Quote - Yes I think ice-boy is trying to create a map to modulate, specifically to decrease, the amount of AO that is applied.

To that end, ice-boy, you should not be putting black into the ears, if you're trying to make the ears less dark. That was your goal, right? You said the default AO calculation for the ears was too aggressive, so you wanted it to be modulated by a map that would prevent it from going too dark.

i didnt want to much AO behind the ear. in this map i tell poser that 100% AO is inside the ear.

Quote - Not sure where you are going with this, but the moment you change the morph/pose/expression, those maps wont be correct anymore. eg if you raise the arms, there will no longer be that much AO in the armpits. Or if you clench your fist, the inner nooks and spaces will be almost black.

This is why AO is always calculated at render time and you can't really paint a map of it. Well not unless your object is never going to move or deform at all, like in a game environment.

the map doesnt tell poser whre AO should be. AO is everywhere.
bagginsbill created showed us how with nodes to control the AO strength . with this map i can control the strength on specific parts.

Quote - Ah stupid me! I should listen to myself.

I said - faces up = brighter = less AO.

Duh.

Give me a couple minutes - will build a shader that does this automatically.

I am so stupid.

whait a minute. what do you mean will build a shader that does this automatic? are you telling me that you will build real AO? 
can this even be done?

The top two balls are original - unmodulated AO - straight from Poser.

The white ball shows my modulation function. This is what you would have put in the control map, but now you do not need to.

The two lower balls show the modulated AO. Compare each to the one above.

i didnt make the control map for the parts that are facing up.

I know that. But on what basis were you going to decrease the amount of AO?

The inner ears are mostly horizontal, but some parts are up and some are down. What exactly were you going to do? Just decrease the whole section?

And the way you were doing it was wrong. You said black is 100% ao, but the whole problem was, according to your comment on my VSS render, the ears had TOO much AO, so why ears = 100%? If you wanted less AO on ears, why did you make your control map have the MOST AO on ears?

Quote - Here you are. This use of AO automatically adjusts to which way the surface faces. If it faces up - there will be almost no AO. If it faces side - there will be 1/2 AO. If it faces down, there will be full AO.

The top two balls are original - unmodulated AO - straight from Poser.

The white ball shows my modulation function. This is what you would have put in the control map, but now you do not need to.

The two lower balls show the modulated AO. Compare each to the one above.

'up' may (is) not really specific enough... is there some way to base it on direction to light-source?

Quote - I know that. But on what basis were you going to decrease the amount of AO?

The inner ears are mostly horizontal, but some parts are up and some are down. What exactly were you going to do? Just decrease the whole section?

And the way you were doing it was wrong. You said black is 100% ao, but the whole problem was, according to your comment on my VSS render, the ears had TOO much AO, so why ears = 100%? If you wanted less AO on ears, why did you make your control map have the MOST AO on ears?

i didnt like  the part behind the ear. i didnt like AO around the nose and cheeks. 

i we dont understand each other. when i have an infinite light infront of the object then i will still see a lot of AO on the face. i dont want this. it makes the face look dirty. this map is for cheating and making a better render.

but what you showed me now is also very good. it will help me and others very much.

Erm.. actually, I was just re-iterating what you said above about this being a partial-solution that may or may not be appropriate for each case.  I guess you could just fine-tune the angle by hand for exceptions... so I was mostly asking if you had enough info/node-control to base it on the (main? brightest? total?) light-source.

[EDIT: cross-posted.. this comment was not meant in response to ice-boy]

[EDIT #2: Hmm... lambert-shaded AO maybe? ;) ]

...just rambling here (brain-fart)...

Let's assume the "up" you mention above is really just another way of describing where the light is coming from (which, it is :) ).

Suppose you found the location / direction of each light-source (unsure atm how to treat IBL, without knowing the pixel brightnesses).

Suppose you also assigned a weight to each light-source, based on it's intensity.

Then you compute 'up' (direction of light) as the average of the positions (or at least directions) of the sources in the scene, weighted by thier respective intensities.

...would that do us any good as 'up', relative to your AO calcs, above?

...I'm not terribly familiar with how all the nodes work, so much of the above may be unnecessary.  The main gist of what I'm suggesting is that the polygon normal should be relative to the light-source - and not the world (which may or may not be doable/easy with Poser nodes in this context).

Quote -
'up' may (is) not really specific enough... is there some way to base it on direction to light-source?

Sure but why bother. Just shine the light on it and get rid of AO.

We're dealing with a particular phenomenon here, which you guys are not seeing in this thread. Ice-boy saw a render I posted and noted that for that particular outdoor scenario, I had too much AO on the ears and face. This is because AO is an approximate technique, not accurate. (Given the approximation, I wish it were a lot faster than a proper GI technique.)

