Page 1 of 1

MLT Sampling Method

Posted: Tue Sep 20, 2011 10:04 am
by ieatfish
If I start two renders and let them run for exactly the same amount of time, will they be the exact same image in the end? Or is the sampling random and I will instead end up with equally resolved images but not exact copies?

I ask because I am finding that (probably due to my poor lighting technique) I have some pretty grainy areas even after quite a bit of render time and was thinking perhaps I could blend two renders to get a more clear render.

Re: MLT Sampling Method

Posted: Tue Sep 20, 2011 10:07 am
by PureSpider
It depends on if you use the same random seed.
If you use the same seed, both images will eventually be the same, if you use different seeds, they will be different even after the same rendering time.

Re: MLT Sampling Method

Posted: Tue Sep 20, 2011 10:09 am
by ieatfish
PureSpider wrote:It depends on if you use the same random seed.
If you use the same seed, both images will eventually be the same, if you use different seeds, they will be different even after the same rendering time.
Where does one set the seed?

Re: MLT Sampling Method

Posted: Tue Sep 20, 2011 10:10 am
by PureSpider
You need to start Indigo (either GUI or CLI) from the command line with the -seed parameter set to some number.

Code: Select all

indigo.exe -seed 12345
for example

Re: MLT Sampling Method

Posted: Tue Sep 20, 2011 10:51 am
by galinette
More simply, if you do not bother with the seed, and if your number of threads is the same (because threads use different seeds), at same spp, images will be identical

Etienne

Re: MLT Sampling Method

Posted: Thu Sep 22, 2011 3:09 am
by Meelis
If you have grainy areas then you are not even close to final image.
In that case different methods have resolved image to different stage.