Weird output with occlusion map

Calabi

I messed with the levels and contrast and found this. All bits projected all over the place, it doesnt look too bad normally but does this mean I've done something wrong?

Mark Dygert

What app... what rendering method... what does the model look like?

Calabi

Rendered in XNormal. The normal map seems to look fine. I dont know could it be if I have the low poly too far forward?

SpeCter

If you are using xnormal take a look at this:
http://donaldphan.com/tutorials/xnormal/xnormal_occ.html

r_fletch_r

looks like photon samples. is your scale totally crazy?

Calabi

Thanks guys. I did check that tutorial, realised I forget the ground plane. It doesnt show up that bad normally.

Is it perhaps because floating geometry doesnt work that well with ambient occlusion maps? If occlusion maps are about bouncing photons off at all angles(well not all, you can see the angles from the image). I guess it must be, just surprised to see the floating objects defined so clearly.

haikai

SpeCter wrote: »

If you are using xnormal take a look at this:
http://donaldphan.com/tutorials/xnormal/xnormal_occ.html

This is my brother's tutorial, made several years ago. It's using an old version of Xnormal so be aware that a lot of the settings used no longer produce the same results. Also, I've heard some people have grievances on this type of "ambient occlusion." Typical AO maps I see don't make much use of a ground plane. Most people dislike having explicit lighting and shading baked in their diffuse textures when used in conjunction with normal maps, but we like to use it because it automates shading that we can use as a starting point on the diffuse.

As far as your problem is concerned, if you're floating geometry then you might have to render the floating elements separately to avoid the casting problems and then composite the two renders by multiplying the floating parts over the rest of it or something like that. It also looks like you need to increase the number of ray samples as your results should look quite a bit smoother than that.

r_fletch_r

Calabi wrote: »

Thanks guys. I did check that tutorial, realised I forget the ground plane. It doesnt show up that bad normally.

Is it perhaps because floating geometry doesnt work that well with ambient occlusion maps? If occlusion maps are about bouncing photons off at all angles(well not all, you can see the angles from the image). I guess it must be, just surprised to see the floating objects defined so clearly.

Its not really photons it just rays. what i was alluding to is it looked like enlarged photons. it could be that your scale is tiny and the samples are being calculated in world space. I have no idea how XNormal does it tho

I dont know if XNormal supports it but most AO solutions have a max distance parameter which restricts the shadowing to only nearby geometry. if the floats are beyong this range from the surface they wont shadow. pretty essential for baking so i guess it must have it.

Calabi

I upped the rays and messed with some of the other settings and it doesnt look to bad now. I'll just mask out the problem shadow parts.

Thanks for the help.

pior

try this :


That tutorial listed above ... it might be good for whatever the artist wanted to achieve at that time, but the result is nothing but an AO map. Then again it might have been really useful - but it simply isnt what he claims it is.

I personally like a 100% layer of real 'ambiant' AO, with a 50% layer of directional AO on top. By the way, Topogun renders great AO maps too (thanks Gunth!)

Calabi

Here's the result.