Milk (regular) optimized

[Meelis]

Thanks i try it out

Now i lowered sss uniform value to 200 from 500 and did some fast hdr tests.
The noise is much more even and not like fireflies
160k s/sec vs 19k before

I gonna restart render with lycium recommended settings

[Meelis]

Yes bidir of with MLT path tracing and the noise will go away very fast, i can even use higher SSS uniform value.

Here is fast preview with HDR lighting.
Also female figures 30cm tall (1st was 20cm). And gemstones are way to big (1cm grid)

And milk material scattering is 2,5x more isotropic.

http://static3.nagi.ee/i/p/692/54/17313693dcc60f_o.jpg

[dag]

PT only at 3k spp.
Looks like 500 sss is too much.

How do you get the liquid to "fill" the glass on the sides, like in real lifes?
Used about 0,3mm space between glass and milk.

[Silmä]

There should be no gap (the meshes should intersect), but difference in materials precedence value. The precedence tells indigo how to clip intersecting mediums (eg. glass and milk).

Higher precedence wins and lower gets clipped.

[Meelis]

I guess value 200 is better than 500, in case the milk doesn't get too transparent

[dag]

In the process of finding out I was dead wrong I made this, which to my surprisement doesn't render correctly in intercrap explorer.
http://www.home.no/faenskap/3d/scatteri ... media.html

Just doing a quick testrender now to check the mat (at 4k s/s) . Woohooo!

[Meelis]

RGB scattering coefficient spectrum was very very slow, atleast in bidir mode.
Maybe i should test it in path tracing mode

And the default Henyey-Greenstein G Spectrum looked very transparent with that low (fast) Scattering Coefficient.

[CTZn]

RGB scattering coefficient spectrum (...) Henyey-Greenstein G Spectrum

>> MLT all the way

[Meelis]

RGB scattering coefficient spectrum (...) Henyey-Greenstein G Spectrum

>> MLT all the way

So with that material MLT use is enforced automatically?
Anyway path tracing with G Spectrum RGB ~ 2000 values was slow.

Uniform value 2k-3k and less isotropic scattering looks much more reall but still way too slow
compared uniform 500 and more isotropic scattering.

[suvakas]

I re-rendered my old milk test.
It's a low-fat milk

Ps. Rendering speed seems quite acceptable. (MLT, bidir off)

[Doug Armand]

Nice one Suvakas!

[Meelis]

I re-rendered my old milk test.
It's a low-fat milk

Ps. Rendering speed seems quite acceptable. (MLT, bidir off)

Looks very good and fast same time.
What's the material?

yes low-fat indeed
------

Why high absorbing material (~ 600 /1m) with non isotropic (measured) scattering (0.98) looks transparent


Here is same material with isotropic scattering (too wax like)

[CTZn]

RGB scattering coefficient spectrum (...) Henyey-Greenstein G Spectrum

>> MLT all the way

So with that material MLT use is enforced automatically?

No, that's not what I meant Meelis, and I for one wouldn't like the idea of enforcing modes. But, in terms of performances with the aforementionned parameters, MLT will (almost ?) always give the best results.

As a conclusion, saying that MLT is always slower is not true (you didn't say this neither, this was a general remark).

However it is interesting to stress the other modes with such a material like you do, optimisation wise !

Nice one suv, that's pretty fast indeed.

[Godzilla]

Finally got around to rendering that scene. I wasn't able to get rid of all the noise, I think upping the supersample level would have helped a bit more (Set at 4 here). I removed the jewels - hope you don't mind.

Small version with noise reduction in PS



Full size

http://farm6.static.flickr.com/5008/529 ... ce93_o.jpg

[Meelis]

Finally got around to rendering that scene. I wasn't able to get rid of all the noise, I think upping the supersample level would have helped a bit more (Set at 4 here). I removed the jewels - hope you don't mind.

Small version with noise reduction in PS



Full size

http://farm6.static.flickr.com/5008/529 ... ce93_o.jpg

Thank you