Daily Builds 1.0 - 1.4

[Ondra]

AFAIK Vray's checkbox on = 1 in corona. 0 in corona = checkbox off. Try experimenting with it in interactive rendering

[PROH]

Well, from my tests, I would say that it acts somehow like a threshold variable. 0 = off, and all other values = on, where lower values means more clamping (0,01 = extreme destruction), and higher values means less/softer clamping. For my test scene I found that values between 5 and 6 worked best.

Edit: Just did one more test, and actually I find that value 1 is working really well when "highlight compression" is off (1). The reason why 5-6 seemed better in earlier tests, was that HC was in action too.

[antanas]

hmm it seems that interactive and regular modes has some differences in vfb msi clamping behavior http://clip2net.com/s/3vz7MeJ or is it just the difference of interactive vs regular renderer which I didn't observe before ?

[PROH]

Are you sure it isn't the denoising that makes this difference? (Render with denoiser vs. IR without).

[Nekrobul]

I found that VFB clamping alo kills all "heat" information in Open EXR. Is it somehow pausible for lights and light mtls to be excluded from this proces?

I have attached EXR files for more noticible difference.

[Ondra]

I found that VFB clamping alo kills all "heat" information in Open EXR. Is it somehow pausible for lights and light mtls to be excluded from this proces?

I have attached EXR files for more noticible difference.

unfortunately it is exactly because it clamps lights that it also solves the AA problems

[Nekrobul]

I found that VFB clamping alo kills all "heat" information in Open EXR. Is it somehow pausible for lights and light mtls to be excluded from this proces?

I have attached EXR files for more noticible difference.

unfortunately it is exactly because it clamps lights that it also solves the AA problems

So it is more like a debug for AA problem scenarios?

[Juraj]

Vray works identically. To retain full dynamic range, they both (sub-p, and clamp) have to be off, and only MSI can be used.

[racoonart]

Hmm, looks like the "Adaptivity recalc passes interval" spinner doesn't do anything, it's still using 10, no matter what is set.

[romullus]

Hmm, it works for me.

[Ludvik Koutny]

Hmm, looks like the "Adaptivity recalc passes interval" spinner doesn't do anything, it's still using 10, no matter what is set.

There's a problem that the pass value for very first interval is hardcoded, only following ones respect the value. :/

[pokoy]

After a long production phase I've installed the latest daily and try to catch up on latest features.
One question - I remember reading that more than 255 lights are now sampled efficiently but the log still warns about the slower fallback algorithm. What's the current state (build from Mar 12 2016)?

[mitviz]

no sure if its the correct thread to post in but i though the denoising happens while the render is going on, i see now that when i end the render it takes a few seconds to denois, also the save all render channel but stil doesnt work sometimes, so weired

[Nekrobul]

no sure if its the correct thread to post in but i though the denoising happens while the render is going on, i see now that when i end the render it takes a few seconds to denois, also the save all render channel but stil doesnt work sometimes, so weired

Denoising only affects beauty and alpha, so far i gues.

[maru]

Hmm, looks like the "Adaptivity recalc passes interval" spinner doesn't do anything, it's still using 10, no matter what is set.

Seems to work fine.
But I still didn't find a way to get rid of "black holes" in some cases. So basically if there is an empty area, the sampler will treat it as noise-free and won't refine it at all (at least that's what I think happens). My guess would be that disabling the adaptivity completely every X passes could get rid of this problem, so we would basically intentionally introduce some little noise at some point to retain some details. Does it make sense?

(a1 = re-calculate every 1 pass, a50 = re-calculate every 50 passes, etc)