Modeling sub-forum?

Is there a modeling sub-forum here?

I don't remember seeing anything like that.

Or will I be better off asking modeling questions elsewhere?

 

Just the sub-forums that can technically include modeling like the Animation section.

 

I don't think any of us are pretty enough to warrent needing one...

 

That doesn't really work in this case.... no animation involved :-\

 

There's really no point in having one when sites like polycount and the like exist, I reckon.

 

I think it would be a useful addition - so many of us are jack of all trades and it would be good to connect with others on different aspects of workflow. Polycount might be useful (haven't really been there all that much) but it's likely not going to be as relevant to Unity users as information found here - so much of what goes on there is portfolio work it seems.

 

Would you categorize it into sub-forums by tool, or would you just do a general modelling forum? I imagine you'd get quite a few people asking general application questions without specifying the application. Although it would still be useful to have a one-stop place within the Unity community about modelling. Might help some people rely a bit less on asset store bundles and create their own stuff.

I guess a good question to ask is: are there a lot of orphan modelling threads out there in other forums that would suggest a modelling forum would be frequented enough to warrant its existence?

 

There's no real reason to make a sub forum per tool, because there are a lot of them.
Also, I wasn't requesting a feature.

The actual question is "where do you go with a modeling question". Blenderartists community is largely lethargic according to my experience. According to @Murgilod it is polycount.

 

There's also "CGTalk". That's a bit older though and I'm not sure if it's frequented much anymore. Honestly, in the past five years I've had issues getting any kind of responses from modelling communities. Usually I just end up finding the solution on my own or finding a different way to go about it.

Polycount should be a good place to go, I think @Murgilod is right on that one. I found a lot of talented artists to work for hire from there.

 

Oh, so it is not just blender community, but modeling resources in general? That's good to know.

I think I'll check polycount next time I need advice with anything.

 

really for modelling, polycount is the place to go whether it be environments or or characters. People mentioned worries about needing a board for each forum, i don't really see that as a issue, most skills and workflows for modeling are pretty transferable between applications. Doesn't matter if im using maya, modo, max or blender i can make the same model in all of them, i just will be a little faster working in modo or maya since that is what i am familiar with.

 

+1 for polycount.

Do you still have that modelling question? Just ask it here in this thread while we're at it.

I don't see what could be so Unity specific about a modelling workflow. The only things I could think of are the fbx exporter and custom scripting solutions. Everything else from vertex colors to UVs etc. are basic blender knowledge or basic 3D modelling techniques, imho. Also portfolio pieces are usually the more interesting ones compared to streamlined production work. Portfolio artworks often are the ones where new techniques are tried first, plus for some artists it's just the only stuff they are allowed to show. You really should check out polycount to learn about modelling and general 3D workflows. I'd avoid any blender forums unless you are looking for things that are blender-specific like the scripting API, exporters, modfifiers etc..

Vice versa, if you are an artist and have a Unity-specific question, come here, don't ask on polycount. Last time I checked their Unity subforum was pretty dead.

 

Yeah I started to check out polycount daily, I'm not knocking it as it's a great resource and has some great info and inspiration there. I guess it's more a case of just being able to talk about the whole workflow with people here (I guess since I spend most of my browsing time here) and not fragment it between different sites where people have generally got a slightly different view on artwork (perfecting everything vs good enough).

Come to think of it though, it might actually not work out very well, since polycount will almost certainly have better expertise and a set a higher bar. I guess it just feels weird not to have a place to talk about modelling here, since I don't really separate it in my mind from the rest of game development.

 

For me that'd be a reason to go there .

 

It was about modeling 3d character head using "non-realistic" reference (think cartoon or anime) + making polygonal strip-based hair using subsurf. In blender. Also my modeling skills aren't very good.

Still interested?

 

Sure, but stylized stuff isn't my strong suit, so the chances of me being able to actually help just got a whole lot lower ^^. Maybe @theANMATOR2b has more experience with it? Just ask here and if no answer comes up, copy your post over to polycount.

 

@neginfinity - I can answer this one for you, the methods are same as for realistic characters. You can have a few or many vertices, you can use a regular or irregular mesh flow, one or many textures etc.

The more advanced, stylized characters often rely heavily on the artists shader knowledge and his understanding of shapes, colors, details etc.

Polycount is the best place for questions that I know of, some people can be a bit frank but that is because some of the users are professional artists at large studios like Blizzard, Crytek, Ubisoft etc... They know what they are doing and sometimes get tired of beginner questions.
A small tip is to search the forum before you post a question. The chance of receiving a good answer increases if you post your work together with the question so that they can see what you are doing and better understand where it goes wrong.

