Now is xpdev doing space station? No, but with the small arc of intense light coming from the ground (pretty much), his scene is closer to space station than snow-covered mountains.

Now I have correct light levels for a nice exposure (as shown by the light meters) but my artifacts are really bad.

You just can't do this picture correctly in Poser.

This is the IDL precalc - observe the density of the IC samples points.

I will post the settings next.

The important ones are the four together: IDL Intensity, Bounces, Samples, and Irradiance Cache.

yeah, and as Poser doesn't do atmospheric scattering, results will be like space station anyway :-). And indeed, whatever high value you put into the render settings, Firefly insists on making splotches in the right-angle corners. But as said, Poser can come close at reasonable resource usages (RAM space, CPU power, wallet power, user time).

Quote - yeah, and as Poser doesn't do atmospheric scattering, results will be like space station anyway :-). And indeed, whatever high value you put into the render settings, Firefly insists on making splotches in the right-angle corners. But as said, Poser can come close at reasonable resource usages (RAM space, CPU power, wallet power, user time).

Totally agree.

In my scene, the light fixture is a scaled Poser box, with Ambient_Color = white and Ambient_Value = 10.

No other sources of light are in it.

Quote - yeah, and as Poser doesn't do atmospheric scattering, results will be like space station anyway :-). And indeed, whatever high value you put into the render settings, Firefly insists on making splotches in the right-angle corners. But as said, Poser can come close at reasonable resource usages (RAM space, CPU power, wallet power, user time).

I agree.  Most users won't care, or notice.  I've read so many people suggesting that Poser's Firefly can create images as good as Luxrender or Octane, so the flawed IDL implimentation in Firefly seems to be "good enough" in many people's eyes.  The problem technically seems to be inaccurate interpolation sampling, and Poser does not offer the exposed settngs to refine this beyond the overly simplified render settings in Firefly.  There should be far deeper control over the irradiance solution exposed in the render settings. Certainly in the "Pro" version of Poser.  The render settings available in PoserPro need a much deeper level of control exposed to the user.

more and more interesting discussion

depressing.  i've been waiting all night for an idl render with Snarly's metals in the scene 

that cornell box is interesting.

found the page with the photo of the box http://www.graphics.cornell.edu/online/box/compare.html

has anyone tried the box in lux?

lux is the same thing as reality?

Quote - depressing.  i've been waiting all night for an idl render with Snarly's metals in the scene 

that cornell box is interesting.

found the page with the photo of the box http://www.graphics.cornell.edu/online/box/compare.html

I like the difference image.

Quote - has anyone tried the box in lux?

Do you mean anyone at all? Google image search says yes, somebody has tried it.

Quote - lux is the same thing as reality?

No - Reality is the scene editing/conversion utility to take a Poser scene, convert it to Lux, then render in Lux. Lux is not Reality, same as Reality is not Poser. They're all pieces of a workflow involving all three.

Not only has the cornell box been "tried" in Luxrender, it's been taken to a whole new level, with volumetric light, caustics, refractive materials, and the works. I'd like to see that attempted in Poser firefly, and see what the render times are by comparison.  Poser doesn't support physical caustics, so we'd have to forgive it that, and some other things will not be possible as well.

** http://www.luxrender.net/forum/gallery2.php?g2_itemId=14532**

I would like to see more on this discussion. I am playing with BB's render settings right now.

wheee. downloaded a box and spheres.  

blender is scary looking, but it has a bunch of render engines to test render in.

is it a better test to close the box?

The photograph of the real box used as reference is not closed, so it's not better to close the box.

If you're thinking about all the many times people here say that IDL won't work well with a missing 4th wall, they're wrong. Ignore that.

Quote - The photograph of the real box used as reference is not closed, so it's not better to close the box.

But it is in a room, and in the link you posted the ceiling does appear to be visible at the top corners of the photo.  Is there a discussion anywhere online of the environment of the box?

Forwarding this link to fun test props  Thanks Mustake!

Quote - I believe that since my version is made of Poser primitives, it is OK to distribute it. So here it is if you want it. Remove the .txt extension when you save it.

kewl!  Thanks!

Quote - But it is in a room

Which by all absence of evidence on the Cornell Box data page, is not contributing to the lighting in a measurable way - one assumes that the room is obscured by a non-reflecting cloth or enclosure.

This page supplies detailed Cornell box simulation data:

http://www.graphics.cornell.edu/online/box/data.html

Nowhere in it is any mention of the factors of the enclosing room - one has to assume that those factors are 0.

