Hello
I am looking for some advice on current generation modelling and texturing for the XB1/PS4.
I thought I would start off with a vehicle
'LAV-3' an amphibious reconnaissance vehicle but feel free to give advice on characters, buildings and environments or any links to interesting articles like
Brian Burell’s Futuristic Russian Helicopter I would really appreciate it.
The vehicle below is one of my student’s files and the advice I am giving is very much pieced together from reading articles on Polycount and then practicing it but I do find my own approach a little varied and I am hoping for a more solid approach. I thought I would ask you guys how would you go about constructing it for current generation? let’s say for argument sake it’s a game loosely based on a game like ‘Kill Zone Shadow Fall / Call of Duty: Ghosts / Battlefield 4. So if you could give me a rough idea of Polycount, textures, materials, lighting.
Also are there particular softwares and pipelines you would recommend?
7th Generation Pipeline for Vehicles/Assets
1. Maya/Modo/3DSmax - (model into game engine)
2. Headus UV Layout/Road Kill
3. Zbrush/Mudbox
4. XNormal
5. Mari, dDo, NDo,
6. Photoshop
7. Clean up back in Maya/Modo/3DSmax (add Substance to the pipeline)
8. Into game engine (might add Substance?)
9. Back do forth for a bit reimporting updates
10. All the time thinking about Edgeloop flow and DirX 11 pipeline?
10.1. - Physically based shaders,
10.2. - Linear workflow/gamma correct workflows.
10.3. - Displacement maps if needed.
11. Bring it all into game engine (might add Substance again here?)
12. Back do forth for a bit reimporting updates.
Here is the of the turret, so if anyone wants to use that to show best practice then that would be great, I can then show this thread to my students and get them to construct files in a similar fashion.
Thanks again your help is much appreciated
Here is the OBJ file LINK to the image below.
Replies
Physical based shaders are in new Killzone, BF4 and COD - using same classic stuff.
You can look at [ame="http://www.youtube.com/watch?v=_29M8F-sRsU"]Digital Dragons - Micha? Drobot (Guerrilla Games) - Killzone: Shadow Fall - Lighting - YouTube[/ame] there is some explanations
Shadowfall and Ryse use approximately 40,000 triangles for non-hero characters,
compared to the ~20-40k for main characters in the last gen. So as a general rule, I'd estimate that triangle counts will hover around 2-3 times the triangle limits you would have had in a 7th gen game (max). But you should never exceed the amount of triangles you absolutely need to get the job done.
Increasing triangle counts may be the most expedient way to make a game look "next gen", but it also reduces an art team's rate of output, and only offers an incremental increase in visual fidelity.
If I were you, I'd keep the majority of my meshes at 7th gen acceptable triangle counts, if only to foster an appreciation for every triangle available to you. Triangle counts can always be augmented with displacement maps a some quality tessellation.
If you can reduce shader (material) complexity by using more triangles. Do so!.
If you can can reduce amount of textures used by increaseing tri amount. Do it!
If you can better normal map, by adding more triangles. Do so!.
If adding more triangles means, you have spend less time on texturing. Do it!
And so on. If you find yourself in situation where adding more triangles means saving time or saving other resource, just add them.
Times where. GPUs were sturggling with 1miln tris are long gone, Now they can render 1 bilion without problem.
If you can run GPU profiler on real-scene with lots of tris, I can guarantee you, that things like lighting, post-process will top the graph. Fallowed by particles. If you can you can also check how dramatic shader complexity decrease performance. The less pixel level masking/blending/other things you use the better. Fallowed by vertex masking. Multimaterials are way cheaper to use.
Wow... where did this number come from? It absolutely depends on your asset. If it's something very detailed, you might want to spend more than 20k, if it's something simple, no. Just saying 20k polys is quite random and doesn't help at all.
Also, the diminishing returns stuff depends on how you plan on using said asset as well. This comment reminds me of the people that believes this image:
You're assuming you have the same amount of detail and you just use more polygons for said mesh.
I hate that image, here's one talking about whats wrong with that image.
It's rather clear that simply subdividing will not add details, it will just mesh more smooth.
More polys. Most Xb1/PS4 gen games I'd imagine are using some kind of PB rendering (the ones simulating realism anyway). Mari and substance designer/dDo would probably be a good start for materials.
Marks - XB1/PS4
As we are not worried about polycount by the looks of what you say how would 80K for the LAV be and then make a level of detail down from there? I will also post up some material stats when I drop it into an engine.
Mari - Physically Based Shader for Real-Time Games
https://vimeo.com/65408722
[vv]65408722[/vv]
I was referring specifically to character art, in reference to a quote by someone at Guerrilla during an interview on KZ:SF, about how even though the poly counts had increased hugely, the corresponding increase in fidelity was only slight (I'll find the specific article later, and post a link). Though, I agree- I was far too universal in my prescription. It's my personal opinion (be it right, or wrong) that unless a character will be consuming a large amount of the screen space, or have a lot of secondary motion in its animations (clothes, flails, chains, etc), there's no real need to use more than 20-30,000 polys for its mesh unless its to correct normal baking artifacts, or increase the quality of critical elements of the silhouette.