Sketchfab is a wonderful place if you need examples. Complete models online for your viewing. Make sure that you have highlighted the 3D window by clicking on it once, then press the 5 key and you will get a Wireframe view You can also find a settings icon in the bottom-right corner when you move the mouse over the 3D Window, it got more settings like changing the color of the wireframe (very useful on some dark models).

A few stylized characters with good mesh flow, textures and vertice amounts.
Tracer Fan Art
Leviathan
Risa

[edit] clarified

 

I think I'll just post the question. Please note that my modeling skills are still quite horrible at the moment.
-------------
I'm interested in simplified non-photorealistic rendering (meaning it doesn't go all the way

Problem #1:
Making polygonal model using non-realistic reference. Anime or cartoon. (the eyes are too high on this one, I'm aware of it, it happened because I messed up hair volume)

Spoiler

Issue: In source material eye placement is simplified (well, it is "symbolic") and partially ignores depth and perspective. As a result attempting to directly use the reference (meaning front/side view) creates very... ahem... strange head geometry which only looks good at original angles. I think that flattening the face a bit (most of polygonal models in this style seems to do this so it might work). Guilty Gear Xrd Sign somehow solved it (but they use 2d eye overlays a lot), but I haven't nailed the head geometry yet.

Problem #1:
Polygonal hair using poly strips made out of subdivision surfaces.

Spoiler

Subsurf modifier creates nice curvy hair strips, the issue is when those strips need to originate from "flat" or featureless hair top, the area where hair top splits into spikes produces a lot of stretchy polygons. Not sure how to avoid the stretchy areas while still using something like subsurf. The common approach seems to be sticking "polygonal hair strips" into character's head so they "grow" out of it or making solid hair with thickness to it.

 

That's what I meant

 

I assume you know this talk?

I suggest to go looking for wireframes or a making of of "knights of sidonia". Afaik it has fully 3D characters and I think they did them quite well:





Standard Polycount advice for hard to model areas is "add more geometry" because beginners often try to be too non-destructive and lowpoly with their subsurfed meshes. Sometimes you just need to add a bunch of loop cuts to get clean quad based topology. Unless you can make the single-hair-strand approach work for you (don't really see why it would be impossible), I'd try something roughly along those lines:



Now for the shading issues at the places where the edges split I'd try to do custom normal editing, which sadly is one of blenders weaknesses. I believe if you watch the GDC talk above they'll talk about how a high level of control over vertex normals is key, and blender has to my knowledge no good way to edit them manually yet. The best I can think of is use a "normal edit" modifier, set it to radial and position an empty object as the source of the radial normal projection inside the head. Like this for example:



That get's rid of the shading issues and you could also mask it with a vertex group to get more custom control over how the whole hair is shaded or you could use several projections and blend between them using vertex masks, to get the shading you want.

This is also the kind of approach that I'd try to use for getting single object strands to seamlessly blend into the top hair shape. Just make sure they share the same source object for the normal projection as the rest of the hair and both have 100% influence of the modifier at the area where the two meshes meet.

Actually I remembered another way that likely is even better:



This modifier projects the interpolated vertex normals from the other mesh onto this mesh and overwrites (or modifies) the old normals. You can see the original normals where the control mesh ends and the raycasts can't find anything to project from. Your control mesh should of course be big enough to cover every angle. That should give you finer control than a simple radial reprojection from an empty object. You can set the control mesh to always draw as wireframe under: object properties tab -> display -> maximum drawtype -> wireframe.

Did that help?

Ah, misunderstood you there. Sorry. I thought you meant that as a reason why polycount wouldn't work out for you.

 

Yes, watched all of it.

That's exactly the style I 'm going after. Sidonia, Ajin, and I think there were few more of those (I heard that new Berserk is also CG, although I haven't watched it). PS2 games by Atlus (Persona 3, 4, Shin Megami Tensei: Nocturne, Digital Devil Saga) also had something similar, except it was done in a cheaper way (because it is PS2 hardware).

I didn't think about searching for wireframes, though.

Uh... In this example, in a place where hair strands just start to separate...
Did you split vertices/edges without moving them away from each other?

Yes. I'll need to look into normal modifiers... and I'll need to make a polycount account.

 

Exactly, after the vertex where they split the next vertex pair shares the same position.

If you make another thread there please link it here, I'd like to see what the experts recommend.

 

@neginfinity - Ok, one thing that is good to know about classical 2D animation. They typically brake the laws of physics around a 100 times per second. They warp, bend, stretch and twist things all the time. They even move stuff around occasionally to achieve a more "dramatic" animation.

Do you have a picture of the source material?

