Need some advice on Normal maps...

Support Modeling
#1

I hope somebody can answer this “slow brain” a question.

http://i84.photobucket.com/albums/k31/Julius_Sneezer/bridge2.jpg

I built this bridge and since I’m still new to Blender(Not an expert, but have been following every tutorial I can get my hands on).

I see people’s scenes and see incredible detail, like trillions of polygons or something.
Do people really use trillions of polygons? My computer wouldn’t be able to handle that many and besides editing to see what something looks like after one 12 hour render would take 100 years.

My bridge had very few polygons, I cleaned it up some and got it later down to 40,000 polygons. But I tried to add rivets to it. Once you add detail polygons the polygon count goes way up! It could turn into several million polygons. So I figured why not use normal maps for all of the rivets on the bridge? I did that, but after about 80 time consuming normal maps created this one model could take forever. So I said to myself, Is this what everybody else does?
Hundreds of normal maps per model? Can somebody give me just a little advice on how better to rivet my girders on my bridge?

#2

Well, usually you bake your own normal maps for stuff like this. The normal workflow would be to make a Highpoly Scene, then bake the normalmaps, make a lowpoly scene and apply them. Some make it the other way round, make the low poly first and then apply additional detail which is IMO not that good for reasons not relevant to this topic.

However either way you will end up with a good truckload of polys and there is no sense in making normal maps for you because you want to render a still for what i understand and not use the model in a realtime engine where you need low poly count. Normalmapping is mostly meant to improve realtime performance, although it can improve rendertime as well ^^
And there are not hundrets of normal maps. Normal Maps are not really used for untextured models. Normal maps are applied as UVmapped textures, which means you would have to unwrap your whole scene.
I wonder what others in the forum say on this topic.

My advice, lower your project expectations or get a faster puter, but nevertheless intrest yourself in normal maps. They are a important part in CG.

And as ultimate advice, use procedural materials and play with the node editor. A bridge is mostly made of tarmac, concrete and steel, mostly painted… Those are materials you can perfectly redo with the procedural materials in blender… take a look at blender-materials.org if you don´t know it already.

An example for concrete:
http://www.blender-materials.org/index.php?action=view&material=361-concrete
Not the best, but many procedural materials include algorithmic generated textures and with the NOR feature in the MapTo panel of the texture you get some decent “bumpmaps” giving your project more detail.

#3

That would be great if it were possible. I noticed it doesn’t seem to work for surfaces other than one at a time, like I mentioned before several hundred normal maps per model? I already tried using fewer maps and it distorted the models.

For example, I tried creating normal maps for the hand railings, but found it distorted them when I applied the maps, even after trying to fit them, they only seem to work at one angle. I havn’t tried the latest version of Blender yet. Something about Tangent baking? Maybe that will solve my problems.

#4

I think you just have to learn UV mapping.
http://wiki.blender.org/index.php/Manual/Unwrapping_a_Mesh

Do you know papercut models? You have to fold them to a certain shape and then glue certain parts together, so in the end you get something like a car… I think the most common is a dice that children use in primary school. Now, Uv mapping is almost this process but backwards. You cut up the surface of your model, so the surface can be “transfered” from 3d to 2d, and after that you can paint on that with for example GIMP. I have to clarify my previous sentence, so, actually this cutting up won’t change the geometry in any way, you just map the 3d surface to a 2D coordinate system “virtually”, and Blender will know which part of the resulting 2d map should be transfered/mapped back to which part of the 3d surface.

This way you just need only a few, or just one texture maps, depending on your object and how you cut its surface.