autoVino's bat projects

[Pat]

Auto that is looking real good there and any progress at all per chance?

[mmorales2]

Is there a place to download your creations because these are very nice.

[autoVino]

sorry guys I've been away for these last few days and for a few more days.  I'll probably have to start again on the weekend of the 27th.

[autoVino]

I did some renders, but nothing new so far.

[art128]

very nice progress my friend 

[autoVino]

I have some considerable progress today: just the modelling basics:


I noticed that the middle of the building seemed a bit dark.  There are two ways to fix this, adjusting the lighting, exposure settings, etc or using GI or Caustics.  I chose caustics because it would produce a more realistic and interesting result.  I went ahead and did a low quality render with caustics:

and got major artifacts.  Thinking this is a quality issue I increased the sample size to one million and density to 100 photons per sample:

Same problem but it is more concentrated.  I tried to convert the scene completely to mental ray, there wasn't any change (that is I chose a mr physical sky and imported the lighting rig into the scene).

This is where I started doing some research on what the cause of the problem is.  I came to the conclusion that there are a few select samples that don't decay correctly, that is their decay rate is d^0 or none at all instead of the physically correct (and set) d^2.  This means that those samples will be at the same brightness as their source, the sun (which they are).  I played with the decay rate (and set the quality down) and got very good results.  A decay rate of 1.9 works really well, but is a bit too bright.  A decay rate of 2.1 is right on the money though.

In the process, though, I changed the glass material slightly so that it would be realistic glass, which is the reason why the artifacts went away.  When getting the glass back, the problem came back:

and a high quality:

what we have now are just some christmas lights.  It's funny how some of the artifacts are not only white, but orange.  Other views produced pink, red, and green artifacts as well, so my decay theory is probably wrong.  I'm going to do some more testing with just the scene materials and castics and see if it isn't just some geometry problem (such as overalapping faces) that makes artifacts.

Anyone have any other ideas?

[SimFox]

what sort of sky are you using in your rig?
Where is it placed? in sync with sun, or straight above? All shadows do look a bit to intense. and overall the feel is a bit odd. On one hand there are clear sharp shadows, on the other hand the sunny side look more like during overcast day. Just take a look at the white paper of building schematics. even in full open sun is sort of gray, too dark. That's not right. If you're using MR Photographic try to lowering dark multiplier that should lighten image quite a bit.

Christmas lights are caused by either GI or FG  hitting overly bright spot and not averaging out properly (too few samples? too small interpolation?)
What is the problem, with your glass? You did give it some defuse color, right? I haven't been using Caustics for a while and am a bit rusty on them. But about decay rate... well you can be physically correct, or you can try to make it look nice. It isn't always possible do both. Particularly given relative physical correctness of MR in the first place. Decay rate of above 2 isn't correct anyway. it should be 2, but with modle like this you may want to lower it to say 1.8 or even 1.6 to get more lihgt into the dark spots. You also may wat to use some trickery with FG bounces, but replacing whatever is proper diffuse color with off-white ray switcher. You can read more on this subject here:
Jeff Patton Blog

at any rate it would be a bitch to light up, or to be precise to expose properly. It will all come done to creative color/tone mapping as the natural illumination differences are so big. Our eye in real world scanning what we see adjusting "exposure on the fly and then our brain recreated the entire field of view from those hundreds exposure corrected elements. So we see relatively well in both shade and open sun. But that is a trick of our brain, not sight. Here you should be a brain and try to trick us...
For some ideas you may refer to the photography and the use of reflectors in it. You need to bounce some light into the yard.

[art128]

this tower is very impressive  really, good job !

[autoVino]

this is a mix of the ies sky and mr sun.  Probably by brightening the ies sky it can look better, but that's not the main problem for right now.
The main problem right now are the artifacts, which are caused by caustics, as far as I understand.

Without caustics the render is fine.  Adding any sorts of caustics adds artifacts (and it always does for some reason when the geometry begins to become complex).  I've attempted to replicate the error in a simpler scene with similar materials, overlapping faces, overlapping normals, etc.  It didn't work, caustics worked like they should each time.  I agree about the correct decay rate should be 2.0, but maybe something brighter may even be better.

In other renders with the same problem, it's funny that it's not one point that causes the caustics problems, but the problem is distributed (usually) around a larger area.  More samples make the problem more defined and concentrated, but the question is why are some points so bright, while others work fine?

[callagrafx]

Not sure I've ever seen a rig with IES sky and MR sun....Surely they aren't designed to interact together properly?

[autoVino]

maybe, but it makes it a bit easier to control the lighting in the scene than when working with the mr-sun-sky setup and is more realistic than an ies setup. 
Maybe the mix is the root of the problem?  I'll check it out.

[Pat]

Hey Auto that building is looking real good there!!! Also a biiiiiig Congrats on joining ITC!!!! 

[Earth quake]

Wow, just splendid, Fantastic work.
Welcome to ITC team. 

[art128]

congrats for the enter in the ITC team my friend 

[MandelSoft]

Welcome to the ITC Team! And nice BATs anyway!

[autoVino]

thanks.

I haven't found the root of the caustics problem, it may be just a random error that arises with speed optimization with the increasing complexity of geometry.  It seems that this error cannot be replicated with simpe scenes and arises with complex scenes.

I have found a half-fix to it:  But lowering the decay rate (as a result caustic shadows and reflections are more intense) the artifacts aren't as large, but they are still there.  By increasing the sample size, density, and number of samples (to about 12mil, with 120 samples per .25msquared) the render seems fine.  It has very sharp caustics (something that can probably be changed by changeing the filter) and very intense caustics.  There is some noise visible, but this can be hidden by addition of more detail.  My only worry is that with more geomtry and on different rotations caustics may react in an unexpected way.

Here are the renders (normal)


current (physically incorrect)


Is it an improvement?
Are the caustics too bright?
Is the noise too distracting?
Are the caustics too sharp?
Are the caustics too distracting against the detail?
How is the lighting now?
My answer would be yes; yes; yes; maybe; yes; it is better, could be brighter.
What do you all thik and what are you all's sugestions?

[Earth quake]

Oh my god!!!! 
This fantastic, really splendid. 
Go on, your excellent work. 

[callagrafx]

I think Simfox is right in that you should try bouncing some light into the centre rather than fight with caustics....

[SimFox]

Another thing you could and probably should do is to add large ground plane. It would bounce great deal of lignt back into the scene - even you roof-tops will get brighter, but particularly the shaded parts, although the courtyard will still likely be a problem without some cheating

[Simpson]

Wow this bat is absolutly splendid and so réalistic