MasterFunk Posted November 12, 2008 Posted November 12, 2008 I just learned how to make Normal maps in another post. Now that I know how to do it I am perplexed as to what the difference is. What is the difference? Quote
John Bigboote Posted November 12, 2008 Posted November 12, 2008 Million Dollar Question! If they actually raised and lowered the geometry, like a displacement...they would be golden. From the examples in that other thread it appears that they do not, but DO in Maya...great. A bump is so simple to create and implement and gives great effect -other than that it leaves the actual SHAPE of the underlying geometry smooth/flat. I would love if somebody could expound the virtues of the normal map (which has nothing to do with normals?) for me and Masterfunk. Quote
zandoriastudios Posted November 12, 2008 Posted November 12, 2008 It is like a bump map, but it uses the 3 channels of an RGB image (rather than 1 channel of a grayscale image) to fake the detail. The best way that I have found to create normal maps is to use a program called "Crazy Bump". http://www.crazybump.com/ Quote
cronos Posted November 12, 2008 Posted November 12, 2008 Normal mapping is a type of bump mapping that involves a normal map, an image that contains an XYZ direction vector ( normal vector ) at each pixel to be used in polygon surface lighting calculations. Some types of bump mapping are done without a normal map. With all types of bump mapping, the goal is to increase surface lighting detail without adding more polygons or vertices Quote
John Bigboote Posted November 12, 2008 Posted November 12, 2008 Nice and succinct, as always- Thanks Will! I like CrazyBump...and the price is right...and they call humans 'puny' (Hulk SMASH!) So it can't be all that bad. I've got to get aboard the 'Normal' train... Quote
Fuchur Posted November 12, 2008 Posted November 12, 2008 I dont know exactly, but if I am not wrong, normalmaped surfaces can be affected by lights in the scene (so the additional detail will be affected) while bump-maps cant do that. Displacementmaps on the other hand can manipulate the surface and will add geometry-details to the model (it wont change the mesh so, it just at it on top). A terrain-map (for example used by the Terrain-Plugin) will change the geometry. I wrote a short explaination on my website on these kind of maps when normalmaps came out... you may want to have a look at it: Difference between Bump-, Normal- and Displacementmaps I am not such an expert as Yves or so, but I think I am quite sure that it is more or less correct. *Fuchur* Quote
Hash Fellow robcat2075 Posted November 12, 2008 Hash Fellow Posted November 12, 2008 Judging from the Crazybump video, bumps made with normalmaps seem to do better at showing a raised surface occluding a lowered surface when you tilt the texture to see it at an angle, while bump maps fail at that. Do Normal maps rendered in A:M do this? Edit: I take that back, the Crazybump people are also extracting displacement maps, so I guess that is what is responsible for the good tilting effect? I dont know exactly, but if I am not wrong, normalmaped surfaces can be affected by lights in the scene (so the additional detail will be affected) while bump-maps cant do that. A:M does take lighting into account when rendering bump maps, otherwise you wouldn't see the bumps since they are suggested by shading shanges on the surface of the object. Quote
MasterFunk Posted November 12, 2008 Author Posted November 12, 2008 Thanks for all the responses. So basically normal maps have more accurate lighting. Neither change the actual mesh. Is that right? Quote
John Bigboote Posted November 12, 2008 Posted November 12, 2008 Thanks for all the responses. So basically normal maps have more accurate lighting. Neither change the actual mesh. Is that right? I think that is a fair assessment. Quote
HomeSlice Posted November 13, 2008 Posted November 13, 2008 Technical explanations aside, I haven't noticed that much of a difference between normal maps and bump maps in actual use. However, there have been several times when I wanted to take an existing model and convert it to a bump map or (rarely) a displacement map. An example would be to render a color pass and a normal pass of a modeled set to use in another set as a 2D background matte. It is really easy to render a model as a normal map using Willi's normal map ambience shader, but it is almost impossible to convert a model to a bump map. So, for me, normal maps still have a very practical use. Quote
R Reynolds Posted November 14, 2008 Posted November 14, 2008 I use normal maps to cheat details that don't need to stand up to arms length inspection, so there's no need to carry the overhead of those patches. Bump maps are for textures that are so small with respect to the overall surface they essentially have no depth. For instance, the cast concrete, octogonal rosettes on either side of the front wall of building I'm working on (shown in the attached WIP image) are normal maps derived from a 600 patch model. They're so far away from the viewer that it's unlikely you'll notice they have no real depth. The second image shows an angled view comparison of the original rosette model next to applied bump and normal maps (both set at 100%) that were derived from the model. IMO the normal map (although not nearly as good as the model) is far more convincing than the bump map. Quote
John Bigboote Posted November 14, 2008 Posted November 14, 2008 WOW! Thanks for your input and illustration, Rodger! What software/process do you use to derive your NORMAL and BUMP mattes from A:M models? Quote
Hash Fellow robcat2075 Posted November 14, 2008 Hash Fellow Posted November 14, 2008 Thanks Roger! Interesting that the mapped versions of the rosette anti-aliased better than the spline version. Quote
HomeSlice Posted November 15, 2008 Posted November 15, 2008 That's a great example Rodger. Thanks! Quote
R Reynolds Posted November 15, 2008 Posted November 15, 2008 What software/process do you use to derive your NORMAL and BUMP mattes from A:M models? Nothing magical. A bump map comes from applying a simple gradient material that sets the color of the top of the model to (255, 255, 255), the bottom to (0, 0, 0) with a smooth transition between. Make sure you set the model's specular color to (0, 0, 0) as well. A render of a top view of such a model yields a decal where height is proportional to gray value (see attached image). As a side note, in the case of the rosette model which is a raised detail, if you applied a gradient that set the top of the model to (255, 255, 255) and the bottom of the model to (128, 128, 128) you would now have a map suitable for a displacement decal (if you can afford the render hit). The normal map is a simple application of aaver's MakeNormalMap material. found here. The only limitation being that you should test how the map renders to make sure it's default settings provide the expected results. Interesting that the mapped versions of the rosette anti-aliased better than the spline version It may be due to whatever interpolation is happening in the translation from decal to apparent surface grey value but it may also be due to the fact that I modelled some of the contours too sharp. Bump, normal and displacement maps don't play well with steep slopes i.e. rapid changes in height. They prefer rolling hills as opposed to cliffs. So I slightly blurred both maps before applying them, creating the softer edges. Quote
Recommended Posts
Join the conversation
You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.