> > > are you guys picking what normals index is supposed to represent what angle manually, using your eyes, or are you letting your algorithms do it for you automatically?

> > > i mean, are you using your programs to identify all instances of your 26 angles on the official voxels and seeing what indexes are actually used most frequently in those instances.
> > We determine the 'facing' (visibility) of the different voxels (FRONT,BOTTOM,LEFT,RIGHT,etc), and then use the analyser to see which normal goes with those facings most of the time (if you download the analyzer and play with the Koen-Index, which is actually the facing, you'll see that sometimes, there are multiple normals possible).

> What is flyby doing? from what i read it seems he's also looking for the indexes to use for your 26 angles. i think he's doing it manually though. there seems to be duplicated work. we should make sure to avoid that so we don't waste anyones efforts. if you're working manually then you should concentrate on finding the whole set of normals indexes and not on finding the set of 26 that is all the autonormaliser can handle.

> btw koen and will, after you find the set of 26 indexes to use can you find another set of 26 that use RA2 indexes and let people select which set to use in your programs? The RA2 set of 26 would be a little more accurate than the TS ones.
We can't detect all 26 indexes (angles) in existing voxels, we can detect far less (it depends on how you count though). But I thought for 3ds2vxl, you need to know the exact angles so you could do auto-normals there?
We (I should say I perhaps) do not calculate the actual angles, just the facings.

And switching to the RA2 set is not a very good idea - because analyzing existing voxels won't give out something like 'use normal 64' there, because there are so many options.

> > It's all that is possible for VXLSE - because we don't know the original 3ds angles.
> > I'm still improving the facings-determining code right now. It's almost perfect.