"gather" is still pretty slow in P7 IMVHO. not much point in using it, unless one hasn't got any other renderer.

Did you know that P7 actually has a Global Illumination feature in the renderer?

They didn't publicize it. There is nothing in the user interface to turn it on. But you can turn it on from the Python interface. I read something about this in the Python forum here. One of our Python gurus found these undocumented functions that actually make it do GI.

I guess it doesn't work very well, or they would have made it visible. If it had a good version of GI, I'd want it, even if it needed all night to render.

Really?

Now that would be something worth investigating.

If the irradiance caching works, it won't take all night to render. Looking at what you said about the gather node's workings, coupled with the irradiance caching; it looks like they maybe trying to implement Final Gathering, which really only needs about 500 rays for production quality. Anything over 1,000 rays is insane. The cost (in render time) to return (in quality) ratio just isn't worth going any higher.
What kind of GI does Poser7 have hidden in it?
Is it photon mapping?
It would almost have to be to properly implement the Monte Carlo. With the Poser nodes, you could probably add various formulae to the Markov chain's probability sampling; and have a native MTL renderer. That would be a serious upgrade to the rendering quality, for the more patient users.
Maybe they just didn't get all the bugs worked out in the rush to put out a new version. That always seems to be the case with software.

Thanks bagginsbill, that was very informative, I appreciate your help!  So a Gather node would'nt work if there were no actual lights in the scene?

*"with a falloff based on distance. So it works like reflection. "
*What does Falloff do for a reflection, can it improve render quality, where is that controlled?

well, anyway, keep us posted on how to activate GI from the python interface. maybe that'll be a feature of sr4.

p.s. increasing the irradiance to 100% actually seems to speed up the "gather" render. increasing "gather" samples to 8 also helps, so it may be analogous to photon mapping. a quick render of the poser lo-res ball (with simple_colour node in both diffuse and alternate diffuse) above the ground (with gather node in ambient) showed that the floor demonstrates the "gather" effect with one lite at 0.000%. after deletion of the lite (no lites in scene), the floor still demonstrates the "gather" effect.

Miss Nancy - May we see the render, pretty please?

This line of GI discussion should probably be another thread, as IMO, it deserves it's own topic as a very important (maybe the most important of all) photorealistic feature that is currently missing from Poser rendering.
Anyway...

Note that It isn't required that a surface be able to generate photons in order for there to be global illumination: Surface photon generation is a shader quality, GI is a property of the overall light model.
A simple standard test for CG light models is to set up a cornell box and observe the render.
Maybe someone with Python script experience will run a CB test or two with Poser's hidden GI feature?

The sphere is not generating photons, unless I'm mistaken. If the sphere were generating photons, then the ground could be a simple diffuse white and be affected by the sphere emmissions. The ground is pulling diffuse color from the sphere. Different concept altogether than GI.
The 'Gather Experiment' here reminds me of Bryce's True Ambience, with the addition of surface shaders. It definetly adds photorealistic qualities to renders; but isn't GI, and it's computationally expensive.
TA (NOT T&A) accounts for specular reflectance also.
Does the gather node pull in specular responses of a nearby surface?
Thats another piece of eye candy that adds realism to renders.
Take it for granted that it would probably increase render times to do so.

Ouh, thank you Miss Nancy, this is very nice!  I wonder why it is purple, I do'nt see anything purple, connected to the PoserSurface node?

Hmm, bagginsbill, what do we think?

angel, it's purple because the poser lo-res ball's got a simple_colour node (purple) plugged into both the diffuse and alternate diffuse channels. as mentioned by dv, it's not actually illuminating the ground, but rather the ground is pulling colour from the ball.

I can see where the inverse square thing is something I should keep in mind.  I'm really concerned about the diffuse lighting acting like specular lighting, though.  This just has to be a bug, and needs to be fixed.  I'll venture that it is the singular reason Poser renders can be so easy to spot to a trained eye!  Certainly I've fought against it without ever realising just what was amiss.

Indeed. And ironically, the specular lighting almost always SHOULD exhibit decreased reflection when facing the camera - i.e. the Fresnel effect. And yet it doesn't - we have to add it to the shaders over and over. In the Modo forums they say "fresnel makes everything look good".

