How do I know what should be normal mapped/baked in an environment?

CognizanCe

I want to create a near future sci-fi city environment.

I assume walls, side walks, roads, and anything "blocky" won't require a normal map baked from a high res mesh.

But what about street lights? Fire hydrants? guard rails? manholes? display kiosks?
This may sound like splitting hairs, but doesn't EVERYTHING look better with normals baked from a high res mesh?

I just see making so many high res meshes, especially if there's a lot of pieces/props, retopologizing and texturing them would take so long it would be maddening, especially modular assets (guard rails, stairs etc.) which are the same them only at different sizes, lengths, heights etc.

Am I just overthinking this? Should I create as much of the environment that I can get away with just flat?
If so, what should I bake?

Here are some examples of the look I'm aiming for:

huffer

You can bake everything if you want to, but you can also keep baking to absolute minimum (close to zero). Take a look at the environments in Alien Isolation or Ruiner for examples.

Large modular pieces wouldn't be baked (like you said, walls, roads, etc), but made with higher res geometry (chamfers) and tileable textures - they won't look blocky at all, maybe even better than if you were to bake them (because you're not limited by texture size, everything is sharp, no artifacts, etc).

But you can keep props non-baked too - make use of mix masks, tiles and decals. If a piece would require too much geometry - then you can bake it instead. Or if there are a ton of tiny details. Guard rails or stairs should use tileable textures (horizontally or both axes) baked from high res elements. You can keep a modular approach for these by planning where to cut and tile, make some caps, end elements, etc.

CrackRockSteady

What you're asking is dependent on a lot of factors.  What type of game is the scene intended for, what's your target platform, how much time do you have, what in the scene is most important as far as visual fidelity, etc.

It's something that you will need to learn to make judgement calls on.  Even if this scene is intended to be a portfolio piece only and you can afford to bake unique high res maps for every asset in the scene, like you said, doing the whole process on every single prop will take a lot of time and you need to decide if you really want to spend your time baking every little asset in the scene.

JordanN

CognizanCe said:

I want to create a near future sci-fi city environment.

I assume walls, side walks, roads, and anything "blocky" won't require a normal map baked from a high res mesh.

But what about street lights? Fire hydrants? guard rails? manholes? display kiosks?
This may sound like splitting hairs, but doesn't EVERYTHING look better with normals baked from a high res mesh?

I just see making so many high res meshes, especially if there's a lot of pieces/props, retopologizing and texturing them would take so long it would be maddening, especially modular assets (guard rails, stairs etc.) which are the same them only at different sizes, lengths, heights etc.

Am I just overthinking this? Should I create as much of the environment that I can get away with just flat?
If so, what should I bake?

It's easier when you think of a normal map as bump map detail. 

It was always intended to fake lighting detail that would otherwise require millions of polygons to do. Since games don't have the same luxury as Hollywood Movies to throw around millions of polygons in real time, normal maps are your best bet for giving your assets high quality surface detail. Otherwise, even if you you end up using other PBR maps like roughness and diffuse, your props will still have that "ultra shiny toy look" instead of the more true to life "uneven and bumpy detail" found on every surface.



Above, the sphere has the same polygon count. But the bump map fakes it so you achieve higher frequency detail without actually increasing the geometry.

And you mentioned ordinary props like Fire Hydrants. Yeah, that stuff requires bump/normal maps too. Go up close and you'll observe all the surface is not the same. Paint is chipping off or the metal is exposed. Both effect how bumpy the overall material is.

CrackRockSteady

@JordanN he was asking specifically about when to bake a normal map from a high poly model vs using tiling textures, not "should i use a normal map", it seems like he has a solid grasp on what a normal map is already.

JordanN

CrackRockSteady said:

@JordanN he was asking specifically about when to bake a normal map from a high poly model vs using tiling textures, not "should i use a normal map", it seems like he has a solid grasp on what a normal map is already.

The tiling textures threw me off.

I always assumed you would still use a high poly even if you wanted a tiling surface like a wall. It just seems more accurate in practice, unless the goal is to create an entire environment procedurally. 

Obscura

musashidan

JordanN said:

The tiling textures threw me off.I always assumed you would still use a high poly even if you wanted a tiling surface like a wall. It just seems more accurate in practice, unless the goal is to create an entire environment procedurally. 

That's one of the main reasons tiling textures are used. To augment the limitations of high/low baked assets. Large environment pieces can't be baked outright as there isn't enough texel res in the 0-1(unless broken into modular kits) This is why there have been many innovative techniques developed:

FWVN

Vertex paint lerping

Trim textures and decal sheets

Mesh decals and alpha mapped texture decals

Mid-poly assets

Multiple UV channels

@obscura edge wear shader

The cost of creating an entire environment - in both time and normal map textures - is significant if everything was a unique asset baked from a high poly.

JordanN

musashidan said:

The cost of creating an entire environment - in both time and normal map textures - is significant if everything was a unique asset baked from a high poly.

