Home Technical Talk

Are normal maps always necessary?

Hi there,

I'm still a beginner, so I've got a beginners question.

Is it always necessary to bake a normal map when creating hard surface models for games?

I'll always bake a normal map for organic stuff, characters etc, and up until now I usually do on my hard surface models, but it seems to be slowing down my workflow a lot for the sake of minor details.

Lets say you have a piece of machinery where the only real details that would benefit from a NM are some rivets and bolts, is it worth making the high poly version, and setting the projection/cage just for what is essentially 5% of the finished piece?

Or is it acceptable (when I say acceptable I mean generally, in the games industry, or by your own workflow) to just bevel hard edges and create a normal map from the diffuse?

I know I haven't explained that very well, but going back to the rivets/bolts thing, I find I can create a reasonable NM from the diffuse which seems to give good results.

Many thanks

Replies

  • CompanionCube
    Options
    Offline / Send Message
    CompanionCube polycounter lvl 12
    you don't have to always make high poly models to create normal map. alot of normal maps are still created or added to by either painting greyscale images or using photos/images and converting them into normal maps using something like the Nvidia plugins for photoshop, or a program like crazy bump and even xnormal has a photoshop plugin.
  • Tom Ellis
    Options
    Offline / Send Message
    Thanks for the reply Companion,

    As I said in my OP, I have created normal maps by simply reworking the diffuse/spec and running them through Crazybump, just that as I'm still very new to creating game art, I wasn't sure if this was the wrong way to go.

    Thanks
  • Mimp
    Options
    Offline / Send Message
    Hello Creation22
    Is it always necessary to bake a normal map when creating hard surface models for games?

    No it's not necessary to bake the normal maps. Creating the rivets and bolts using an application such as CrazyBump or the NVidia filter in Photoshop is a perfectly good way of doing this.

    Something you should keep in mind is that creating rivets and bolts in geometry is not difficult, and should be considered. This is because creating normal maps from "real" geometry, gives you a better result, though it is slightly more work.

    The more you work on creating high poly assets, for normal mapping, the faster you will be able to do it, which is always a bonus.
  • EarthQuake
    Options
    Offline / Send Message
    I think you're confused in a sense, saying that the only details that would benefit from normals are rivets and bolts. Really, more important than that is the smoothing, soft-edges and overall highpoly look. Sure you can take a simply lowpoly mesh with tons of hard edges and photoshop some bolts on it, but its not going to look anywhere near as good as a nicely done HP + bake.

    At the end of the day if your HP mesh is only contributing 5% to the overall result, you either need to spend a lot more time learning how to do quality hp stuff, or you're only working on very simple unimportant objects with no complexity, which can make sense to just stick to a bump only.

    So is it acceptable? If you just want to "get by" and not have your work stick out from the crowd, yeah it is. But if you want to make really solid, high quality work that would be competitive in a portfolio, you need to get your HP -> low workflow down like it is second nature.
  • mLichy
    Options
    Offline / Send Message
    Everyone else pretty much has summed it up I think.

    But it also depends heavily on how big the object is in relation to the player,how important the piece is, and what kind of texture budget is left.

    At work we can do normals for things, but sometimes it's few and far between. It really depends what it is. For some really big shit, we just use tiling textures with detail maps to cover up the tiling.
  • Tom Ellis
    Options
    Offline / Send Message
    Thanks for the replies,

    It's not that I can't make high poly models, or even make them quickly, just that building a high poly model for a hard surface piece with a lot of 90 degree angles and hard edges and getting it to bake nicely seems excessive in certain scenarios.

    Earthquake,

    I get where you're coming from, but I sucked at getting across what I was trying to say in my original post.

    I fully understand that a detailed high poly mesh is paramount in making the most of the ability to use normal maps, and wherever possible, I would of course make good use of the opportunity to build a high poly. I'm not trying to 'cut corners' so to speak, just don't want to waste time.

    I'll use my current project as an example, I'm modelling this, or something very similar:

    http://www.made-in-china.com/image/2f0j00kewTWEuBvtzpM/Two-Post-Car-Lift-DTPF607-.jpg

    Now to me, I can't understand why you'd spend a whole bunch of time making a high-poly for that. But I'm not saying that is the right way to think, as I said I'm a complete beginner. Ok so some of the detail on the forked lift portions could use some high poly detail, but for the most part, especially on the pillars, they're box shaped, so would it not be a more effective solution to bevel edges, add smoothing groups, and add surface detail / corrosion in a Crazybumped normal map rather than a baked one.

    Again, I am merely asking... not saying my method is right.
  • EarthQuake
    Options
    Offline / Send Message
    I get what you're saying here, but i think the flaw in that logic is this:

    A. Item isnt complex enough to warrant HP

    BUT

    B. Simple objects like this should take next to no time to create the HP for, so you're not actually gaining much time by trying to fake the same thing with just a bump map.

    A painted bump will never compensate for smoothing, so you'de either have to use lots of hard edges on those pillar shapes(and have an aliased, very "gamey" looking asset), or add double edges to every hard edge to get it to appear soft, and it still wouldn't look as a good as if it was generated from a HP source. So you'll get some slight smoothing errors here without a HP, now is this the most important thing in the world? No, not really, but to me when you have some simple, flat shapes like this, even slight smoothing errors can throw off the entire asset, you lose the material feel as the specular wraps around the mesh in weird ways.

    However, some guy has come out with a method that may help you get around these problems:

    http://boards.polycount.net/showthread.php?t=66139
  • Tom Ellis
    Options
    Offline / Send Message
    Thanks a lot Earthquake, this is what I like about posting here, you get opinions and ideas from perspectives that aren't apparent in your own mind.

    The link is also a great help.

    Thanks again.
  • praxedes
    Options
    Offline / Send Message
    three are no sharp edges in real life. That's the main reason I can see for making a HP of the kind of model you've got there. All edges have a slight curve to them which catches light in a certain way- otherwise it'd be razor-edged on each side! The HP and bake will reflect this. It might seem subtle but the result to the viewer is considerable.

    ~P~
  • Tom Ellis
    Options
    Offline / Send Message
    Thanks for the reply,

    I see what you mean, and I am currently in the process of building the model in question. You're right, and as Earthquake said earlier, I can definitely see how touches that the HP will create are what sets apart average work from good work.

    Thanks again.
Sign In or Register to comment.