I hope someone finds a way of unlocking the hidden GI....i will be keeping my eyes peeled

Baggins... a thought occurs to me. Rather than using the fresnel effect on all the object materials, in order to give more accurate specular effects; could it not be pluged into the specular channel of the lights themselves? By making the specular channel of the light behave properly, then it would save having to do it to each and every object we find in our scene.

Just a thought, and I realise it's entirely possible that this wouldn't solve anything.

JonTheCelt

Jon,

Absolutely - you could probably fix a lot of scenes by putting fresnel on the lights.

Since we've discovered that the diffuse light is inverse fresnel (i.e. it gets brighter towards the camera and darker away) then fresnel on that would improve things.

And so if you want fresnel on the diffuse (to cancel the inverse fresnel that isn't supposed to be there) and also fresnel on the specular (because its missing) then just stick it on the light intensity.

Of course, if you want to control the fresnel and anti-inverse fresnel separately, then you should put one edge blend on the diffuse and another on the specular.

Either way, I bet this would give you a lot more control that could be put to good use.

However, a word of caution. If you are using materials that already properly include fresnel on their specular response (like any of my shiny materials or my new skin shader) then the addition of fresnel on the light itself will accidentally square the effect and produce undesirable results.

My head a splode...

Well, I haven't messed with the noodles enough to get to that anyway.  I've played with bump, and displacements (Damn, every material that looks good has values in the <.2 range...  Shouldn't they recalibrate those?) environments and ambient...  Pretty much mastered those, and moving to math functions and procedurals next.  It wouldn't cause me too much trouble to put fresnel (how's that pronounced anyway?) on the lights, though I'd never have thought of that by my lonesome.  Things is, originally I had some crazy fantasy of foregoing rendering altogether and just running some filters over the preview output. Now I'm wondering if the enhanced hardware preview will render the corrected lights properly? (I've temporarily  given up on that partly because I like bumps and displacements, but mainly everything looks too floaty w/out shadows.  I'm seriously hoping for GLSL bump and parallax displacements with shadow mapping in P8's preview)  

While I'm on that topic allow me to post my first ever image to the forum:

Yeah, that's just one polygon.

That is correct. Also, forgot to add that the last pic uses bagginsbill's inverse square node together with the Stewer's GI script - putting those two things together makes a big difference. take care jdc

Oh my oh my - this is getting interesting. The GI actually is a big improvement in this scene. I"m noticing so many differences that are meaningful. 

Could you do more? If done right, then bright "sunlight" coming in the window should light everything in the room. You probably have to raise the infinite light so much that it blows out on the directly lit things, but that happens in real life too when you set you shutter speed and aperture to pick up the indirectly lit elements of the scene.

Here's my continuing saga with the Great Room setup - I'm very nearly at a point wher eI might slip a figure in and render it off as a finished image (how exciting, eh?)

I switched from point lights to spots, and set them to a 180 degree angle. When I did this, I noticed something that's worth keeping in mind for other users of this script - you need to set both angle start AND angle end to the same setting - don't keep angle start at the '0' level. If you do this, then the lights don't work properly. I would imagine this is because the light material now controls all the falloff, so combining the linear fallof of angle start/end with the inverse square falloff of the material causes curious things to happen... on the other hand, that might not be it at all: I'm sure Baggins can exaplin this better than I.

Removing the point lights meant that the ceiling was no longer lit at all which was unrealistic - in the real world, due to GI, the ceiling is lit indirectly by the light bouncing off all the other surfaces. To compensate for this, I disabled the floor's "cast shadows" option, and set an infinite light pointing straight up at 50% intensity to act as a bounce light. I turned the bounce light's shadow setting down to 0.3, and its shadow blur radius to 20 (the highest you can get on raytraced shadows). Without any furniture in the room, this is what you get:

I like this, but at the moment, the ceiling looks incredibly flat. So I added in some furniture, and increased the render settings (minimum shading rate down to 0.5, and bucket size down to 32). The following render took about 6 hours to complete (I am SO upgrading in the next month or so - only on an AMD Sempron 2300+ with 512 RAM at the mo):

Looking at this, I'm inclined to think that perhaps the ceiling is a little bright - in the next render, I might bring the intensity of the bounce light down to maybe 30-35%, instad of 50. I like the soft shadows projected up there from the furntiure, though, especially the subtleties captured in the dome (which looked horribly flat in the previous render). OH, yes... the only other material I dickered around with on this was the glass setting - I used BagginsBill's fresnel glass shader on the windows and door glass, in order to give the reflections in the glass and the ground/skydome outside. Other than that, all the materials are as they were in the original Great Room package, and I'm really pleased how they've come out. I love the reflections in the poished wood floor, and the edge effect on the sofa works beautifully, too.

