3D models Help (Blender)

[Yuki/MidnightHowl]

Me and a friend are planning to start a turn-based rpg later this year, I'm doing the art and we decided to do it in 3D. Here my questions:
1- Somebody got tutorials for furry/anthro models? I found some on youtube but most of the time they are low quality or not a tutorial at all like RoseRedTiger vids
The one i use for the head below here was this

You must be registered to see the links

, but Keliff never do a tutorial on body so xd
2- What are the Poligons/faces/etc common for in game models? (you know for not get extreme lag)
3- We actually decide to do this a mix of furry/anime world, Like https://f95zone.to/threads/erolon-dungeon-bound-v0-15-alpha-sex-curse-studio.21837/ world, I'm going to look for anime style model tutorials on my own, but if you know any good ones and want to leave it here I'll be grateful.

I attach some models i made 2 years ago. Thanks in advance <3.

 

[MidnightArrow]

#2 - Depends 100% on the genre, platform, and how many characters will be onscreen at once.

You must be registered to see the links

PS3 era had about 15,000 - 20,000 triangles per character. If it's a desktop game and you're not doing Dynasty Warriors fights you can probably go a little over that now. Needless to say if it's a mobile game, divide that by ten.

 

[Yuki/MidnightHowl]

#2 - Depends 100% on the genre, platform, and how many characters will be onscreen at once.

You must be registered to see the links

PS3 era had about 15,000 - 20,000 triangles per character. If it's a desktop game and you're not doing Dynasty Warriors fights you can probably go a little over that now. Needless to say if it's a mobile game, divide that by ten.

TY so much, this was my main concern, we probably use unity.

 

[Tompte]

Modern graphic cards are pretty good at crunching triangles in the millions (but that doesn't mean that you should.) By "modern" I mean anything Nvidia 1080-series and up.

The answer to #2 is as few triangles as you can get away with for the level of detail you're after. Say you have a model with 100,000 triangles. If you can reduce that to 50,000 without any visual difference, then that is clearly better for performance. If you can reduce it to 25,000 that's better still. If you can do it in just 1,000 that is probably for the best. It's not a matter of asking "what is the magic number?" because the answer will depend on how many objects you want to draw in a frame (60fps = 16ms) and your target machine. That said, the number of triangles you can put in a model is probably higher than you think because triangle count isn't that big of a bottleneck these days.

Honestly, you don't really need an obscene amount of triangles to describe an organic shape like a game character, or fox in your case. Whatever you're doing is probably correct already. If you need a higher level of detail, you can use a normal map to add the fine details in and the player won't be able to tell it's not a super-high poly model.

There are other bottlenecks to consider, like batching and number of draw calls. For instance, you could probably draw a million triangles no problem if they all share the same material and can be rendered with a single draw call. But if those triangles were divvied up into 100,000 individual low-poly instances, each with their own individual material and thus separate draw calls, then that would perform much worse. Both Unity and Unreal (I presume) batches automatically and will tell you the draw call count in the performance metrics.

Also keep in mind that post processing effects can come at a very high cost because they will incur several render passes per frame and full screen pixel-by-pixel operations. For instance, for DOF the game would need to render a depth pass. For shadows it renders one shadow map per shadow casting light. SSAO will require its own pass(es). Any blur effect will sample a texture several times per pixel. The more effects and render passes you stack up, the more those triangles will matter because you're not just rendering them once. As soon as all of this takes longer than 16ms, your fps will drop below 60.

That being said, I think if your game is not really that visually intense, you have work really hard to make your game run slow on modern machines, or do something really stupid. Don't feel the need to optimize until there's actually a performance problem. But it is good to understand how and why things can get slow, if you care about this stuff, and tweak your game accordingly.

(This was probably overkill for what you were asking but I'm hoping other people could learn something from this as well.)

 

[Yuki/MidnightHowl]

Modern graphic cards are pretty good at crunching triangles in the millions (but that doesn't mean that you should.) By "modern" I mean anything Nvidia 1080-series and up.

