Questions regarding Bump Mapping

In my quest to master Blender, I’ve started to play around with Bump Mapping. Trying to keep it simple, I’m testing the techniques on a dice. However, the outcome is far from satisfying, and using YafRay as renderer, they look even worse (it’s quite possible that this will be solved for both renderers at the same time).

Since I’m modeling a dice with numbers on it instead of dots, I’ve played around with UV face mapping (or whatever it’s called) too. This appears to work fine. I’m using the same image file for both Bump Mapping and Texturing. The Bump Mapping is set to use UV values. The problem is that the outcome looks fake. Basically, the Bump Mapping is supposed to give the numbers on the dice a bit of depth, but unless the angle is perfect, it’s not even noticable (this might have something to do with the fact that the dice is pure black with white numbers).

So either I need to fix this (and somehow manage to increase the bump values – Nor is at max), or I need a good way of actually modeling the numbers directly onto the dice (which I’d actually prefer).

Suppose I were to model this out (which I indirectly asked for in another thread), how should I go about it?

A few points that should be noted are:
I’m using subsurf (catmull-clark) to smooth things out.
I’d preferably work in a way so that the numbers would be reusable in another model (i.e not a dice).

The implimentation of Nor (bump) mapping in 2.37a has bugs, it was much better in 2.2 versions (before Normal Mapping was introduced). However, it’s been fixed in CVS.

There are some things you can do. When using Nor use it on it’s own texture slot and turn off the value in the Disp slider.

Don’t use the “Normal Map” button in the Image buttons (it’s for real normal maps, not bump mapping).

The values of the black and white colors is what makes it work, so the whiter the white, the better the result.

You can get ‘real’ displacement with the Disp option and its slider. Works like Nor except that it works on geometry at rendertime so you need a very subdivided mesh to get decent results.

Modelling the characters and extruding the geometry from the dice is also possible (some may advise to do it the other way round) but I’ll wait for your reply before I search for threads that have discussed it in the past.


Going the textured route (which is probably the easiest) you will want to use Displacement Mapping probably.

Proper or Correct Normal mapping takes the RGB values of an image and uses that to change the normal mapping.

Look at: for information on Both mapping types.

Modelling would be hard. It would take a lot of vertex pushing and pulling. Though I suppose it is possible. Another method, though messy would be to make a Text object, convert it to a mesh and attach it with a lot of cuts and joins to your cube shaped object. All of this said, go with the texturing route unless its absolutely imperative that it be modelled.

I really only care about the rendered product in this case, so it really doesn’t matter what method I use. The reason I want to model it out is because you don’t have to worry about the bitmap (should you choose to resize/zoom in a lot). And I take it Blender doesn’t support SVG or similar as a texture/bump map?

Since I’m partly doing this for learning purposes, I’ll read through those links and give texturing another try, although I did manage to model some numbers (which took more time than it took me to make the whole textured dice).

Ok, I tried the displacement map thing. It worked, and would be the thing I’d use, if I owned a render farm. I’ve only got 256MB of RAM, and nearly ran out of swap when I tried to subdiv it to get an acceptable result.
With my hardware, I guess I’m left with the manual model way, if I don’t want to go with regular bump mapping. The question is just, how do I attach the numbers to the sides of the cube without distorting the surface (catmull-clark smooths everything out, leaving artifacts when turned on)?

Edit: Btw, is there a way to make sure that two objects’ surfaces line up exactly according to one axis? Right now I’m kind of guessing. Not that it’d make much of a difference.

When you are using subdivision surfaces everywhere there is a crease you should have 3 parallel lines of verts with just quads between. Another way of saying this is that you should have 2 face loops at all the creases. If you do this you won’t get any artifacts. When I went through the exercise of engraving text it took me about 5 to 10 minutes to do each number. When I get home I will post an example and a brief description of how I did it.

To line up 2 surfaces select a vert on 1 surface and snap the cursor there ctrl-s cursor to selection. Go to object mode and press center cursor in the editing buttons. Then select a vert on the second surface and snap the cursor there. Go back to object mode select the first surface again and ctrl-s selection to cursor.


I’d appreciate it. It seems I’ve forgotten some of the basic things you should keep in mind when working with 3D. I guess that’s what you get for not using things for years.

Took me a while to find it.

Blend file:


  1. add text. alt-c to curve. alt-c to mesh
  2. delete interior edges (bb in edge mode makes this fast)
  3. you now have a outline of your character in a mesh. knife in a vert either side of sharp corners. offset the mesh inwards – blender doesn’t have a command for this yet so you have to do it manually. Select 5 or 6 consecutive verts and extrude slightly inwards and tweak. Select the next 5 or 6 and repeat all the way aound.
  4. fill in the rest (triangles don’t matter here)
  5. select the edge verts and extrude (select non-manifold helps here) up slightly. The two bottom rows of face loops have been created.
  6. extrude again most of the rest of your depth, extrude one last time just slightly.
  7. extrude the verts outwards slightly this time as before.
  8. extrude out the plane and scale to the appropriate axis.
  9. done.

If blender had an offset coomand the whole process would only take a minute or so. Manually doing the offset changes the time required to 5 to 10 minutes depending on the character.

Good Luck,

I feel that, for dice, suitable effects can be achieved using bump maps (rather than displacement maps), and as you have seen, one is considerably more expensive than the other!

You get the most bang for your buck by putting an image on each face of the dice, then using a separate bump-map that is aligned with it.

If that doesn’t look good … then you are much too close to your dice! :slight_smile:

That might be the case, which is why bump mapping might not be too good for this sort of thing.

GreyBeard: Thanks, I’ll definitely give that a try. Oh, and I really like your Video Tutorials. Saved me hours of trying to find my way around Blender.