where that technique ends... it's OK when seen from the front but from the side you can't even see that the mouth is moving, and you would expect the silhouette of the model to move like with a normal model, rather than something inside the face.

>>389272
Heh, I'm not really even sure which is easier now. I liked this technique I'm talking about because you don't really have to worry about topology much or at all, just make facial parts and as long as the shape is right the topology can be anything animatable, more or less. More specifically it allows you to just concentrate on how best to form the face parts for the purpose of animation without worrying about how they connect to the rest of the face. Plus model parts are really easy to exchange(to change eyes you just swap out eye parts, for example), and model separately period as opposed to making a complete head, and it should work in any 3d application that allows for one-sided faces.

It sounds like this method you're talking about might be more powerful, though. You said that you do have to model the face differently(like moving the mouth mesh back and creating a duplicate placeholder on the face mesh), but are you still really able to get good shading/shadow results if you do that? I don't mean to question you about a complicated topic on 4chan, though. I don't imagine there are any good examples of it in action?

Also I'd like to note that this model is popular. Like, really, REALLY fucking popular. Historically the most popular MMD model of all time, even though it isn't the best. All of the best people in the community used it, literally. Just saying it has appeal.
http://depositfiles.com/files/ta0wogx8n
You can get the textures here
http://bowlroll.net/up/dl3934

>>389278
What? Just because Lat didn't do it doesn't mean it can't be done. He just didn't model the mouth moving, that's all. That's not the fault of the technique. Regardless, it's a technique with applications, mouths aren't the only thing. Sonico's model in Sonicomi uses it to change the shape of the face from the front and sides/back.

And I'm not sure how much anime you watch, but as far as I know this is actually a pretty accurate replication of what's actually done in anime/many people's art styles. Obviously not everyone wants this but you don't have to do it this way either.

Also, you can attach the mouth to a bone and move it forward and backward, or rotate it manually if necessary.

>>389282
Thanks a bunch for the link.

>>389282
But I already pointed all of that out in my other post... Yes you can use that technique, but not by itself, you have to create a whole rig around it that's controlled by the camera angle, it's not suitable for real-time.
And TDA's model is more popular as well as superior in many ways, and does not employ that technique.
Yes some artists draw a separate mouth inside the face even in side view but it looks like shit. Generally it just becomes a more anatomical mouth.

>>389284
Hope it helps. I doubt this is for everyone but it's easy to do if you understand it, although like I've discussed there are limitations.

>>389285
Yeah, even if it's something I don't end up using I love looking at new techniques and seeing how other artists solve these issues. I'll try making some stuff tomorrow and tackle it as I go along.

>>389286
>Yes you can use that technique, but not by itself, you have to create a whole rig around it that's controlled by the camera angle, it's not suitable for real-time.
You're talking under the condition that you DON'T want the mouth to be back like that on the side view, right? You're absolutely right if you want to have the mouth actually be at the front from the side view AND have it appear to be "back" at other views, you would need to do some special shit. But I wouldn't write it off as it is, because it's a popular and common method of expression even if you think it looks like shit. And it's been shown to appeal to people, so I wouldn't say you 'can't use it.' If Lat's popularity is any indication, people like it.

And either way, it doesn't matter if you don't view the model from the side much/a all. It still provides you with something that you can't get without the technique you're speaking of, which sounds like it might be difficult to produce in simpler programs, or games for that matter. Depending on the animation you're doing it the view from the side may not even be much of a question.

And yeah, TDA is a much better model, there is no arguing that. In fact it blows Lat out of the water as a model in general, technique or not. Because of that, I don't really think it says anything about the technique, though. I think if you made a model of TDA quality with this technique(or yours, if possible) it could be very successful.

Either way, I wanted to introduce it to anyone interested. It seems useful, easy, flexible, and efficient, even if it has some flaws/not for everyone. Or at least usable in your bag of tricks if you're creative.

>>389294
Ugh... too many typos. I'm getting tired.
>Depending on the animation you're doing, the view from the side may not even be much of a question.

>>389287
I hope you'll post if you make something.

>>389296
Yeah, will do! I'd like to mess around with different techniques and get some feedback. I don't have time for anything tonight, but I'll probably be starting it tomorrow.

>>389096
I attempted this solution, and of course placing a crease on the adjacent edge resulted in a dark shadow.

So I pulled a little panel out behind it and creased that instead. Seems to work.

>>392167
That's pretty ingenious, or at least I might never have thought of it. Or noticed it, because I'm only thinking of flat shading, where all of these seem kind of equal.

Without going into a wall of text, I went and tried out the different ideas again. Here are two images of the results. It seemed like the vertex-adding method was better at first, but I think in that image I moved the corner of the eye a little too far back. After moving it into a less extreme position I got the results on the right. So, I'd say yours is probably the best option, if we're considering shading the model, at least in terms of shading. I imagine it's probably easier to create vertex morphs with than adding two vertices as well.

I also tried using the >>389093 method and creating an ngon instead of using tris, which also gave a(different) unwanted shadow, no surprise there.

Thanks for posting that.

>>392167
That's pretty ingenious, or at least I might never have thought of it. Or noticed it, because I'm only thinking of flat shading, where all of these seem kind of equal.

Without going into a wall of text, I went and tried out the different ideas again. Here are two sets of images of the results. In the left set of images, it seemed like the vertex-adding method was better at first, but I think in that set I moved the corner of the eye a little too far back. After moving it into a less extreme position I got the set of results on the right. So, I'd say yours is probably the best option, if we're considering shading the model, at least in terms of shading. I imagine it's probably easier to create vertex morphs with your creasing method than adding two vertices as well.

I also tried using the >>389093 method and creating an ngon instead of using tris, which also gave a(different) unwanted shadow, no surprise there.

Thanks for posting that.

I also wanted to post that there is a possible disadvantage in using eyes that are just textured onto a face, like some people suggest. There's a common technique among Japanese models where you give the eye a vertex morph that makes it a concave shape. This gives the illusion that the eye is always looking at you, so it's useful sometimes since it means you don't have to animate anything to have the model(appear to) focus on the camera.

Although I'm sure there are techniques for this in other modelling packages which will work better, this is a simple optical illusion so it works with any engine or medium, and I'm not sure how it would work without a standalone eye part that's fairly high poly. Maybe it would work fine to just make the textured area concave, but it seems easier if the eye is a floating part.