Took an old head model I´ve done and tested to do a normal map.
Found tutorials and scripts (used the search button) and found it to work quite well I think.
In the cvs version there is a great funktion in the Ctrl+E called EdgeLoop Delete. The only thing thats missing in that function is to merge the UVs after it removes the loop of vertices.
Btw is the ATI standard for Normal maps alot difrent than nvidias?
nVidia’s ‘Melody’, ATI and ORB will all produce similar if not the same quality normalmaps, the main difference between them really just comes down to one of usability. ATI is pretty much ‘command line’ stuff, Melody is buttons and file loads, ORB is inbetween.
Apperently you dont need have uvmap on the highpoly version of the mesh. Just UV map the lowpoly version and it will work great.
I used the world space normals on the previous meshes when i should have used the tangent space normals.
I even tried it on a sphere and it still worked. Like magic.
Still both normal maps act strange to the light that has been added.
On the world space normal map the shading is inverted like the lamp is on the back instead of front.
On the tangent space normal map it seems that the light has to be placed higher than the mesh to light the whole head. If its below then the mesh is dark (unlighted).
Open the image into photoshop (or whatever photoediting app you have) and go into the ‘channels’ tab - this displays the red, green, blue channels of an RGB image - and select the green channel (up/down iirc) and ‘invert’ it. Resave the image. If that’s still not right just select and invert each channel until it displays correctly (it’s usually the red or green channel that needs inverting).
What are you using to render this btw? A game like D3 or Blender? I’d need to know this to tell you what’s happening on the Tangent space problem, games like D3 don’t like solid models.
@kat-II: D3 disliking solid meshes? would astonish me as this is required for stencilled shadows <.=.<
@bigbad: blender is not able to use normal-maps correctly. it works with tangent space but only in one axis direction. your face has to lie in the ground-plane looking up to have any chance to get it right. i fiddeled around with normal maps in blender since some times and it is prolly said total useless implemented. unless this is fixed in an upcoming release you won’t get much from normal-mapping, only bump-mapping.
By ‘solid’ I mean meshes that aren’t ‘split’ (all the vertices are conneted to the same single object). To get smoothgroups (and the resulting light interaction) working correctly in D3 you sometimes have to split the mesh to create vertex seams (basically ‘subobjects’ in the palance of 3DS Max). In otherwords a ‘solid’ mesh sometimes exhibits the problem mentioned above where they don’t get lit properly (lighting the front lights the back of an object and so on). To solve that problem you often need to create a vertex slit somewhere (usually at the bottom of the neck).
I use blender internal render. But it´s not all useless. At least you get a preview look how it might look like in a engine.
My computer isnt the best either. It has internal intel graphics card that doesnt support realtime normal mapping. So I´m glad thats its even possible.
One thing to note here is that there are differences between engines in terms of how they like to see the orientation of each R,G,B channel. Although once you’ve got your maps they easy to fix, you might find that the same image in HL2 doesn’t look the same in D3 or FarCry becasue of the way they’re looking at the channels.
