Fireflies

[contegufo]

Hi

The presence of curtains at the windows (Diffuse Transmitter) and frosted glass can help eliminate the fireflies?

Thank

[Oscar J]

Just use arch glass and a SS value of 4 or 5. Should make the fireflies disappear. If that doesn't help, use BiDir MLT.

[Zom-B]

For curtains blend DT with diffuse. I your Curtains have a mesh volume use 1.0 gamma of DT, if it is single sided use 2.2.

Portals help rendering faster, but also tend to produce (way) more FFs from Sun or HDRIs with sunlike areas...
For some interiors I prefer to render without portals because of this issue.

Again a rendering would help us a lot!

[thesquirell]

I think it can only generate more! Sorry for hijacking the thread, but this question is somehow related..Zom-B, where did you get, and why, this gamma difference in relation to mesh volume/single sided curtain? What's the system behind this gamma difference? Also, read it some time ago that exit portals don't work well with BiDir MLT.

[Oscar J]

I don't think exit portals work that great at all if I'm honest. They give more fireflies, and if your windows are big, it often converges slower than without exit portals. The indigo team were talking about improved exit portals for 3.8, though I'm not sure what happened to that?

[Zom-B]

Zom-B, where did you get, and why, this gamma difference in relation to mesh volume/single sided curtain? What's the system behind this gamma difference?

I get that simply by testing. DT Material also has no (!!) diffuse reflection, so it behaves a little strange...

I don't think exit portals work that great at all if I'm honest. They give more fireflies, and if your windows are big, it often converges slower than without exit portals. The indigo team were talking about improved exit portals for 3.8, though I'm not sure what happened to that?

Yes, EPs have a lot room for improvement, and afaik the last approach to enhance them simply had some issues and maybe was put into the waiting Q since v4 development is prioritize.
I think EPs are super important tool to enhance renderspeed, and would love to see them optimized!
This recent paper here looks promising, without me understanding anything of the text ^^
http://www.cs.dartmouth.edu/~wjarosz/pu ... ortal.html

Cycles already implemented it: https://developer.blender.org/D1458

[Polinalkrimizei]

Zom-B, krass was Du so nebenbei aus dem Ärmel schüttelst!

@Glare: This info (suggested gamma-value depending on mesh thickness) belongs in the manual!

Also interesting paper (judging by the pictures ). If anyone cares to experiment: the example scenes are available for free at blendswap (user: wig42), funny that Disney research uses this resource and probably also Blender...

[Zom-B]

Zom-B, krass was Du so nebenbei aus dem Ärmel schüttelst!

Its kind of a hobby of me to stay well informed about the academic research done regarding raytracing and related topics.
The fun fact here is that I usually only understand ~10% of this stuff... I usually like the pictures

@Glare: This info (suggested gamma-value depending on mesh thickness) belongs in the manual!

I talked to Ono regarding this, and keeping Gamma @ 2.2 is the suggested behavior, also for volumetric meshes. I just noticed that for my artistic point of view, gamma @ 1 creates more predicted results like I noticed with that guy: http://www.indigorenderer.com/forum/vie ... 41#p131341
Also gamma @ 1 is not perfectly right, but easier to remember and close enough ^^

Also interesting paper (judging by the pictures ). If anyone cares to experiment: the example scenes are available for free at blendswap (user: wig42), funny that Disney research uses this resource and probably also Blender...

Yes, such papers always try to shine with good examples where their new magic code has the best possible output.
I also have zero clue about if it is easy, hard or impossible to include such stuff into Indigo, so take it with a grain of salt...

[thesquirell]

I would love to jump on that hobby wagon and read up more of those papers! Where do you get them?

[Zom-B]

I would love to jump on that hobby wagon and read up more of those papers! Where do you get them?

A great source always is this here: http://kesen.realtimerendering.com/ke-s ... gelog.html
On the ompf2 Forum you'll find some too, and people do discuss them there as well: http://ompf2.com/

But true pearls lie hidden on the web in some publication archives of universities etc. ^^

If you are interested I can spam you by mail with the latest of my findings, just pm me your mail

[thesquirell]

Oh, that's brilliant! I'll PM you my email, so I can read up! Thank you very much!