Quote - and in the link you posted the ceiling does appear to be visible at the top corners of the photo.

Which link is that? The only link I see above in my posts is:

http://en.wikipedia.org/wiki/Cornell_box

That wiki article does not have a photo - it has a render from POV-ray and I don't see any ceiling, so you must be talking about something else.

If, instead, you're talking about my render of MY version of the box for Poser, it has no ceiling either. The box walls and ceiling, however, are actually thick and have a front face, and also my floor is the Poser ground which extends beyond the box, and the ground shows light spilling out beyond the box, which then reflects back on the front faces of the walls and ceiling. If that's what you're referring to, then it's not a mystery that they're lit - nor are such lit faces present in the official cornell box which by all appearances has nothing whatsoever outside its interior boundary.

Quote - Is there a discussion anywhere online of the environment of the box?

I have not been able to find any, not even in the official Cornell web pages. Given the care and sophistication that goes into their testing and simulation, I very much doubt that you're supposed to think there is any environment contribution at all. Imagine if there was a contribution and they forgot to take it into account - would they not look incredibly stupid?

Quote - Poser does not offer the exposed settngs to refine this beyond the overly simplified render settings in Firefly.  There should be far deeper control over the irradiance solution exposed in the render settings. Certainly in the "Pro" version of Poser.  The render settings available in PoserPro need a much deeper level of control exposed to the user.

use d3d's FFRender script.

Quote - use d3d's FFRender script.

You're talking about this script?

http://d3d.sesseler.de/index.php?software=poserpython&product=render

Sorry, while that's a nice script, all it does is provide a more convenient way to access the settings in Firefly that are already exposed to the user.  There's nothing in there that will help fix blotches or noise, beyond what is already available.

I'm talking about deep level of control over the irradience solution, which would require the ability to tweak interpolation, and adjust raycast parameters.

Let's take the level of exposed control for the user end in something like Vray, as an  example:

http://www.aleso3d.com/blog/wp-content/uploads/2011/11/vray-render-settings-interior-rendering.jpg

The ability to get a perfect marriage of speed vs quality is available to the end user, in abundance.  I'm not suggesting Poser users would want all this control, but we can not achieve good renders in some architectural situations, and that's problematic for advanced users, using Firefly.

This level of control is useful to achieve optimum results in most render situations, using biased render engines and GI.  Unbiased can be far more simple, but that's another story.

Quote - > Quote - and in the link you posted the ceiling does appear to be visible at the top corners of the photo.

Which link is that? The only link I see above in my posts is:

http://en.wikipedia.org/wiki/Cornell_box

That wiki article does not have a photo - it has a render from POV-ray and I don't see any ceiling, so you must be talking about something else.

Oops; sorry Ted, it was Honzu's link...

http://www.graphics.cornell.edu/online/box/compare.html

Attached Link: Next limit (Maxwell) - Photography techniques to speed-up interior renders

Quote - This discussion reminded me of an article on the Next limit site which appears to be trying to solve a similar problem (and suggestion of replacing indirect light with direct to remove artifacts) - added link in case its of interest to someone.

Yes, except that Maxwell is a physical based render engine, so some of the techniques used there might not translate well to a render engine like FireFly.  Maxwell's "direct" lights" are far more realistic, and based on area emitter physics.  The problem they are solving is how to make the GI calculations faster, and more efficient.  In an entirely enclosed room, adding direct emitter lights in the windows, which approximate the size and shape of the windows, will accurately assist the simulation and calculation time, and speed up the render.  Poser's lights aren't physically based, and the problem is a little different, although this would be interesting to test with IDL and some object emitters in the windows.  I have a feeling, without the necessary parameters to tweak the irradience interpolation and other params, this kind of physical simulation won't work in Poser well at all.  There won't be any way to clean the results.  Using the non-physical standard lights in Poser will help, but the results won't be as realistic as emitters.  Wouldn't hurt to try though.

The technique provides a good example of something that might work well in Luxrender/Reality, or Octane though.  Worth a try.

I have to admit that I completely lost track of the question between all those great answers around. Can someone rephrase the issue at hand? I'm not trolling, I'm lost.

this is what I got with bill's box.

artifacts not too bad IMVHO. did anybody else try it?

With that amount of samples and basically without caching, how long took it to render this, on which kind of hardware?

interesting.

 16 threads, are you running dual Xeon or so?

 pixel samples 8, that's 8x8 antiliasing. 3x3 would do I guess

 irradiance cache OFF ? and still artifacts "not too bad". Hmmmm.

 smooth polygons OFF ? and everything else is about high quality.

 render time? 

