CoolColJ's test pics thread

Post by **OnoSendai** » Fri Sep 14, 2007 5:51 pm

Bump should work on specular.
Caustics on a pool bottom viewed from overhead are possible, they're just very slow to converge, much slower than caustics viewed form inside the water.

CoolColJ · Post by **CoolColJ** » Fri Sep 14, 2007 6:36 pm

ok bump maps are probably what I need to get that high frequency caustics effect - yeah I do remember using em in Sketchup, time to bug Fused

Ono - take a look at the pics below

rendered in Bidirectional MLT mode, Indigo V0.9 test 7

I wanted to point out some issue with Diffuse shading vs Phong with textures.
The cylinder with Phong+textures shades fine, but the other pic with Diffuse+textures shades with a hard edge, which I used to mistake for a smoothing error. If I remove the texture it renders like the Phong version
Is this a bug? Looks it

Apart from the bug, the Phong version seems to have more shading even with an IOR of 1. Is this more correct than using Diffuse for this type of texture?

UV mapped texture tiling in Cindigo is working nicely as you can see

I tiled the textures in cubic mapping mode and then assigned UV co-ordinates to the material to freeze this in place.

Used a 66,000 poly water mesh here, a bit calmer than the other.
no underwater caustics yet, and there are plenty down there, but I will keep it rendering and hope they show up eventually. Currently at 250+ samples per pixel

Kram - that's your blue Sapphire material there

CoolColJ · Post by **CoolColJ** » Fri Sep 14, 2007 10:24 pm

well well the caustics did show up in force around the 650 sample mark!!

With Bidirectional MLT, about 2 hours for me on my P4 3.2ghz machine. A Q6600 quad core would probably show them around the 15-20min mark
Rendered up to 900+ samples here, 3.5 hours and not quite finished yet. They are getting brighter and denser, and more showing up on the wall. They arrived in pretty dense formation even then. Not these little dots I was expecting. I used 100,000 bounces BTW

Ono - something strange though - there are caustics in the sky or maybe it's in the camera?!

A bug I suppose of the new V0.9 test 7 code?

I'm rendering up the other Phong version now and let it go for longer. I don't consider a render fully finished until it reaches at least 3000 samples, and usually I go much higher into the 5000-9000 range

I keep seeing new things all the way up there too.

matsta · Post by **matsta** » Sat Sep 15, 2007 1:32 am

wow the water looks great dude! dig the spere ball. the pat is very nice. only crit is that the mat on the cylinder is showing its mid line. perhaps rotate a bit so its not showing

greetz

Kram1032 · Post by **Kram1032** » Sat Sep 15, 2007 1:43 am

AMAZING!
Cool caustic clouds

It's extremely clear, already

thanks for using my material, I'm honoured

I second matsta on the UV coordinates...

CoolColJ · Post by **CoolColJ** » Sat Sep 15, 2007 6:38 am

matsta wrote:wow the water looks great dude! dig the spere ball. the pat is very nice. only crit is that the mat on the cylinder is showing its mid line. perhaps rotate a bit so its not showing

greetz

Guys it's a shading bug I mentioned above....
Look at the Phong version with the black sphere, no shading problem there

It only happens when I use textures with diffuse material objects, when I remove the texture it shades like the Phong version. I'll make a formal bug report and show it better with a scene file as well for Ono.

Now I'm rendering the Phong version with the black sphere, which uses Inpox.nk, and the caustics haven't even started to show at the 1000 samples mark...I guess there is a lot of luck involved

Still I can't believe the speed difference, there is almost no point using plain MLT anymore, as it's not much faster sample count wise and with lots more disadvantages accuracy wise. Hybrid in either version is so slow now in comparison

Kram1032 · Post by **Kram1032** » Sat Sep 15, 2007 10:43 am

ah, I see, THAT'S the actual bug

Maybe, it's the expoter? C4D?
Try Sketchigo/Blendigo

CoolColJ · Post by **CoolColJ** » Sat Sep 15, 2007 11:01 am

I'm pretty sure it's a universal bug - but I will try sketchup to be sure

now that other render has been going for 4.5 hours more, at 2786 samples, and no sign of caustics...

I have restarted a few times to get new random numbers. If nothing showing by the time it hits 4000 samples, then something is wrong!

may to have try a higher mutation probablity..render is already clean as a whistle!

CoolColJ · Post by **CoolColJ** » Sat Sep 15, 2007 11:11 am

here is photo of real caustics - it kinda looks like a render for some reason

To the left is a real life example of caustics taken at a beach in Rättvik, Sweden. The water is about half a meter to a meter deep where the picture is taken.

Post by **OnoSendai** » Sat Sep 15, 2007 11:40 am

That caustics in the sky bug is bizarre.
Can you send me the test scene please?
Also if you could reproduce it in a simpler scene, where they show up earlier, that would be good.

CoolColJ · Post by **CoolColJ** » Sat Sep 15, 2007 11:58 am

OnoSendai wrote:That caustics in the sky bug is bizarre.
Can you send me the test scene please?
Also if you could reproduce it in a simpler scene, where they show up earlier, that would be good.

sure, but first I have to find some consistant settings that will have the caustics appear all the time, right now that doesn't seem to be the case. I have only had them show once as per the above render

I restarted the renders at smaller size now and trying to find a combo of things to get them to appear all the time and earlier etc
Render pretty fast at 320x240 and 160x120

maybe higher MNCR, Large mutation probability etc

CoolColJ · Post by **CoolColJ** » Sat Sep 15, 2007 2:05 pm

ok I did several more renders with the same scene, although at very small sizes to get then to render real fast

all with different settings, and I took one all the way up to 7500 samples, and not a single one showed any caustics under the water!

That single render above must have been one hell of a lucky fluke

I'll keep trying....must have caustics..... could be a bug or some optimisation needed? I'll try again with V0.9 test 6 as well

CoolColJ · Post by **CoolColJ** » Sat Sep 15, 2007 10:41 pm

getting frustrated, I have not been able to get any caustics to show since that last render above! I even took one render up past 10,000 samples and still nothing!!!!

caustics should not be this hard to come by, Ono you may need to tweak something. I mean the above render shows it's possible to happen by 650 samples, but that was one try out of many

I redid a simliar but simple scene with a 4000+ poly water, with regular wave pattern. It renders faster and yet I don't get anything.

Ono I'll send you this scene for you to check over.
You can also check out that shading issue I mentioned before.
Left cylinder - diffuse + texture, shading issue
middle cylinder - phong + texture, shade fine
right cylinder - diffuse, no texture shades fine as well

might try this scene in Maxwell/Fry and see what I get

CoolColJ · Post by **CoolColJ** » Sun Sep 16, 2007 8:18 pm

The glass spheres render after some 20+ hours!! 3000+ samples now
Very very slow, gonna re-render it in V0.9 test 7 in larger size, the speed increase and better Specular/SSS material will help

Exit portals and this much glass is always a killer

so far it looks like V0.9 test 7 is getting all the tiny details in the glass spheres, much much earlier in a larger resolution!!

suvakas · Post by **suvakas** » Sun Sep 16, 2007 8:39 pm

Hmm..so much indirect light...nice.
That's exactly why i love unbiased renderers

CoolColJ's test pics thread

Who is online