Anyway, the fault lies with material based AO and the inaccuracy it produces for directional lights. We should use light-based AO and then directional lights don't matter. Everybody needs to bug Smith Micro about this. Then we can get the problem dealt with and stop hacking.

I don't see a point to dealing with directional lights in AO because AO is supposed to be for non-directional lights. The kind of detailed adjustment needed for directional lights could only work if we could actually tell if the directional light is shining on this currently rendered point or not. We can't tell that in the shader. Also, we can't tell which direction is mostly occluded either. I'm just showing you another hack that - under particular circumstances - will reduce the error you observe.

If you want to modulate it - modulate however you like. I'm thinking this modulation is pretty helpful.

I'll post some images in a second.

I used a mirror so we can see the ear from two angles.

Click for full size on both images, and flip between the two windows or tabs in your browser to compare. It is a subtle difference.

Sure, but that's a lot of work. I'm just saying, for my standard outdoor setup, what should you do? Answer - react to the sky as dominant light source, after the sun of course.

You are talking about subtle differences here now that would not show up in a render as being obvious problems. We have bigger issues, like unreal armpits and such. We've moved well past the AO being the #1 issue preventing us from fooling someone into thinking a render is a photo.

One more image - this time using light-based AO. This avoids suppressing the "sun" light, but it has artifacts. If you don't see what I'm talking about - then you're good to go and you should use light-based AO. I see them and they bug me. If you see them, right a note to SM and tell them to fix this renderer. I'm really tired of the whole "ray bias" bullshit. It's a lame excuse. Other "free" renderers do not have this issue.

Gotcha... I was not thinking about the difference between light-based AO and material-based AO.  I guess I (and you, in this instance) was trying to come up with a psuedo-light-based AO hack on material-based AO nodes - and what we really need is correct/fixed light-based AO.

did you use here 100% AO? or did you use teh blender node?

i use AO only for very bright lighting. so for sun light. when the figure is outside. and then its always above the figure. so this will work great.
but can we control here also the strength?

ice-boy I posted 4 pictures as follows:

#1) AO Node from VSS PR2 skin shader. This is 100% AO into Diffuse_Value
#2) AO Node into Blender. Blender modulated as above, using N node.
#3) No AO at all.
#4) AO on IBL LIGHT, at 50%.

In all four images, I had Beach lights, IBL = 60%, Main light = 100%. I think I moved the Main light from the default VSS package. However, it is the same in all four of these images.

so the second pic has also 100% AO.

No it does not have 100%. If it had 100% why would it be a different picture, since the first picture is 100% AO?

I said the second picture is modulated by the N node. That means that the N.y value (which goes from -1 to 1) is being multiplied by .5 (thus -.5 to .5) and added to .5 (thus 0 to 1). So the amount of AO is between 0% and 100%, not always 100%.

Sorry, but I see no difference between the second and third image. Are you sure #2 has AO?

As I said we're now into very subtle differences. The objective of the modulated AO is to avoid producing AO in some places. For example, on either side of the nose, 100% AO produces a noticeable darkening that is overdone. The modulation based on N.y decreases the AO there, because the cheek bone area under the eye and beside the nose actually points up a bit.

To see such small differences, first of all you have to have your monitor calibrated very well. Second, you need to do a flip-test comparison.

If you are using the Internet Explorer (or other tabbed browser) click on both images to load them into a tab by themselves. Then alternately click one tab then the other. With the images in each tab, they should be perfectly lined up on your screen, so that flipping tabs will replace one image with the other. That makes it easy to see which pixels are different.

Oops, sorry! I did that, but I must have clicked one of the images twice.

... just ignore me ...

ok i will first post some pics.
when you render with AO on you get this . right?

but there should be no AO from where the light comes from.

i did this in photoshop. i rendered an AO pass and a matte.

i am now experimenting if i can in the shader make a matte. i was thinking maybe with the diffuse node? so that it then ignores the white part of the diffuse and makes AO only where the ambient light comes from?i think this would make AO more realistic. i think this could be done since bagginsbill used the diffuse node to make a shadow catcher.

what do you think?

ice-boy what do you mean there should be no AO where the light comes from?

AO is supposed to be used to block the IBL. The same spot may also have a directional shadow. The directional shadow should correspond to the blocking of directional light, and the AO shaodw should correspond to the blocking of ambient light - usually implemented as IBL.

If you use light-based AO on the IBL, this will work perfectly.