The next step for em here, I think, is to get the ceiling looking right. For the moment, I think I'm going to forgo the GI script, although I have dutifully copied it and saved it away in my runtime for later use - I just don't thnk my computer has enough power in it yet to pack that kind of punch. So it's just a matter of tinkering with the bounce light's settings until I'm completely happy with what comes up.

Any feedback on these images, including sugestions on where to work on them, would be much appreciated.

JonTheCelt

in image 2, the ceiling is o.k. IMVHO. it's an excellent render. I found the GI script (sans bill's inverse-square node set-up) is causing artifacts along bucket edges that I can't get rid of, trying the usual fixes. but the GI fx are very striking and interesting to see for the first time in poser, for me at least.

Jon that looks excellent. It seems the approach you're taking is worthwhile, as that looks very very good.

Yeah, those artifacts in the GI render are a real pain - I guess that's why the feature isn't officially supported. The markings are very specific and don't change under different lighting conditions, so it may be more of an issue with the 3D mesh. Oh well, like I said it ain't perfect but it might be fun to tinker with. Hey jon, that room is looking good. later jdc

Thank you MN - very interesting. Hmmm. Hmmm.

Is there a way to use this gi script to light a scene without using any lights at all? Something like in Lightwave.

it almost appears that AO was on. there are several more variables that may be of use in the script, but I dunno what the variables control:

        useRenderer p5
        settings 
                {
                shadowRenderShadingRate 16
                hairShadingRate 8.000000
                useSumAreaTables 0
                rayAccelerator 0
                occlusionCulling 0.000000
                maxError 0.500000
                maxICSampleSize 16.000000

Could be, I'm still pretty new at this, Thanks for your help. jdc

Quote - Jon,

Absolutely - you could probably fix a lot of scenes by putting fresnel on the lights.

Since we've discovered that the diffuse light is inverse fresnel (i.e. it gets brighter towards the camera and darker away) then fresnel on that would improve things.

And so if you want fresnel on the diffuse (to cancel the inverse fresnel that isn't supposed to be there) and also fresnel on the specular (because its missing) then just stick it on the light intensity.

Of course, if you want to control the fresnel and anti-inverse fresnel separately, then you should put one edge blend on the diffuse and another on the specular.

Either way, I bet this would give you a lot more control that could be put to good use.

However, a word of caution. If you are using materials that already properly include fresnel on their specular response (like any of my shiny materials or my new skin shader) then the addition of fresnel on the light itself will accidentally square the effect and produce undesirable results.

So much info...  Fresnel cancels inverse fresnel, does it have to be increased then I wonder...

bm

Anyone else have Poser hang at render-time with this GI script?  Renderbar just won't move.

That could be down to the fact that GI renders in Poser take approximately forever. 

it's not practical to attempt large renders with IL enabled and hi-quality settings
if scene contains complex poser figures.  all known desktop computers will be
very slow.  enabling depth-mapped shadows and/or AO (in lite properties) may
cause the machine to freeze/crash/slow to a crawl.

Ah.  Sigh.

I tried out the script, but I'm having the same problem.  It seems to make Poser lock-up just when it would start rendering.  Either that, or it just takes a ridiculously long time and I just got impatient.  I did make all shadows ray-traced.  Any ideas?

It takes a ridiculously long time to render.  GI calculations are quite intensive even for software that's designed to use it - the GI in Poser is an unsupported hack

bookmark

the developer used OS X, hence maybe these variables are in a more stable environment
in OS X than in windows.  I can usually finish a poser GI render, but it's mind-numbingly slow.
however, if ya jack up all the GI settings in Carrara to the max, that renderer is also real
slow.  that's why 3d users take various shortcuts when they're on a deadline.  it can look just
as good without all the settings at the max, if one follows the various lite/shadow/shader tips
posted here by baggins et al.

in movies they actually dont use GI a lot of times. i was shocked to found this. ok not when i found out how long it takes to render.

pixar is stylized. but did you know that they are using in some shots like 100 lights?

That's how they fake GI - take a look at [Digital] Lighting and Rendering some time - gives lots of good ideas about CG lighting