> > > > > 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?

> Both. They should agree with each-other. Statistics can point at a pattern, but manual search is needed to be sure.

> Therefore, the normals value will be by a file-lookup. Anyone can edit the file to tweak the autonormals. And you can always finish touching the model by hand.


> > > 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.

> Flyby's work is invaluable. It will be his results that are used as the lookup table.

> > > 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.

> As we only have basic facings instead of polygons to work with, the RA2 normals have more detail than we can use, so we don't need to tackle them.


well since i think TS normals format has only 6 o'clock directions and you are assuming 8 then you are going to get some indexes repeated. besides, with more possible angles in the ra2 format, the angles available are closer to the ones you want.


> > 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.

> Very true.


can't you pick the indexes that are used most frequently for each facing you distinguish?


> > > > 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.

> It is almost perfect. Much better than the old voxel editor :)