Exterior with pool

[Kram1032]

ssf only slows down the render, if you run out of RAM....

Max Num Consec Rejections (MNCR) doesn't speed up or slow down your render.
It simply reduces Fireflies but also caustics
in this release, 1000 is quite ok Don't use lower values. 2000 is ok, too, I guess...

But the genius, in this forum, with these values, is CoolColJ

[slamdog]

thanks all.
i wasn't thinking that supersamples had that meaning
actually the floor is modeled, and the board are slightly skewed.
I'll try to make the more realistic.
bye bye

[slamdog]

maybe last update...
changed textures with higher resolution ones, some other minor changes.
must study for exams so maybe this will be the final version. i will render full resolution for 11-12h.
some last hints?

bye bye

[Kosmokrator]

vere very nice results man.......give alast render about 20h.....and add more contrast to the table texture.....the roofs material-model is super....trees-water-env...super!!!
i like it!!!

p.s..... add some green leavs in the left as in the real photo...
ps 2....play a litle with the color of book...looks strange!!!

[slamdog]

hehehe... many many thanks kosmokrator
i will begin the final rendering tomorrow

[CoolColJ]

some technical questions...
i'm using these settings:
LM prob=0.1
Max Change=0.03
Max num cons rej=2000
Max Depth=600
SuperSample=3
low values of Maxnumconsrej speeds up convergence? and can affect the "realism" of the scene?
Other question: there's a way to reduce the render time per sample?

bye bye

don't use a LM that low, try 0.4 to 0.5!
don't use a max change that high, it will smooth faster with 0.005 or 0.003
or you can try default of 0.01

[BbB]

Great work. I agree with Kosmo, you really need to give it some time. 20h would be nice, maybe more. It will make all the difference and bring details out you never thought you would see. If you have the memory, you should even try rendering it at twice the resolution and downsample when it's done. That way you can get a super-clean final image even if the original render is not completely clear.

[slamdog]

Hi all
Here's a VERY little preview of how the final version will look like.
Rendering full resolution at 1536x1904 (the preview is after 6h of rendering)
Some details already came out, but water seems to take a looong time to clear. I'll give 20h like you said (if my laptop doesn't burn , for now it's emitting strange sounds )
bye bye

PS: i'm planning to re-tonemap the hdri with reinhard05 and do some postpro with the gimp.

[Kram1032]

uhm... how do you want to extract the hdri?
Igis aren't supported by any tonemapping prog, afaik...

you'll need to export multiple pngs with different post-scale settings: post-scale is like Exposure stops

[slamdog]

in the blendigo exporter i've checked the option that saves the untonemapped exr, then i will edit with qtpfsgui.

bye bye

[Kram1032]

ah, I see I didn't think of that

[Zom-B]

in the blendigo exporter i've checked the option that saves the untonemapped exr, then i will edit with qtpfsgui.

bye bye

Be aware of the untonemapped.exr

It seems they are the clamped output of Indigos internal 96 bit (???) Accuracy, so sometimes this 32 bit image is totally overbrigned!
The tonemapped.exr has all color informations and is still very useful for tonemapping!

[Kram1032]

96 bit? O.o
if, then 128 bit but is it really that accurate? O.o

[slamdog]

this is the re-tonemapped hdr...
no problem...only seems that the dynamic range isn't so extended...
however it does a nice job: the reinhard05 tonemapping is quite good... maybe better than the indigo camera (I know... they have completely different approaches... )

[slamdog]

New update...