My First Game: How many UV layouts should I create?

[Jesse Davis]

I’m in the process of creating my own video game using the skills I’ve learned from CGCookie. I have no formal education in video game development. Please excuse my lack of knowledge, everything I know involving video game development I’ve picked up from CGCookie, Youtube or some random corner of the internet. 

I imagine this will be the first of many questions. 

I know how to UV unwrap models and lay them out in the UV editor. I’m conscious to conserve space and pack the UV’s as tightly as possible while considering pixel density. 

What I’d like to know (on average) is how many maps should I layout? I would imagine most video games need to use more than one UV map for their game assets. For a one-man dev team, how many UV maps should I use? Roughly… ballpark? 

Thanks

-Jesse

[Jesse Davis]

Did my question make sense? 

[bitbuilder]

Yeah it doesn’t make sense, you want untextured  uncolored game assets?  why?

 At the very least there should be one UV map for each model,  not every object is going to need baked in normals or bump/height maps, Are you planning to have the shader material be a replacement? even then that requires unwraping and I’m not sure how easily that transfers.

[Jesse Davis]

My apologies for the confusion. Of course, I want to have colored and textured game assets. 

I have modeled every item in the game. Now, I’m beginning the process of unwrapping the models and laying them out in UV space. I’m wondering how liberal or conservative should I be with UV space? What would be considered too many UV maps? 

[bitbuilder]

The bigger the area on the uv the more pixels and thus detail are applied, unless your messing around with Udim but that a different fish entirely. 

Application again is key, what parts need to be viewed up close if any? is this going for realism? what are the requirements of the device?   

The shader can replicate what it will look like. but then again there is a wide gap in application, a top down mobile strategy game is going to rely more on simple shapes and colors viewed at a distance while one of those find the x puzzle/adventure games are all about up close intricate set pieces.

[Jesse Davis]

The game I’m developing is a First-Person puzzle/adventure game with a heavy emphasis on manipulating small objects. Therefore, I want to have as much detail as I can. This game is being designed to run on a PC. Being my first game I didn’t want to have to deal with the limitations of a console or mobile. 

I knew when I asked this question that the response was likely to be “it depends”. And for me, this process will include a lot of “try it out, see if it works, if it doesn’t, figure out why and adjust”. 

Thank you for the advice. 

[Jesse Davis]

Would it be more accurate to ask “how many textures should my game have?”

[Tomas Plasil]

I dont think there is a definitive answer for that, its more the less the better without sacrificing visual quality.  So if you have a larger model, it can have one texture set and if you have a bunch of tiny models you should create one texture atlas for them instead of giving each model an individual texture set.

Ideally models should have consistent texel density based on how far from the player they are. Objects that are viewed really up close need more texel density than objects in the background. Check out this for more detail: https://www.artstation.com/artwork/qbOqP 

[Jesse Davis]

Tomas Plasil. 

Thank you so much for that link! I immediately downloaded the PDF. It was quite informative! 

[silentheart00]

There are many different ways you can approach textures for assets based on different criteria.  Not sure what kind of art style you’re going for or what assets you need, but there are some different techniques you can look into and see if that works with what you’re trying to achieve:

• polycount wiki has plenty of game dev techniques. Highly suggest going through and reading up if you’re not familiar with the website.
• MultiSubObject Materials are great if you have a lot of meshes that use procedural materials; can save texture space this way by just having 1 texture set, say, for a brick material, but that texture can be used across a lot of “building” assets without each having their own brick texture set. This also allows you to create “material slots” that you can then plug in procedural/tiling textures. I hope that makes sense. I can pull up a more visual example if needed.
• An amazing post about how Star Citizen gets such crisp detail on their ships without using crazy high texture sizes.
• Another thing to add to your arsenal is using detail maps (this is the name I use for this particular map, it may be called something else), which can save texture size, as well.  What I mean by that is your greyscale maps (so Roughness, Metalness, AO as the standard, but any greyscale texture can be stuck in a detail map) can be stored in the RGB channels of a texture, so instead of 3 separate 24 bit textures, you have one 24 bit texture with 3 maps in it.  Then you plug in the corresponding channel for what you’re looking for.  Each channel has varying degrees of “detail afforded,” so if you need something with the best detail, stick it in the Green channel, then Red, then Blue is the worst.  I hope that makes sense.  So, using this technique, you have three 24 bit maps per asset instead of five 24 bit textures.
• You can combine varying sizes of maps if you need to.  For example, you have your textures, but only 1 has some small detail, like an emissive map.  If it’s only a BW texture, you can totally scale that down to something as small as 128 if you want to.  There should be minimal loss of visual impact, but this is something to experiment with.  Or heck, you don’t have to resize it; you can chuck it into the DET map in the Alpha channel, increasing that map by only 8 bits.  Make sure to save it with alpha.
• Vertex painting/colors is also something to look into. You can totally use a DET map with BW textures other than Roughness, Metalness, AO, such as some sort of grime texture instead in each channel to drive how the vertex painting is masked out. It can save on mesh density while still providing visual density.

There are definite costs associated with each technique here and as yet to be discovered, so it’s about balancing each cost against if you need it or not.  Why spend the resources on something the player will never get up close and personal to when those resources can be used to make something the player will see really, really good?

I hope this is a good starting point for you.  Good luck!

[Aisnley Hayes]

Hi. I’m just starting out too and can’t seem to find the answers to my questions, please help. What factors should be considered when choosing between high and low resolution UV unwrapping for game models to ensure optimal performance and graphics quality in the final video game?

[Martin Bergwerf]

Hi Ainsley @ainsleyhayes ,

As mentioned above: it depends.

Performance and graphics quality are more or less opposites. You’ll need to find a balance.

But apart from that, UV Unwrapping is usually done on Low-Poly models…I don’t know what you mean by “choosing between high and low resolution UV unwrapping”.

[Aisnley Hayes]

@Martin  understand, thank you…

[Adrian Bellworthy]

@ainsleyhayes when modeling for games, a high poly model is created first with all the detail.
Then a low poly model is created from a duplicate of the high poly model, by removing vertices without effecting the silhouette of the model.

It will take lots of practice and experience, check out TREAD: Hard Surface Asset Creation for Video Games and The Game Asset Creation Playlist

[Aisnley Hayes]

@Adrian Great, I’ve already started watching based on your recommendation… useful.

[Author]

Posts