sm3 map (beta) TwoSeas

[SwiftSpear]

When posting SM3 textures please include normalmaps...

This is a bit much to ask..... Considering how easy it is to create them.

Normal maps are a bit more complicated then just running an image through the nvidia plugin. While the nvidia will produce a fine workable normal map in most cases, better normal maps can be produced using other techniques, different autonormal generators produce different results as well, and if you understand how normal maps work you can even make customizations yourself. I'd like people to post normal maps when sharing textures because of the diverse ways you can potentially make normals, different normal maps may be better or worse, and normal map artists may be able to improve your normal map and get you a better version if it's shared. Additionally if multiple people work on the normal map for the same texture and all post their results then map artists will potentially have several different normal maps of varying qualities to choose from. I think it's just a good standard to start when the new format really starts rolling and people start compiling texture databases and what not.

[edit] for example, I can tell from that normal map that it isn't optimum, it thinks the vegetation is scratches in the rock face, and it isn't rendering much depth in the crags, that normal will make the rock face look a bit like a ripped up brick wall. Tolerable but not optimum...

[mufdvr222]

Stuffing around with different textures, including Forbodings spiffy rock texture

<edit> Yet another texture change,, sandy bottom and darker grass.

[Das Bruce]

Better, but theres still something missing.

Oh and never use an underwater water texture. Please.

[mufdvr222]

HAHAHAHA Its alright, its certainly not staying, Im actually thinking; something dark.
I think the "theres something missing" is a decent texture variation in the grassy areas,, its too I dunno,, errr green!

[SwiftSpear]

Better, but theres still something missing.

Oh and never use an underwater water texture. Please.

The minimap?

[Weaver]

HAHAHAHA Its alright, its certainly not staying, Im actually thinking; something dark.

I am not a big fan of water maps but I think we a still waiting to see a good sea bed texture.





My other thought is that we don't usually set water absorbtion high enough to properly represent the depth of water at the scale of the game.

[hrmph]

Well swiftspear feel free to post the normal maps for your textures when you post them... Muf I'm really liking the look of the sand, although I think the heightmap needs a bit more stretching on the beachfront areas to make it seem more like a coastline. (I even like the underwater texture )

[SwiftSpear]

Well swiftspear feel free to post the normal maps for your textures when you post them... Muf I'm really liking the look of the sand, although I think the heightmap needs a bit more stretching on the beachfront areas to make it seem more like a coastline. (I even like the underwater texture )

Absolutely I will. I hope to be able to renormal some of the existing textures that have come out of the various texture packs posted so they are properly 3D normaled rather then 2D pass normaled.

[mufdvr222]

Swiftspear can you elaborate a little, Your basically saying we are not making propper normals,, Can you look at the screenshot and point out the shortfalls of the Nvidia developer tool most people are using or what I/we are doing wrong?

[Forboding Angel]

Swiftspear can you elaborate a little, Your basically saying we are not making propper normals,, Can you look at the screenshot and point out the shortfalls of the Nvidia developer tool most people are using or what I/we are doing wrong?

Ya, especially considering that I made whakamatunga riri using the nvidia plugin, and I think it looks frickin awesome.

[SwiftSpear]

In order to answer that question I'm going to have to delve a little into how normal maps work and what the nvidia plugin does when it produces it's normal maps. What a normal map does to your texture when applied in a game engine is tells the game what direction light hitting the texture should be bouncing off the normal map. the red channel contains the direction the field is pointing in the x dimension, the green channel contains the direction the field is pointing in the y dimension and the blue channel contains the direction the field is pointing in the z dimension. Each pixel therefore will reflect light in a different direction then each adjoining pixel based on the normal map values. What this practically means is that the only perfect way to produce a normal map is to know the 3d dimensions of the original texture your 2D texture image is based off of.

The nvidia plugin does not do this, what the nvidia plugin does is some fancy guess work. The plugin reads the surface of the texture you provide and looks for dark spots and light spots then guesses based on what your dark/light distribution is what the height map and x/y maps should look like for the texture. For many textures this isn't that bad, it largely does a very good job, but for some, like the foliage rockface texture that has lots of light variation due to the different shades of rock and the different lights and darknesses caused by the plants on the face it ends up thinking there are alot more cracks then there should be and that the texture is much more flat then it is intended to look.

There are various methods of manually creating a normal map. Hand drawing it, while possible, will rarely produce good results. If you are making your textures off a photo referance, using this technique can be incredibly useful. However, in the case we're dealing with here what I intend to do is make high poly models that represent the way the texture is supposed to look the most ideally and then using a normal map generator that will digitally use the process shown in that tutorial on the model and produce the results automatically. Some normal maps will see a massive improvement from this process, others, like sands and flatter surfaces will most likely look best when normal mapped through the nvidia plugin.

To clarify, the nvidia plugin isn't the wrong way to do it, it just isn't a method that can't be improved upon in some cases. For a perfect example of the nvidia plugin's limitations look at ficwill's post with the large images half way down the page here. The normal map the nvidia plugin has produced for the macdonalds logo hasn't estimated the correct shape of the macdonalds logo, and will make it look like a metal plate rather then pyramided the way it is on macdonalds signs. The only way to improve that would be to create a model of the logo and make a normal map for that model, ficwill says he doesn't want to do so in his next post after mine in that thread.

[edit] Fun normal mapping tutorial I found

[mufdvr222]

Ok, I see your point. I used that,, err large fast food company logo and applied the nvidia normal map filter, the result was nothing like the example in that post.