Freakin’ Awesome !!!
Only dancing pony’s I could find 
what about
? 
VCM is not unbiased
why do you say that?
anyway, i just want to test it very very very soon, i need it :s
he is right, it is not unbiased, but consistent - it means that when rendering the bias eventually goes to zero with time.
i see , but it’s doing a great job here, maybe you should make a 2 hours render of a full test scene with VCM so we can see better.
I see you don’t have those funny white splotches, Do you keep “use merging” off ?
FYI VCM with “use merging” off equal regular bidir
I see, you noticed those splotches I am talking about ?
Also edges get over exposed as soon as I used an environmental light, Coronasky or Hdri
Actually both Bdir and VCM are doing great with direct lights but it gets very noisy and slow if I use environment lightning.
Chakib: Saying that because it is (I am one of the method’s authors). But it doesn’t really matter, as you want low noise, not some math buzzwords 
grab the newest build, there were some bugs in older ones (including incorrect environment weighting). If the problem persists, post screenshots. Also you usually need to tweak the search radius parameter to get good results (there is tradeoff between blurring and fireflies)
does it include the legacy build ?
*(there is tradeoff between blurring and fireflies) → But both are resolved and gone in time, right? Anyway, would also like to test in legacy build if you find time…
Tested yesterday, it’s working very well with glass/refractions/caustics, but antialiasing seems to be less effective than standard progressive path tracing.
I use the latest builds
It’s easy to replicate, you can use your cornell box scene, turn off direct lights and turn on environment light, you should see those bright edges
It can help you to identify the problem: it happens only with geometry without thickness, if I add a shell modifier to my cornell box, the problem is gone, but there seems to be a relation between the thickness amount and the VCM lookup radius.
Hi looks like vcm need more time to render less noise or its more for night lighting, how do i download the alpha version with vcm??? or sam you are a tester for that version that will be released??? thanks i aprecciate any answer.
too much noise…
There is no glass and metal in your scene, can you add some ?
If the scene contains no caustics (or even no glass/mirror), then VCM is overkill. It will do lot of heavy lifting to solve problems that are simply not there.
For examples of scenes it alone can handle have a look http://www.smallvcm.com/ (the complicated scenes from paper).
I would look at it as method that opens new artistic opportunities, rather than speeding up the current ways of doing stuff. (At least, that was the hope when we came up with the method)
I am afraid there is no silver bullet (yet).
I got some ideas for time-adaptive algorithm, that is pretty much unpublishable, but could make production guys happy.
(Will bother Keymaster with it once his deadlines are gone).
In the meantime, I will have to re-check with him today, but I am pretty sure there is still a bug in his implementation of environment lights. These things are damn hard to catch, especially since the example implementation uses different way of envrionment light (and I am quite sure still has bugs of its own).
Good luck guys your doing great job