I think most of the assets have been released with 3D files (.blend?) so you could render your own top-down sprites.

I think the enemy sprites and spell effects should work without change.

If you're talking about the tileset sprites, sure I'll make some time to render those another way.  What dimensions should the floor tiles be (e.g. 32x32 square?).

Ok. I'll have to experiment.  RPG maker style games use square base tiles (32x32), but in this 3/4ths perspective the base would be rectangular (32x16 for example). I'm not sure the best way to handle this (when rendering side tiles, put two tiles together? stretch the perspective?).

This is actually a fairly common issue with pixel art.  The Les Forges tutorials point out that the perspective used in RPGs is an odd sort of othographic projection that maps as follows (assuming your origin point is in the lower left:

• x' = x
• y' = y + z

What that means is that a square is still square, and a cube is precisely 2 tiles high.

In my personal opinion, that looks kind of odd.  That said, if you're doing something more like:

• x' = x
• y' = y + z*2/3

You end up dealing with some odd issues because you lose the one to one mapping between y and z.  On the other hand, this perspective looks a lot more natural.

I'm pretty sure that if you wanted to achieve the first perspective, you would have to have an orthographic camera that looks down at exactly 45 degrees and then stretch your aspect ratio vertically by a factor of 1.414 (that's a wild guess though).

> I'm pretty sure that if you wanted to achieve the first perspective, you would have to have an orthographic camera that looks down at exactly 45 degrees and then stretch your aspect ratio vertically by a factor of 1.414 (that's a wild guess though).

That will give you: y' = y + z*0.707

I would go for : y' = y + z*0.5 (You can scale z by 0.707 as well to achieve this effect.)

How about shearing your model, so that z' = z + y  (with x, z as floor, and y up) and then rendering orthographically top down?

(exactly what bart said, but two steps: afine-transform, render-ortho)

This thread is fun to read for people who have no idea about perspective or 3D modeling. :)