The answer to #2 is as few triangles as you can get away with for the level of detail you're after. Say you have a model with 100,000 triangles. If you can reduce that to 50,000 without any visual difference, then that is clearly better for performance. If you can reduce it to 25,000 that's better still. If you can do it in just 1,000 that is probably for the best. It's not a matter of asking "what is the magic number?" because the answer will depend on how many objects you want to draw in a frame (60fps = 16ms) and your target machine. That said, the number of triangles you can put in a model is probably higher than you think because triangle count isn't that big of a bottleneck these days.

Honestly, you don't really need an obscene amount of triangles to describe an organic shape like a game character, or fox in your case. Whatever you're doing is probably correct already. If you need a higher level of detail, you can use a normal map to add the fine details in and the player won't be able to tell it's not a super-high poly model.

There are other bottlenecks to consider, like batching and number of draw calls. For instance, you could probably draw a million triangles no problem if they all share the same material and can be rendered with a single draw call. But if those triangles were divvied up into 100,000 individual low-poly instances, each with their own individual material and thus separate draw calls, then that would perform much worse. Both Unity and Unreal (I presume) batches automatically and will tell you the draw call count in the performance metrics.

Also keep in mind that post processing effects can come at a very high cost because they will incur several render passes per frame and full screen pixel-by-pixel operations. For instance, for DOF the game would need to render a depth pass. For shadows it renders one shadow map per shadow casting light. SSAO will require its own pass(es). Any blur effect will sample a texture several times per pixel. The more effects and render passes you stack up, the more those triangles will matter because you're not just rendering them once. As soon as all of this takes longer than 16ms, your fps will drop below 60.

That being said, I think if your game is not really that visually intense, you have work really hard to make your game run slow on modern machines, or do something really stupid. Don't feel the need to optimize until there's actually a performance problem. But it is good to understand how and why things can get slow, if you care about this stuff, and tweak your game accordingly.

(This was probably overkill for what you were asking but I'm hoping other people could learn something from this as well.)

Not overkill at all, it really helps.

 

[shark_inna_hat]

I agree with Tompte. Mostly. The number of batches (draw calls, individual objects, meshes, call it what you will) will have a much, much, much, much bigger impact on the performance than the polygon count. Keep the number of meshes on screen under 200 and you'll be golden.

Also keep in mind that post processing effects can come at a very high cost because they will incur several render passes per frame and full screen pixel-by-pixel operations. For instance, for DOF the game would need to render a depth pass. For shadows it renders one shadow map per shadow casting light. SSAO will require its own pass(es). Any blur effect will sample a texture several times per pixel. The more effects and render passes you stack up, the more those triangles will matter because you're not just rendering them once. As soon as all of this takes longer than 16ms, your fps will drop below 60.

That is not entirely true. You need to render the scene (a depth map) from the perspective of a light for it to cast shadows but, the scene is usually not rendered multiple times for effects like AO, DoF, AA, SSLR, etc. and even if those effect need a depth/normal/world pos/velocity map it can be done efficiently simply using multiple render targets (the geometry is processed once and the fragment/pixel shaders output to multiple texture buffers) - that's why these are often referred to as post processing effects or screen space effects. The geometric complexity of the scene has no impact on how fast these effect run (even with if causes, the shaders regularly execute both branches because how wavefront/warp SIMT? architecture works). These effect can and will have an inpact on the framerate, but not because of how many triangles are there to draw.

One more thing to think about when optimizing meshes is animation.
If the mesh is animated you may want to keep it as quads and not decimate it too much and in fact you may want to add additional edge loops around joint/bones so the mesh keeps it's volume when you deform it... on the other hand - animating a mesh is hard work for the hardware and worst of all - it's often done on the CPU, so each of the 100 000 verts you got on your character need to be multiplied by the ~20 bone matrices before shipping it to the GPU, and that ain't free.
Optimizing the geometry of static meshes is usually wasted time (it might load faster, but it won't run faster), just slap some auto-generated Level-Of-Detail on it and it's good - for animated meshes it might actually have an impact.

 

[Tompte]

Thanks for clarifying.