Well I learned something new today. I thought high poly models took preference over the use of procedural generated ones in AAA production.

I had been developing an environment that was 95% hand sculpted, and the other 5% were procedurally driven. I'll look into changing this in the future.

Alex_J

Judgement call. 

As a general rule, a workflow that allows you the greatest ability to make changes quickly and easily is good. You may make a really nice hydrant by sculpting in zbrush, but you can make 50 unique hydrants in half the time with a few procedural materials and mask that might still hit your target visual quality. I think its important to practice both the high poly sculpting/baking workflow and other means so that you can understand the pro's and con's of both methods. Just different tools in the toolbelt.

How much of a given environment contains customized, sculpted assets versus assets relying on other less time consuming means for detail probably depends on how many artist are available to work on the environment, what the target visual aesthetic is, how much money and time there is, etc. I'd expect being able to do the most with the least is a good skill to have, aside from just being able to sculpt beautiful masterpieces.

JordanN

BIGTIMEMASTER said:

Judgement call. 

As a general rule, a workflow that allows you the greatest ability to make changes quickly and easily is good. You may make a really nice hydrant by sculpting in zbrush, but you can make 50 unique hydrants in half the time with a few procedural materials and mask that might still hit your target visual quality. I think its important to practice both the high poly sculpting/baking workflow and other means so that you can understand the pro's and con's of both methods. Just different tools in the toolbelt.

This is true.
I'm just really caught up with attention to detail that even spending a lot of time hand sculpting a single asset to match a reference completely  just feels more worth it and a lot of fun, than generating a bunch of masks and presets and customizing as you go along.

But I can see this type of method might prove too expensive in the workplace so I'll keep this level of detail for personal projects.

musashidan

@JordanN Well let's take your environment as an example. 95% hand-sculpted. A lot of assets that need to be individually sculpted, high/low created, baked, and I assume you're texturing them all uniquely? Destructive workflow. Locked into your bakes. Pain in the arse to resculpt/rebake. That's a lot of work and a LOT of texture memory.

So if you created it using the mesh decal technique, for instance, you could break it down like this:

Sculpt/create a single 2K texture with all of your edge details/cracks/chips/variety of small/mid/large damage decals. This would be your main texture that would be used across all your assets.

Create a tiling RGB packed stone texture. This would also be used across all your assets

Create a mesh decal set to match your main NM texture. This could consist of generic pieces and a few 'hero' pieces for variation.

Assemble pieces together in your DCC. Make use of 2 UV channels.

Optional step: create a second tiling packed texture(moss/different stone/dirt/etc) and also create some 'hero' stones with extra tessellation to be used for vertex painting with this second texture.

So, using this method you are only using 3 texture maps in total, for the entire environment. You will save massively on the texture footprint and definitely save time on asset creation. Plus this is very flexible as you can chop/change/iterate all the way through the project as you are not tied into any uniquely sculpted or baked assets.

Then when you have all this work done you could add in some unique baked assets for even more variation.

Obscura

Time is money and honestly, noone cares about some unimportant prop or thats just there to fill up the space, or some generic tile floor so hand sculpting paint chips or dirt on it would be very pointless and waste of time. Also, have you seen Substance Designer and Painter, or the Quixel apps? There are much more time - efficient ways of doing thing these days, while they can give the same quality as a hand sculpted one, plus the free variations. And its not only that but imagine if all the thousands of assets in a game would have a unique normal map both made from a highpoly and stored uniquely on acceptable resolution. Video memory would be eaten up so quickly you could only fit a few rooms in into a consumer grade card. 

Sculpt organic assets, and characters, it makes sense to have a unique normal map for these, but shader tricks are still used very often. A huge rock for example. Instead of having all unique textures, you have one baked normal and multiple tiling textures.

Anything hard surface can be done using the techniques mentioned above, and in a bigger scope, its usually cheaper to have them like that because you cut down a huge amount of video memory by using those tiling textures, masks, more polygons instead of everything baked, etc. There might be some examples where modeling the most of the things out would be too highpoly but you can always use a good mixture.

Setting these up can still take a lot of time but the reusability compensates.

danr

Hang on a second ... before someone else starts using the word like the actual meaning isn’t important... 

Generally speaking in environment art, the alternative to baking out a high poly asset is not “procedural”. As @musashidan pointed out in his (not exhaustive) list of techniques, these all require a deft and highly manual touch. “A bunch of masks and presets” is a generalisation so weak as to be completely irrelevant at best, offensive at worst. 

Of course there are ways of procedurally generating environments, where the procedural nature is at the very core. None of these are being discussed here

Alex_J

One of my favorite games of late is theHunter : Call of the Wild. Amazingly beautiful natural environments, and very well optimized. Almost anywhere you go in the any of the three large open world environments, the beauty if like, post-card level. And yet, if you look at any individual prop up close, its really low texture quality. I have no idea really, but I expect they started with a lot of scanned assets and just took the texture res down and chopped the poly counts down to a level that allowed for dense, realistic forest and a consistent framerate. 