If you use material-based AO it will not work perfectly, because the darkening will affect the reflected light from all light sources, not just ambient IBL light sources. That's why it would be better to use light-based AO. But if you get too many bad results from the LBAO, you have no choice. So that's why I modulate the MBAO, to take into account that not all of the diffuse reflection is actually from ambient light, so the amount of darkening has to be tweaked. It is not based on iterating over each light in the shader, and only doing MBAO for IBL lights. There should be a switch on the MBAO that tells it to automatically adjust for the proportion of the diffuse light that is from IBL versus directional, but there is no such switch. So we have to hack it. That's why I use the N node.

this just occurred to me.  is this assuming you're already using very accurate directional lights?  that is, if, with no IBL, if your image produced glowing areas like nostrils, parts of the ear, etc.,  wouldn't you still need AO to affect the work of directional lights?  i ask because i'm sure i'm not the only one who can't get rid of horrendous grain in raytraced shadows, and i know for a fact that most people don't put up with renders that take as long as mine.   i'd bet that most people still use depth mapped shadows, and those can be hard to get to work well and on a detailed level. 

but maybe i'm wrong and people work with more accurate lights.

Depth-mapped shadows can do small-scale if you work with the shadow cams. There's no reason to use AO just because you have DM shadows. I wrote about this years ago - the No Nostril Glow thread.

If you like the speed and control and quality of soft DM shadows over RT shadows - use them.

By the way, AO is just as grainy as blurred RT shadows if you set the samples to the same level as is used for RT shadows - it uses the same mechanism. But RT shadows don't let us adjust number of samples, whereas with AO you can, and that's how you get rid of AO grain - more samples.

And I don't think blurred RT shadows are that bad. Have you experimented with shading rate? The shading rate of the surface receiving the shadow will influence the number of samples used to generate the directional shadow.

Quote -

And I don't think blurred RT shadows are that bad. Have you experimented with shading rate? The shading rate of the surface receiving the shadow will influence the number of samples used to generate the directional shadow.

raytraced 20 blur shadows and a shading rate of ',2 and i still see a lot of noise. if you use a bright IBL then you can get away with it. but if you use dark shadows its very bad. 

Quote - ice-boy what do you mean there should be no AO where the light comes from?

AO is supposed to be used to block the IBL. The same spot may also have a directional shadow. The directional shadow should correspond to the blocking of directional light, and the AO shaodw should correspond to the blocking of ambient light - usually implemented as IBL.

If you use light-based AO on the IBL, this will work perfectly.

i tryed this and it loosk better then i remember in poser 7. a little confused right now. hmmmm.

i really think in poser 7 when i used light-based AO it still made a lot of AO where the direction light was.

Quote - But RT shadows don't let us adjust number of samples, whereas with AO you can, and that's how you get rid of AO grain - more samples.

AAAAMEN. i have been saying this for 1 or 2 years now. other software have this for raytraced shadow. samples.

why do i have to use the best render settings for soft shadows? it should be a seperate option. i really really really hope that poser 8 will have this. because if not then its really pathetic.

Quote - Depth-mapped shadows can do small-scale if you work with the shadow cams. There's no reason to use AO just because you have DM shadows. I wrote about this years ago - the No Nostril Glow thread.

i know.  i followed it.  and had lots of problems in different situations.  i encountered enough that i eventually switched to raytraced, despite the grain.  and i'm pretty decent as Poser users go, and not but so complex with my scenes. just from complaints in the forums, i'm betting most people aren't willing to take the time it takes to make them work well in most scenes, nor able to master them well enough to avoid serious problems. 

basically, i think you're right, it's technically possible for some scenes if you're willing and able (given your hardware) to get your shadow cam settings right.   but after alll that tweaking, some of which makes your light inaccurate or uses huge resources (avoiding clipped shadows and getting the right res shadow can be difficult at larger sizes), you still have to wait for it to render for much longer than most people are willing to wait.

edited to clarify:  basically, i think you're right and i'm not disagreeing with you, just trying to make it clear that this information is to be used in conjunction with other info.  it only negates all those threads about using AO to get rid of nostril glow if you are using your shadows accurately in the first place, which i think most people aren't.

Quote -
By the way, AO is just as grainy as blurred RT shadows if you set the samples to the same level as is used for RT shadows - it uses the same mechanism. But RT shadows don't let us adjust number of samples, whereas with AO you can, and that's how you get rid of AO grain - more samples.

i know.  i do that with AO, and the "quality" settins on reflections and refractions.  i thought maybe the pixel samples settings would affect the shadows as it does the depth blur (if you use it), but that doesn't seem to be the case.