And I still don't know the question behind this answer. 

question: how to grasp irradiance in poser.   answer: don't use irradiance cache (IC).  machine: ordinary imac from 2009, 2 hr. render.  was thinking of getting newer faster one to watch blu-rays, but old ones are fast enuff for poser, and new macs don't ship with blu-ray drives.  avoid smoothing these primitives when possible.

Quote - question: how to grasp irradiance in poser.   answer: don't use irradiance cache (IC).  machine: ordinary imac from 2009, 2 hr. render.  was thinking of getting newer faster one to watch blu-rays, but old ones are fast enuff for poser, and new macs don't ship with blu-ray drives.  avoid smoothing these primitives when possible.

So without the interpolation (IC) enabled, Poser's IDL method is about the same as brute force Monte-Carlo GI methods in other render engines, except there are still render artifacts that can not be cleaned entirely.  Although the artifacts are minimal in your last render of the Cornell box, if one were to include additional geometry in a much more complex, but similar scene; with reflective or refractive materials, more sharp angles, and adjoining corners, the results would more than likely be troublesome, with render times not much better than something like Luxrender (depending on your hardware).  If we only had access to control the probabilistic subdivision sampling, we might be able to finness a virtually clean result from Firefly, but I don't think it's possible until additional parameters of the IDL sampling are unlocked/exposed to the user.

theory: if orthogonal joins are micropolygonal and/or microbeveled, then it would work better.  possibly related problem with poser cone (P7 or P8), which was solved by replacing tip (apex) with tiny empty polygon.  it still looked like sharp point, but eliminated rendering artifacts.

Quote - theory: if orthogonal joins are micropolygonal and/or microbeveled, then it would work better.  possibly related problem with poser cone (P7 or P8), which was solved by replacing tip (apex) with tiny empty polygon.  it still looked like sharp point, but eliminated rendering artifacts.

Hmmm. Has anyone tried increasing the subdivision surface density of the geometry, in areas with concave angles and coincident faces?  Looks like most of the images posted so far are very simple geometrically, which you would expect to render good results, but might actually create problems.  Higher resolution geometry might force the raycaster to increase sampling in areas with concave angles and coplanar faces, improving the IDL solution.  Then again, it might worsen the problem, at the cost of even more render time.

I'm just curious if it would make a difference (for better or worse) or at all.  Instead of walls and floors that consist of 2x2 or 4x4 polygons, try geometry that's subdivided to 32x32 or 64x64 at least, and chamfer or bevel the corners of the walls, allowing more geometry in those tight areas to force more sampling.

is the color walls supposed to bleed into the objs' shadows too?

Quote - is the color walls supposed to bleed into the objs' shadows too?

Yes.

Though technically the colours are bleeding everywhere and are being washed out by the Main light.

Which leaves the colour-bleed visible only in the areas the Main light can't reach directly, i.e. the shadows.

Another question.

You suggested this thing:

Quote - use large softboxes instead

 

I tried to use this technique but the problem is that "large softboxes" are visible on every reflective surface and sometimes even in the eyes of the characters.

There is something that I do not know about "large softboxes" or is there a way to hide them?

Or may be there are "large softboxes" that emit light but are totally invisible on reflective surfaces?

Many thanks

I didn't see an answer to this question. I have the same issue. If I use some large object to light up the scene I can make it invisible to the camera, but I can't stop it from reflecting in eyes and everything else.

If you haven't updated with the latest service release, you should:  there is a very significant improvement in the blotchy IDL artifacts.

This is the kind of thread we want to see more of here. Thanks. Making this a sticky.

If we can get enough others to make similar threads, I will make those sticky as well and move them to the Poser Technical forum with a stickied link page from here as a ToC, (so that all the stickies don't pile up too high).

:)

you wanna reduce cornea surface reflection of large diffuse sources?  will cause classic poser 4 "dead doll eyes" look (mostly specular), unless segregate light panels from other posersurfaces by clamp or limit diffuse.  will consider it over weekend (quitting time ~3 min. frm now), but I daresay BB already solved this.  would start with typical refl/spec math as shown below, then moderate with diffuse node. e.g. would only accept diffuse below certain limit, in case light panels had ambient > 1.0.

Artbee, Thanks for this post it has inspired me into trying

light settings I've never poked around into, Please continue !

Anyone figure out a FAST work around to fix those black splotches?  I don't want to set the samples so high that the render takes forever to finish.