A lot of things are hard to tell without a view of the wireframe of the model (the edit view) but typical, good looking 2D characters that are made into 3D have the same face mesh as a realistic model. The difference is that all the vertices (vertexes) are moved around to create the stylized shape. The reason for this is that the realistic face mesh is proven in animation and can even handle really strange facial animations like those used in cartoons (yes, it is that good).

The most important part is how the quads (squares) move from one area to another. This is the reason the mesh gets its smooth finish and why it works if you want to animate it.

Don't be afraid to make dramatic shapes, those gives you shadows and are the reason that people can read the face in the end. Typical Anime / Cartoon shadows are on the other hand impossible to get by just editing the face mesh. Normally, anime shadows are achieved by one of two methods.
You make a second face model with a face shape, bake it into a normal map and use it on the "pretty" face model. The other method to do it is by editing every normal by hand (very tedious).

Common areas that might get simplified in an Anime / Cartoon face are the mouth and nose like you would expect.

You can google "anime face wireframe" or "anime face topology" to get examples. You can also change anime to pixar or disney, maybe your favorite character in one of their movies?

Example (2 seconds Google but good result)


Here is an example of what a good, proper flow looks like, notice the similarities above and below.


Change the amount of quads, the number of "rings" in the flow between each area and you might get this.


You can easily turn it into something that looks like an Anime character if you carefully manipulate it.

Something I did notice is that you are using the subsurface modifier. That is okay but! You are applying too many levels of it, look at the top of your Blender window, it says 49.000 triangles. This is way to high, I can do a complete model with equipment at a total between 7.000 to 20.000 depending on how many details the clothing, hair and equipment got.

Guilty Gear Xrd is a wonderful aim but be careful and remember that they made a fighting game so they could make characters with insane amounts of triangles, animation bones etc. They also used special techniques that they researched themselves, wrote their own Shades based on the research and added lots of special cases to the system.
Example, every character has its own camera, lighting is done on a per animation frame basis... :-o

Regarding subsurface meshes, your model and hair. The subsurface modifier need to get applied before you use it in the game engine and yes... that is a mess...
The normal workflow is to bake the subsurface version of the model into a normal map to capture all the details.
Then remesh the detailed, pretty model with a more simplified mesh and apply the normal map to it.

This is how kings at 3D modeling can make a complete model with 10.000 triangles that looks like it has a few million triangles without a lack of detail.

Wonderful example from The Witcher

1. Actual model,
2. smoothed model
3. smoothed model + normal mapped
4. smoothed model + normal mapped + ambient occlusion map + diffuse map and maybe some shader magic etc.

 

Spoiler

Spoiler

First wire is fairly bad, though, mostly because it was one of first decent models I managed to create, and because I didn't get the hair volume right. First ref is also less realistic (see eye placement), which causes a lot of trouble when trying to model it.

Yes, I've looked into that, and I'm familiar with the colored loop image you posted later.

There's extra modifier applied to model in order to produce outlines. It effectively doubles polycount. The other thing is that subdivision level is set to high (Ctrl+2) because it makes it easier for me to spot problems with model's shape.

I'm familiar with contents of Xrd GDC talk, but I'm not exactly aiming at Guilty Gear. What I"d be interested is something similar to Sidonia's screenshot like the one Martin_H posted. I'm planning to use some of Guilty Gear's tricks when I get to them, though.

And the main purpose was to get a hang of modeling techniques, not to immediately plug the result into game engine as soon as I can. Basically, I had my own character which I tried to model first, then I ran into trouble with it, then I tried to use someone else's pictures to see if those will work better for practicing.

I've looked into sculpting, as far as I can tell, it does not work well for the visual style I'm interested in. The issue was getting shape right using unrealistic reference and dealing with stretchy polygons around hair strands.

 

For poly strip hair you make an image of a hair lock with the ends faded to alpha transparency. If your modeling app had the ability to extrude edges then pick edges on the head, extrude them to strips trying to keep a squareness to each poly in the strip, and then extract and separate them. Lay each individual strips polys in the UV layout to overlay the single hair lock image. If you cannot extrude edges then just create the strips and "plant" them.

For the head..box model it. Start with a cube and slice it along the main anatomical axis..vertical face center, horizontal eye level, nose and mouth. A few more slices along the eye outer edge plane, inner eye edge plane, center eye on each side. This should give you enough polys to start pulling about the get a basic shape. Select the eye rects and extrude inner a few times to get the eye lids and center eyeball polys for manipulation. Do the same for the mouth and again for each nostril..if yer cartoon character has them. You now should have enough points to pull about the determine your shape to a significant degree. If you need more detail in any area then knife that area further, welding points that are not necessarily needed for either weighting a face rig to it or add nothing to the curvature of the head. Try to keep quads when you can as this makes the whole pushing points around much easier to follow. 3D apps often have a relaxer type tool that will round off squareness and try to set the vertices at an even set of local distances and angular changes. This can rapidly get the roundness from the box modelled structure.