Quote - And I don't think blurred RT shadows are that bad. Have you experimented with shading rate? The shading rate of the surface receiving the shadow will influence the number of samples used to generate the directional shadow.

yep.  experimented with pretty much all the settings in the rendering options.  for instance, i found that if you render large enough, and keep your shading rate too low, your depth mapped shadows become too accurate and make what looks like a contour map on surfaces.    i haven't done a methodical test in ages, though.    unless i get some wildly upgraded hardware, doing so would probably take a long time. 

it's probably not worth it right now since i'm still on P6 (though i think a significant amount of other users are, too) and will upgrade in the future.

baginsbill:
there was a thread where someone asked you something about ambient occlusion. you showed him a trick to control the radius of AO with nodes.
i can not find the thread. do you know the title? 

Do you mean the AO "distance"? I don't remember talking about that. I do remember talking about AO node Strength doesn't work, on page one of this thread. I use a Blender to implement strength.

http://www.renderosity.com/mod/forumpro/showthread.php?message_id=3099237&ebot_calc_page#message_3099237

this happens when people are writting strange titles.

i am now searching for more then 30 minutes.

you posted screenshots. a white plane that looked like a wall and you also used a ball. the ball was next to the plane. and then you showed i think some math functions. i think it was bias.

Good clues. I found it.

http://www.renderosity.com/mod/forumpro/showthread.php?message_id=3195532&ebot_calc_page#message_3195532

THANK YOU

how did you found it?did you remember the title? 

No, I typed into the search here:

Find "AO Bias" posted by "bagginsbill" older than 100 days sort by relevence

Then I found it in the first page of results by reading the excerpts to see what I said.

aha ok.
 so in this htread you connected teh AO to math bias and then to blender. you can make it lighter and softer.

how to make it darker? to connect a match function in blender input 1?

One way to make it darker is to change the Bias value to below .5. In the examples I showed, I only demonstrated raising the bias above .5.

The Bias node is mathematically identical to a Pow(er) function. It differs only in that the exponent is expressed in a more convenient way.

Mathematically, Bias is defined like this:

Bias(a, b) = Pow(a,  ln(b) / ln(.5))

where ln is natural logarithm.

[ Curious note for math-heads: Strangely, the Bias function is commutative, i.e. Bias(a, b) = Bias(b, a) ]

Anyway, the way I demonstrated that shader was:
Blender(1, Bias(AO, b), d) where b is .5 to .7 and d is 1 or less.

The b value selects an exponent such that Bias(.5, b) = b. In other words, given the input value is 1/2, the output value is b. The rest of the curve is constrained to be an exponential curve that passes through the points [0, 0], [.5, b], [1, 1].

When b is .5, it means the curve is a straight line from 0 to 1. When b is less than .5 the curve bends downward, and when b is greater than .5 the curve bends upward.

When the curve bends upward, the visual effect is that the shadow is brighter (less dark). When the curve bends downard, the visual effect is that the shadow is darker.

In either case, the Blender is blending the result with white. That means that as you decrease the blending value, d, below 1, you get more of that mixed with white. This effectively changes the value when the Bias produces a 0 to something higher than 0.

So, to tie it altogether, d selects the maximum darkening (0 means no darkening at all, 1 means darken all the way to black) and the bias, b, bends the curve between the darkest and lightest values. 

thanks.

i think i actually understand what you are talking about.

Quote - ice-boy what do you mean there should be no AO where the light comes from?

AO is supposed to be used to block the IBL. The same spot may also have a directional shadow. The directional shadow should correspond to the blocking of directional light, and the AO shaodw should correspond to the blocking of ambient light - usually implemented as IBL.

If you use light-based AO on the IBL, this will work perfectly.

If you use material-based AO it will not work perfectly, because the darkening will affect the reflected light from all light sources, not just ambient IBL light sources. That's why it would be better to use light-based AO. But if you get too many bad results from the LBAO, you have no choice. So that's why I modulate the MBAO, to take into account that not all of the diffuse reflection is actually from ambient light, so the amount of darkening has to be tweaked. It is not based on iterating over each light in the shader, and only doing MBAO for IBL lights. There should be a switch on the MBAO that tells it to automatically adjust for the proportion of the diffuse light that is from IBL versus directional, but there is no such switch. So we have to hack it. That's why I use the N node.

so there should be no AO where light comes from.so since material-based AO doesnt work how it should i made a shader.

just playing around with nodes a little. i used gain to make more contrast in the diffuse. and used bias to make it darker.

That's a very interesting idea, ice-boy. It certainly works in this situation. You have to tune it for each scenario, of course, but it's a useful technique I think.

Go put a notch in your belt, you're the first to teach me something in the material room in a long time.

well even if we connect the diffuse node direct in the blender value it still looks good IMO.
 its not like the light based AO. but from what tests i made it looks better then just normal AO.
of course my test renders are not always the best.

i am just trying to use your teaching from every thread and i am playing around.

''You have to tune it for each scenario, of course, but it's a useful technique I think.''

where for example? you mean the Gain?

Well not exactly - perhaps. It depends.

ok