jwiede
Electron wrangler
Nice render, but still quite a bit of noise at 1h10m. Guessing that's without noise reduction?
Show your LW2018 Test Renders
- Thread starter luzlight
- Start date
Nice render, but still quite a bit of noise at 1h10m. Guessing that's without noise reduction?
jbrookes
New member
What happens if you lower the Roughness setting for the surfaces?
I find that a tiny bit of that parameter goes a long way.
That might smooth things out.
Schwyhart
Member
Chernoby
Member
Why in gods' names did this take over an hour?
jwiede
Electron wrangler
Semi-useful, if only because it shows how much noise is still hiding in there and where (to better direct your noise reduction efforts).
Schwyhart
Member
There was only just a tiny amount of noise in the raw render, so I removed it. Then I added noise to give it a more film look...guess I failed
One more:
Last edited:
jwiede
Electron wrangler
Well, I didn't really get a sense of "film grain" from the noise, clearly. In particular, look at top of elevator door header on the right, and top of left escalator towards the left (where it goes behind light pole/panel/whatever) -- btw, meant in _03, not _04. That kind of pixellated random (but largely monochromatic) noise doesn't really read as "film grain" for me, it reads as regular CG render noise. Maybe try adding less back, or try different types of noises to see if you can find one closer to "film grain"? Getting some chromatic "dispersion" in the noise might help it look more like "film grain" too.
pbaroque20
Polygon Folder
Some work in progress model I'm currently close to finishing. Easy rendering setup. Fiberfx works very nice in this version and appreciate. I did some quick photoshop retouching to contrast and saturation.
DonJMyers
Studio Animator
it is eggbert!
pbaroque20
Polygon Folder
lol. I added Fiberfx to the model. I'm happy to say this module is really stable on my end. Was working with it for over an hour and no crashes.
Attachments
Kryslin
Well-known member
Ok, to make up for the really bad test render I first posted, I tried applying what RH has been saying to the scene:
1) Keep Camera AA Passes as low as you can.
2) Raise lighting samples, Refl. Samples, Refr. samples, and SSS Samples first, before Camera AA passes.
So, that being said, and I was getting frustrated, I rendered this image out @ 4x resolution, and scaled it to final size:
(Time : 2h16m39s @ 3200 x 2400; No Adaptive AA, 4 AA passes, 1 Environment light @ 25 samples; 4 / 2 / 2 Refl/Refr/SSS Samples)
(No Noise Filtering, Sinc Filter, Radius 3.0)
And things look wonderful.
(Note, it appears that render times scale pretty much linearly; My estimate of 2h40m was just a bit off the actual time, based on a 10m time for an 800 x 600 render with the same settings - I took the square of the resolution multiplier (3200 x 2400 is 16x the number of pixels of 800 x 600), times 10m)
So, I started with everything as defaults @ 800 x 600, and started with 25 AA passes / 4 samples on the environment light, and started reducing AA passes while bringing light samples up.
(Time : 14m27s @ 800 x 600, No Adptive AA, 7 AA passes, 1 Environment light w/ 16 Samples, 4/2/2 Refl/Refr/SSS Samples, Sinc Rec. Filter, Rad = 3, No Noise Filter)
After this point, fewer AA passes start to degrade the FFX fibers.
Things I have discovered
1) Denoiser shouldn't be used with FFX, it obliterates fiber details.
2) Adaptive AA incurs a fairly hefty performance hit in this example - about 25% longer. I thought AAA was supposed to help save us time?
3) FFX likes and needs AA passes; The lowest I would go is 6. After that things degrade quickly.
4) I need to work on making my fur textures stand out/come through in FFX.
1) Keep Camera AA Passes as low as you can.
2) Raise lighting samples, Refl. Samples, Refr. samples, and SSS Samples first, before Camera AA passes.
So, that being said, and I was getting frustrated, I rendered this image out @ 4x resolution, and scaled it to final size:
(Time : 2h16m39s @ 3200 x 2400; No Adaptive AA, 4 AA passes, 1 Environment light @ 25 samples; 4 / 2 / 2 Refl/Refr/SSS Samples)
(No Noise Filtering, Sinc Filter, Radius 3.0)
And things look wonderful.
(Note, it appears that render times scale pretty much linearly; My estimate of 2h40m was just a bit off the actual time, based on a 10m time for an 800 x 600 render with the same settings - I took the square of the resolution multiplier (3200 x 2400 is 16x the number of pixels of 800 x 600), times 10m)
So, I started with everything as defaults @ 800 x 600, and started with 25 AA passes / 4 samples on the environment light, and started reducing AA passes while bringing light samples up.
(Time : 14m27s @ 800 x 600, No Adptive AA, 7 AA passes, 1 Environment light w/ 16 Samples, 4/2/2 Refl/Refr/SSS Samples, Sinc Rec. Filter, Rad = 3, No Noise Filter)
After this point, fewer AA passes start to degrade the FFX fibers.
Things I have discovered
1) Denoiser shouldn't be used with FFX, it obliterates fiber details.
2) Adaptive AA incurs a fairly hefty performance hit in this example - about 25% longer. I thought AAA was supposed to help save us time?
3) FFX likes and needs AA passes; The lowest I would go is 6. After that things degrade quickly.
4) I need to work on making my fur textures stand out/come through in FFX.
RebelHill
Goes bump in the night
Yeah... ffx seemingly creates an exception to the low cam samples thing... but really it just creates a TON of "spatial details" that need to be resolved, and AA is the only way. So the rule applies, but the "as few AA passes as needed" is still often a higher number.
As for ffx + GI... i'll just leave you to that, id rather finish my renders before the end of time.
As for ffx + GI... i'll just leave you to that, id rather finish my renders before the end of time.
Snosrap
Active member
Pretty cool Terb! Did you use weight maps around the nose and mouth etc.?
Terb1097
New member
Thank you. No weight maps. Just 3d texture painting and exported texture maps, metal, roughness etc..
antik22
Member
Hi everyone great tests!!! I have problem with glass surfaces in 2018, I've tried Dielectric as well as BSDF, but always getting same result. When two glass panes in behind other its looks pitch black. Played with background colors same results. I have image backdrop in my sciene. Maybe it's have to do with Shadow Catcher though I ve tryed different settings under the Shading Model tab. Is it a bug?
Attachments
Try to increase Refraction Recursion Limit
Look at:
https://docs.lightwave3d.com/display/LW2018/Render+Properties+-+Render
antik22
Member
Asticles thank you for reply tried all the these. Still the same. Looks like it did not have any samples.
Please note that with BSDF material 100 percent transparency the same, but when I set it to 85 percent its goes away.
p.s. My apologies I also asked about this issue in this thread http://forums.newtek.com/showthread...arency-actually-working&p=1539879#post1539879
Please note that with BSDF material 100 percent transparency the same, but when I set it to 85 percent its goes away.
p.s. My apologies I also asked about this issue in this thread http://forums.newtek.com/showthread...arency-actually-working&p=1539879#post1539879
