Quote: "If you are indeed able to render your model with the bone animations working in your screen pass I don't see why it shouldn't be possible to do so in the depth pass as well - if anything it would stand to reason that there are less constant registers in use in the depth mapping shaders than in the final render ones."
Quite true! Thanks for the reminder.
The only downside is a loss of performance because doing the animations in the shader results in a performance boost - but that means nothing if the object is too complex for the shader! So, yes, you should be able to get dynamic shadows very easily using a simpler version of the shader which relies on CPU animation.
Quote: "How this works with multiple objects, for example multiple characters in different positions, what do I have to change on the code? just adding the 1st shader on each character and the calculation on the do-loop cycle will be 1 only? also what about the 2nd object (the floor), if I apply that effect just to the floor it would reflect the shadow only on that object, so that I have to apply the 2nd shader also to for example walls? In terms of performance will be heavier? both to use multiple shadows and different areas where to project the shadows"
You may need only one shader - and apply it to each object in the scene. The shader doesn't care whether it's a character or a wall or floor.
You'll only need more than one shader if the main render pass of the scene uses more than one. In that case you'd need a shadow technique in each shader. Depending on how complex your scene is you may need more than one shadow map - but that's another matter.