 

That's an excellent method to approach such problems.

I would agree.

Thumbs up for Foodwars! I started watching it recently on crunchyroll and it's one of the best anime shows I've ever seen!

I see that you are using the crease edge feature in blender. That one is kind of non-standard and it has some potential downsides to use it. It can be very handy in some circumstances, allows nice control of edge-sharpness without blowing up the polycount or making the basemesh harder to edit, but you're also using a non-standard modelling technique that can't easily be transferred to other modelling applications and upwards of 95% of subsurf modelling tutorials out there will be using a different technique. They'll be using traditional edgeloops to control edge-sharpness and that kind of workflow also helps to force you towards a workflow with better quadbased topology. It gives you more control in some situations and the way creased edges behave in blender you might hit problems that you could have fixed with the traditional method.

Check this out:
http://polycount.com/discussion/56014/how-the-f-do-i-model-this-reply-for-help-with-specific-shapes

Sorry for just linking you a 130+ page thread, but I truly believe that some of the absolute best 3d modeling advice that exists, is in this thread. I'm also sure lots of questions keep being repeated in there. For a start I'd just skip ahead to the middle or so (partly because of linkrot on linked images) and read a few pages in the thread each day.

In blender afaik there is no good default tool for that, but you could try enabling the "Mesh:relax" addon, that is deactivated by default. I haven't used it though.

 

The problem is not how to model but how to get the shape right when ref is not "perspectively correct".
If you check out ref I previously posted, you'll see that in the side view eye is located very deeply within the head, which results in very strange head geometry, if you attempt to model it as is and match both views.

I also found out that first making edges that match character silhouette then extruding them and sewing them with quads works better than starting with a cube to me.

Hmm. To be honest, it didn't work well for me and I enjoyed titles like Ajin and Sidonia no Kishi much more.
However, this is one of the shows that had few character designs I really liked, which is why I grabbed one of the chars for practice.

Hmm. Dropping edge crease.... your arguments are good, and I'll think about it, but most likely I'll drop it at later date, when I get better hang of modeling in general.

Alright, I'll check it out. Already saw an interesting talk about pentagons being easy to subdivide here. I noticed something similar with triangles on rare occasions (when three triangle strips join at a triangle, sometimes it forms decent quad-based geometry after subdivision).

You can "sorta" relax geometry by pulling random vertices with proportional edit, or by applying sculpting "smooth" tool without dyntopo enabled. Both are suboptimal and in general in blender you jsut try to never end up in situation where you need to relax anything.

 

Alright, so I checked online for sidonia/ajin wireframes, there are none. At the very least they don't seem to be reachable via english language search.

However, somebody in this thread mentioned sketchfab earlier, and as far as I can tell it is very useful as a learning resource for this kind of models. So I guess I'll be looking through that, for now.

 

Why don't you just use this for reference?
https://www.assetstore.unity3d.com/en/#!/content/18705

 

You can use this free app from the ZBrush folks to sculpt and paint with. I am not sure why this isn't mentioned more.
Sculptris

 

To be honest, I forgot about it, and another reason is probaby because unity-chan has "painted-on" face and I was looking for something that has geometry for eyes.

I"m aware of it, and I might give it a try... it is just I already tried sculpting something similar in blender, and while it is fun, for this kind of style (smooth surfaces, almost no micro details):

Spoiler

Subdivision surfaces seem to be working better. Of course, It is possible that I just don't know a good sculpting workflow.


Anyway, I do not think I have any more questions regarding this at the moment. Just need to find more time to finish the last model....

 

Sculpt your form as it is like working with clay->use the low poly brush to get the tris down and create a smooth surface->use that model as a base for retopologizing it properly with the appropriate edge loops to accommodate animation with proper vertex weighting. Zbrush is actually the ideal tool for this work but Sculptris is free and some of it's tech found it's way into ZBrush such as dynamic tesselation. Both spit out nice UV maps with good coherence between the polys in 3D versus the polys on the UV map. ZBrush has great retopo tools. So does 3D Coat which you can probably use the demo of.

The issue with SubD is that you have four times as many polys than the cage in the first level of subdivision. Look great in 3D apps renders and viewport as a smoothed model and will subdivide stretched poly when joints are animated and kind of tame loose vertices..but.It does not work inside current game engines..so if you want fidelity then just model using the basic polygon surface .

