3000 s/px and hotpixelremove ratio comparison .

[patryn]

Dear All,

after than i was explained that DL is far less realistic than PT or PMC , i decided to make a fast test bench .
as you can see there is an ugly scene ( done by me) with 3 ligh emitters (one windows and two lamps ) . the windows transmit the natural dayligting and the two lamp are base on blackbody emission property .

after a render about 3000 s/px i decided to export pictures and try several value in the imager for the hotpixelremoval tools .

hereafter you can find the result .

here are my conclusions :
Hot pixel removal bring a blur that kill sharp in the picture . don't like it .
i didn't see any big difference for these value .


But i'm asking:
Maybe a postprocessing is mandatory ? but which one ?
what is wrong in my scene ( texture, lightning ?...) to not have a clean render at 3000 s/px step ?
maybe 3000 is not enough for my scene , what are the clue in order to predict it ?
PMC maybe not the best choice . PT will be better ? if yes why ?

sorry for the amount of question , but octane made me very enthousiastic and when i see the other people work , i want to try reach at least a small part of their gift . gallery made me a badly jealous guy...

thanks in advance and sorry for my bad english .

Pat

[kubo]

usually, but this is not always the case, you use PT when the light source is quite direct on the scene, in those cases (in your scene) PT solves it faster and cleaner, but when the light emitter is hidden and most of the light in your scene is indirect, then PMC works best and solves the scene with less samples than PT (and some times PT won't solve it no matter how many samples).
The number of samples needed for each scene is unique, but keep in mind that with only sunlight in PT you need a very low amount of samples (usually) the same scene but with an HDRI needs little more samples (lowering the gamma makes the fireflies go away faster), add lights and the number of samples needed go up exponentially, and if those lights are "hidden" from view and are shedding indirect light then switch over to PMC 'cause PT will probably not solve it at all.
The great thing about octane is that you don't need to "finish" the render to see if it worked. Try it, see which one is doing a better job, get the feel how many samples you are going to need and "cook" the rest of the series accordingly.
Hope it help, cheers and welcome ;P

[patryn]

thank you very much to just save me about 10 years of frustation trying to understand what you just clearly explained to me !

thanks you again !

i think that you were right dude.
please find hereafter a picture rendered with PT kernel, 3000s/px , one light ( portal surface ) 0.15 hotpixelremoval , only livedb material .

[ribrahomedesign]

hi patryn

nice render the last one . my tip : render the same image at 3x your normal resolution and then downscale it .
your image will clear in half the time.
cheers
Rico

[patryn]

thanks for the tip , but i try to understand well the setting in order to have the render as much realistic and clean as possible without post production . and if PT and PMC are obviously more realistic in light treatment , i believe that materials and texture quality are now the next big milestone in order to try to reach great render quality . in the picture above , i believe that my ceramic texture ( issued from livedb and modified with a little turbulence pattern in order to remove the repeated tile effect ) is a little too glossy for my scene .so fireflies occurs . i have tried my first material macro yesterday . What a powerful but complex tool ....

[patryn]

here is a 16000 s/px render using only one portal light emitter . the render was almost 7 hours long . please find hereafter a picture with none hotpixelremoval value . as you can see , there is a poor quality and a lot of fireflies remain .
any clue why the result is so poor even whith PT rendering during 7 hours . ???


the second image is with max hotpixelremoval , the blur is ugly and lot of artifact remains ...grrr the path is a long way ....

[MaTtY631990]

Try using pmc on this render as it will help with the fireflies. Using these settings might help in this case, otherwise you could freeroam with
them.

Exploration strength - This setting try to find new paths to help compute difficult areas better in the most difficult scenes.

Direct light importance - this determines what areas are to be sampled the most in the scene. Set to 1 for the most direct illumination to be sampled, in your case this would be the sunlight on the walls. Setting to 0 would mean the indirect light would be sampled the most.

[gabrielefx]

Try using pmc on this render as it will help with the fireflies. Using these settings might help in this case, otherwise you could freeroam with
them.

Exploration strength - This setting try to find new paths to help compute difficult areas better in the most difficult scenes.

Direct light importance - this determines what areas are to be sampled the most in the scene. Set to 1 for the most direct illumination to be sampled, in your case this would be the sunlight on the walls. Setting to 0 would mean the indirect light would be sampled the most.

I always use pmc but believe me, Arion has the best antialiasing (for 995€), no fireflies. I hope that Refractive will develop new algorithms to reduce noise automatically in pathtracing.

[patryn]

you're right .

here a picture done with arion (wich i'm testing too) .but arion memory management is weird . it crash often when the scene became complex . octane is far more stable in order to load heavy scene and video memory management .

Matty.

thanks for the tips i'm trying it right now .