But, leaving my opinions behind, here are actual numbers I've been able to find so far (I'll update as I encounter new numbers):
Marius (hero, Ryse) -> 85,000 polys
Source
Helghast soldier (Shadowfall): 40,000 polys?
(Source was on the Geurrilla site, but it's down right now so I can't confirm)
Currently emerging workflow as I understand it:
1. build a library of physically correct materials w/tiling textures
2. construct mesh as you would for any 7th gen title (UVs should be made with tiling materials in mind)
3. assign materials to geometry (per-poly, or with a mask), as well as baked maps (normal, bump, etc.).
Also MP vs SP player also has considerations to be made
I'd rather say. Pull your shit together and stop using DirectX..
This video should provide a more complete explanation:
[ame="http://www.youtube.com/watch?v=R7WNcUotwSQ"]Inside Unreal - Character Aesthetics - YouTube[/ame]
generally for something like this you'll have some material value index to check out to start your texture off with, but you'll still be tweaking those values, adding overlays, and applying unique details to finalize it. now that PBR is fashionable, it's becoming more common to get lookup tables for material values, but really this is something that sharper technical artists at studios have been doing for years, and isn't endemic to PBR.
tiling textures will remain primarily used by environment artists.
So, since there's no way you'll find access to Killzone files, I'd suggest scouting around the internet for communities who managed to extract BlackOps2, Ghosts, BF3 or BF4 files and follow the specs of these files exactly. These files will tell you all you need to know about how the assets were made, without the need for any educated guess.
In general, I don't think any realtime modeling assignment (with an emphasis on proper pipeline and tech) can really be given without at least a couples reference files of that nature. If such files don't exist for the latest cool game out there, then simply fall back to games/engines where such files are readily available for sure, like the stock assets of UDK and Cryengine. And when it comes to shaders and materials, if these engines don't support the material type you want to work with, simply use the Marmoset2 standards and "render" there.
Also I would advise you and your student to let the tank model sit for a while and figure all of the above first with an older, already made model (or any of the available SDK kits here on Polycount). That way you'll cut straight to the chase and figure out all the tech aspects first. There's nothing more frustrating than having to redo art because of an unforeseen technical limitation.
[edit]
Quick google search reveals that the information is definitely out there. Just dig around these forums, ask around, and so on - I think you'll get your hands on some files in no time !
http://facepunch.com/showthread.php?t=1320985&page=11&
That's kind of a sweeping blanket statement. And one that isn't necessarily true either, depending on where you work / what your art pipeline is like.
The implementation at Ready At Dawn uses a library of tiling materials and then each object has a unique unwrap with a mask texture to interpolate between materials on the object. Localized details like rust, corrosion, dirt, damage... Are all created by blending between parent materials. It is basically the same thing you would normally do in a PSD file by masking between layer sets. I think there are more local controls beyond that (map to affect color?) but I only did modeling for them and am not sure.
Well. Lets put it this way. To add unique details you need unique texture in some way. Either mask or diffuse.
Sometimes you need it sometimes you don't.
For clean models, you probably can do just with tileble textures. For more rough and dirty you would need some uniqueness.
Yes exactly.
What a genuinely surprising attitude from an ex-CDG'er
look in the mirror
someone is still salty about like 2007 or whatever, get over it and stop derailing threads
Good point, I have added it to the pipe line. I also know of lots of companies switching to Modo for UV creation as well as modelling.
7th Generation Pipeline for Vehicles/Assets
1. Maya/Modo/3DSmax - (model into game engine)
2. Headus UV Layout/Road Kill
3. Zbrush/Mudbox
4. XNormal
5. Mari, dDo, NDo,
6. Photoshop
7. Clean up back in Maya/Modo/3DSmax (add Substance to the pipeline)
8. Into game engine (might add Substance?)
9. Back do forth for a bit reimporting updates
10. All the time thinking about Edgeloop flow and DirX 11 pipeline?
10.1. - Physically based shaders,
10.2. - Linear workflow/gamma correct workflows.
10.3. - Displacement maps if needed.
11. Bring it all into game engine (might add Substance again here?)
12. Back do forth for a bit reimporting updates.
Here is the of the turret, so if anyone wants to use that to show best practice then that would be great, I can then show this thread to my students and get them to construct files in a similar fashion.
Advantages:
1. It allows you to group parts together in boxes when packing
2. Watching the uv islands relax with shift f and shift b is very relaxing and rewarding.
3. The cut tools, edge loops and being able to rotate around in 3d space.
4. Moving parts off to the side or hiding them in the edit window
5. Relax brush to smooth out small areas of uv islands
6. The segment tool that uses a face angle, particularyly good at squarish objects when cutting them up and laying them out flat in uv space
Disadvantages:
1. UVlayout is 32bit software, anything over 500k polygons in a obj file will cuase the software to crash a lot and run out of memory. That's a problem for me as I do a lot of high poly model unwraps not for games. So that means more maps that I use and breaking the model down to smaller chunks of obj's
Give it a try if you get the chance.