 

Looks like everybody has some really good advice. Ill just add most modeling topics Ive helped with were listed in external tools section. But this is as good a place as any.

One suggestion on workflow havent seen mentioned (probably because it isnt as popular) is to extrude polygons/edges along a spline.
Some people find editing splines in 3D easier I guess - so they create splines for each hair segment they want to create. Then extrude the polygon/edge along the spline. A little vague with specifics because each software likely performs this technique slightly different.

Also I think its important to define the 'part' in the hair and extrude polygon planes from this area.
To be honest - I have not seen many 3D examples of extruded polys coming off the edges of a helmet like hair mesh - although Im not really into anime so this may be a technique used in that style.

 

Thanks for the pictures, things are making a lot more sense now

Lets concentrate on the face first, it is good that you know of the different face loops and mesh flow. These concepts aren't laws by any means and don't apply to extreme low-poly models. They simply make it easier to achieve a natural, organic shape on medium- to high-poly models. They become really important on the other hand when you want to animate a model because they have a habit of stopping different animations from affecting unwanted areas and sometimes even help the system to bend the mesh in a natural way. They also minimize the chance of odd artifacts to appear in the mesh during animation.
As you might have noticed online, many artists have developed variations on the core concept that better suit their own needs and purposes. Don't be afraid of experimenting and even the best artists get stuck sometimes with odd cases and have to thrown in a triangle among the quads or make a pole vertex (this is when a vertex connect to more then four other vertexes, considered bad because it can have bad effects on animation but poles are also a good way to direct the flow of the mesh, use with care).

Enough theory and more fun with modelling, here are some suggestions and why:

I would begin with adding a loop perfectly follow the jawbone and pass just in front of the ear. The rest is personal taste, some combine it with the lopp I will mention next, some make it curve around the ear and others allow it to pass over the ear and to the back of the head to give the skull more shape. - The main purpose of this loop is simply to give the face some of its most important shape lines. It is also a lot easier to change the features of the face at a later point when you got a fixed loop that follows the jawbone. Some tweaking and the face will go from a triangular anime face to a more european / american rectangular face.

The other loop follow the the forehead, pass behind the cheek bone and typically go below the jaw to better define the area where the jaw meets the neck.

An example where both have been combined into one can be seen below. Some people skip the area below the jaw, especially if the area where the neck meets the jaw is less defined on the character.



I would personally add a loop of quads along the nose ridge to give the nose more volume. Anime characters actually have more volume to the nose then you think, this is the reason why the whole nose seems to be a bit "off" in different 3/4 and side views. You can actually notice this effect in your reference screenshot. That guy, his nose is a bit to far back, follow the lines and try to imagine the nose volume and you will realize that it all makes sense if the line-work actually follows the distant side nose ridge, the closest edge of the nose edge is hidden by the strong, cartoonish lighting.
You could actually allow the loop to continue along the whole head so that you get more geometry to play with.

This brings me to my final suggestion, I would personally add more loops and vertexes to the head in general. The area beneath the eye and on the side of mouth isn't really correct and this is probably because you don't have enough geometry to work with. More geometry also give you more points that you can more to shape the head with. I'd say twice or thrice the amount you have now is a good goal

Still, good work if this is one of your first decent models. I can see that you are getting there.

Ok, polycount is still a bit to high but it looks like it is the work of the subdivision modifier then. Are you using the mirror modifier?

The basics aren't really that different but Guilty Gear is more of a dream scenario for artists. The faces are definitely the same though and that was the important case for us at the moment. GG just shows what is possible if you have highly detailed face mesh that makes good use of the loops and flows.

The bodies are on the other hand a bit different. Sidonia seems to use a regular (aka even, balanced) flow to the mesh while we know from seeing the GG models that they use an irregular flow that follow the shape of the muscles. GG has to do it that way because of their trick with the textures.

Take your time and brake a few eggs, just modelling the mesh is an art in itself. That is why many large studios have artists that specialize in concept, sculpting, retopology, texturing etc... All of us have been there in the past xD

Both sculpting and traditional modelling can get the same job done. Some people will tell you to go this or that path but it is in reality more like picking a NVIDIA or AMD graphics card. Its just different ways to approach the same problem and the truth is probably a bit personal and somewhere in the middle. Lots of AAA professionals have many different programs installed and use a specific set for specific tasks.

Find your own way and you will be fine, it should be easy and fun to get the work done.

 

Might be relevant:

 

Ha. I just saw this today. I don't quite agree with the author though.... It is more like there are several studious researching CG direction and they haven't found the right approach yet.

And speaking of "ruining"...

Didn't disney pretty much close their 2d animation department?