Hi, I want to do bumpmapping on an animated model, a smd(halflife1) skeletal animated model. I'm using tangents/biTangents/normals, but of course, they are not correct anymore once the model starts changing. I think I need "vertex skinning" to solve this, but I'm not sure though... So, what technique to use? And, are there any nice demo's of it somewhere? Greetings, Rick

well if you only pass the normal you can recalculate the tangent/bi in a vertex shader.
check this out.
Clicky

hope it was what you where after.

So I don't have to pre-calculate the tangents/biTangents at all? I always thought the calculation of tangents was depending on more than just 1 normal.

Anyway, thanks for the tip! But it brings to me to another problem. It has nothing to do with tangents, but it must work before I can do proper bumpMapping. The lighting is wrong. For now I'm just using simple lighting ( dot(L,N) ). But the model can rotate itself, so I have to convert the lightPosition (and camera) first to object-space in the vertex shader. Otherwise the comparison between the normal and the lightVector will be incorrect. However, I'm doing something wrong.

/*** VERTEX SHADER ***/	// Bring the vertex to world-space	float3	ow	= mul( modelView, in.pos.xyzw ).xyz;	// Calculate lightVector	// Light pos is an absolute coordinate, world-space	out.lightV	= normalize( lightPos.xyz - ow );	// Bring the lightVector to object-space	// modelView is a 3x3 matrix, don't know if that gives	// problems... lightV is a float3	out.lightV	= mul( modelViewInverse, out.lightV );	// The normal is already in object-space, just pass it	out.normal	= in.normal;/*** PIXEL SHADER ***/	// The result color	out.color	= dot( in.normal, normalize( in.lightV ) );

In the first case the lighting seems to be right, but when starting to rotate around the object, it's not correct. The lighting seems to be dependant on the camera angle(and how much you zoomed into it), but it shouldn't be.

I'm not good with matrices, so I guess I'm converting in the wrong order or the wrong things. For example, I have my doubts with "bring vertex to world-space". The model has not been animated/rotated/scaled or transformed yet, so the in.pos is already world-space. When multuplying it with the modelView matrix I get different results though.

So, how to bring the lightPosition(and camera) into object-space so that I can make a good comparison with the normal (and later bring it to tangent-space)?

Greetings,
Rick