 

[DyingStar87]

1- Somebody got tutorials for furry/anthro models? I found some on youtube but most of the time they are low quality or not a tutorial at all like RoseRedTiger vids
The one i use for the head below here was this

You must be registered to see the links

, but Keliff never do a tutorial on body so xd

Answer to your question is right in the video you linked. He's working from a concept art. If you created head model by following tutorial, you can use same/similar workflow to create other heads or body (but I would advise to do proper retopology instead of using auto-retopology tools, especially for characters as they need proper edge loops for proper deformation). So your workflow should look like this "visual research (gather reference images) ---> Create concept ---> create hi-poly mesh (if necessary) ---> low poly mesh ---> texturing ---> rigging". You can basically re-use any tutorial showing body modelling and apply same workflow on creating any other body type. Once you get base body mesh, you can simply modify it to fit any body type you want.

- Have a look at face and body edge loops, to get proper muscle and joints deformation on characters.

2- What are the Poligons/faces/etc common for in game models? (you know for not get extreme lag)

- Try not to overuse transparency/translucency. It is very expensive.
- Have a look at "channel packing" to limit texture calls.
- Have a look at occlusion culling.
- Merge meshes where possible before exporting from 3D Software to limit draw calls.
- Texture resolution. There's no need to have fancy 1024x1024 texture for eyes, if 99% of the time eye won't take more than 100x100 pixels on the screen.
- Trim sheets. Easy way to create rich, but performant environments using few textures.
- Try to think what objects need dynamic lighting, and what objects don't. Use shadow maps for static objects (if you are not planning to implement real-time day/night cycle)
- All meshes are triangulated in game engines, so watch triangle count instead of polycount (many 3D packages show quads amount as polycount). Having a mesh of 20 000 quads = 40 000 triangles.

Last small advice. Don't bother with creation of amazingly looking assets at this point. Just create a stickman or somthing and use it to set up prefabs and materials. You will replace it later for final assets. This way you'll get basic game functionality done much faster. Don't need to bother with graphics and visuals at the beginning of development.

 

[Yuki/MidnightHowl]

Answer to your question is right in the video you linked. He's working from a concept art. If you created head model by following tutorial, you can use same/similar workflow to create other heads or body (but I would advise to do proper retopology instead of using auto-retopology tools, especially for characters as they need proper edge loops for proper deformation). So your workflow should look like this "visual research (gather reference images) ---> Create concept ---> create hi-poly mesh (if necessary) ---> low poly mesh ---> texturing ---> rigging". You can basically re-use any tutorial showing body modelling and apply same workflow on creating any other body type. Once you get base body mesh, you can simply modify it to fit any body type you want.

- Have a look at face and body edge loops, to get proper muscle and joints deformation on characters.



- Try not to overuse transparency/translucency. It is very expensive.
- Have a look at "channel packing" to limit texture calls.
- Have a look at occlusion culling.
- Merge meshes where possible before exporting from 3D Software to limit draw calls.
- Texture resolution. There's no need to have fancy 1024x1024 texture for eyes, if 99% of the time eye won't take more than 100x100 pixels on the screen.
- Trim sheets. Easy way to create rich, but performant environments using few textures.
- Try to think what objects need dynamic lighting, and what objects don't. Use shadow maps for static objects (if you are not planning to implement real-time day/night cycle)
- All meshes are triangulated in game engines, so watch triangle count instead of polycount (many 3D packages show quads amount as polycount). Having a mesh of 20 000 quads = 40 000 triangles.

Last small advice. Don't bother with creation of amazingly looking assets at this point. Just create a stickman or somthing and use it to set up prefabs and materials. You will replace it later for final assets. This way you'll get basic game functionality done much faster. Don't need to bother with graphics and visuals at the beginning of development.

Thanks, yeah im struggling with the creation of characters, but its normal since i begin with 3dmodel now, most of my problems are with the body, most of the time it creates triangles or broke on some point and idk how to fix it, im following Shonzo "tutorials" but i got problems follow him, like in the body tutorial he use the clay tool in sculpt mode to inflate the parts but when i use in add or substract mode it deinflate the model no matter what, so yeah shit happens

 

[Yuki/MidnightHowl]

Some update:

 

[osanaiko]

Those arms remind me of something....

 

[Yuki/MidnightHowl]

Those arms remind me of something....

View attachment 2096049

XDDD, I haven't touched the arms yet, it's the last thing I'm going to touch of this model, since I'll do the hands, feet and head separately.