Anyway, the point is, all of the individual enviro props work together to serve a purpose. Individually they are pretty low quality, but altogether they make a masterpiece. Average player isn't sitting there observing single trees. They see a scene as a whole. 

Alex_J

danr said:

Generally speaking in environment art, the alternative to baking out a high poly asset is not “procedural”...these all require a deft and highly manual touch. “A bunch of masks and presets” is a generalisation so weak as to be completely irrelevant at best, offensive at worst. 

Hopefully nobody was offended by my poor classification. 

Ok, you've said what is wrong, but you didn't say what is right. What, if anything, do you call the common, modern methods of modeling and texturing that do not involve baking from sculpts? 

CrackRockSteady

danr said:

Hang on a second ... before someone else starts using the word like the actual meaning isn’t important... 

Generally speaking in environment art, the alternative to baking out a high poly asset is not “procedural”. As @musashidan pointed out in his (not exhaustive) list of techniques, these all require a deft and highly manual touch. “A bunch of masks and presets” is a generalisation so weak as to be completely irrelevant at best, offensive at worst. 

Of course there are ways of procedurally generating environments, where the procedural nature is at the very core. None of these are being discussed here

musashidan said:

JordanN said:

The tiling textures threw me off.I always assumed you would still use a high poly even if you wanted a tiling surface like a wall. It just seems more accurate in practice, unless the goal is to create an entire environment procedurally. 

That's one of the main reasons tiling textures are used. To augment the limitations of high/low baked assets. Large environment pieces can't be baked outright as there isn't enough texel res in the 0-1(unless broken into modular kits) This is why there have been many innovative techniques developed:

FWVN

Vertex paint lerping

Trim textures and decal sheets

Mesh decals and alpha mapped texture decals

Mid-poly assets

Multiple UV channels

@obscura edge wear shader

The cost of creating an entire environment - in both time and normal map textures - is significant if everything was a unique asset baked from a high poly.

Alex_J

Ah, ok. So there isn't really a good single term to categorize the myriad techniques that exist beyond sculpting/baking. 

danr

you actually clipped out the answer to your question while quoting me? Man, that's weird

Alex_J

Well, I didn't think it was exactly answering what I was getting at, but it's really a matter of semantics and I didn't think it worth pressing and derailing the topic, so I conceded when crackrocksteady reiterated your same point. I don't see what being sarcastic adds to anything.

Eric Chadwick

What exactly were you asking? Maybe we can provide some clarity.

musashidan

@BIGTIMEMASTER if what you are asking is 'do all the techniques I listed above have an umbrella term?', then no. They are just methods that have been developed over the years to overcome real-time engine limitations.

It might be just a case of saying uniquely baked assets Vs non-uniquely baked assets.

poopipe

^^ I've always viewed it as reusable props vs unique level geometry  

Props are usually high detail/loddable&reusable,  unique geometry is usually none of those things. 

You'll tend to do unique bakes for the props because you get a lot of benefit for the time/resource investment (they're scattered all over the place and you only pay once) 

Unique geometry will tend to use tileable materials because  if the geometry is used only once, the only way you can reuse resources is in textures. 

Alex_J

Eric Chadwick said:

What exactly were you asking? Maybe we can provide some clarity.

Mushashidan got it with the next reply. Basically I used poor terminology in ignorance, and wanted a better alternative. 

CognizanCe

CrackRockSteady said:

It's something that you will need to learn to make judgement calls on.  Even if this scene is intended to be a portfolio piece only and you can afford to bake unique high res maps for every asset in the scene, like you said, doing the whole process on every single prop will take a lot of time and you need to decide if you really want to spend your time baking every little asset in the scene.

BIGTIMEMASTER said:

Judgement call. 

As a general rule, a workflow that allows you the greatest ability to make changes quickly and easily is good. 

Obscura said:

Time is money and honestly, noone cares about some unimportant prop or thats just there to fill up the space, or some generic tile floor so hand sculpting paint chips or dirt on it would be very pointless and waste of time. 

Sculpt organic assets, and characters, it makes sense to have a unique normal map for these, but shader tricks are still used very often.

Anything hard surface can be done using the techniques mentioned above, and in a bigger scope, its usually cheaper to have them like that because you cut down a huge amount of video memory by using those tiling textures, masks, more polygons instead of everything baked, etc. There might be some examples where modeling the most of the things out would be too highpoly but you can always use a good mixture.

Setting these up can still take a lot of time but the reusability compensates.

BIGTIMEMASTER said:

Anyway, the point is, all of the individual enviro props work together to serve a purpose. Individually they are pretty low quality, but altogether they make a masterpiece. Average player isn't sitting there observing single trees. They see a scene as a whole. 

Thanks a lot guys, you've been most helpful and have given me some real world perspective!