PDA

View Full Version : Caustics question



Snosrap
04-03-2015, 09:24 PM
How is Chris getting caustics to work? He doesn't even have caustics turned in the Render Globals panel. Is it a 2015 thing? See here:
https://vimeo.com/123362353

spherical
04-03-2015, 10:29 PM
What I want to know is, how'd you get your embedded video to be a thumbnail. It even plays!

Snosrap
04-03-2015, 10:48 PM
Go advanced and click movie icon.
127728

Oh ya and for what it's worth it's that silly "draft" mode on VPR keeping caustics from working. :)

spherical
04-04-2015, 05:18 PM
Go advanced and click movie icon.

I do that. It comes in at the usual embedding size; like all of the others seen previously, here.

Tartiflette
04-04-2015, 06:11 PM
How is Chris getting caustics to work? He doesn't even have caustics turned in the Render Globals panel. Is it a 2015 thing? See here:
https://vimeo.com/123362353
Doesn't have Flash installed so i can't see the video you linked, but caustics are "just" a side effect of GI with "directional rays" turned on. (you would need to check the "use transparency" button as well)

The good thing is that these caustics actually work, as opposed to the one you get with caustics turned on in the Render Globals panel, which fails miserably almost 100% of the time ! :D
The bad thing is that it takes longer to render a complicated scene (that is, NOT a ring and 2 balls) when "directional rays" is turned on, so that's the biggest trade off.
The other thing to consider is that you will never get the crisp caustics you can get from a bidirectional path tracer as Maxwell, Thea, Indigo or LuxRender (this "list" is non-exhaustive), but it's already good enough in a lot of cases.

And to reply more directly to your question it's NOT a 2015 feature at all, you can get this with LightWave 9.5 if you want. :)
What is new in 2015, though, is the "BG Importance Sampling" feature that permits this kind of caustics to show up even when only using a HDRI as a background image, whereas these caustics were only accessible through light panels before...

Hope it helps.


Cheers,
Laurent aka Tartiflette. :)

Snosrap
04-04-2015, 06:48 PM
Thanks Laurent for the additional info. We've upgraded to 2015 at work but I haven't gotten around to it on my home box. I was using the Discovery version of 2105 when I discovered the VPR Draft mode checkmark trying Chris' setup and once I figured that out it worked in 11.6.3 as well. That Importance Sampling is a nice upgrade to 2015. :)

Tartiflette
04-05-2015, 04:16 AM
That Importance Sampling is a nice upgrade to 2015. :)
It sure is ! :)


Cheers,
Laurent aka Tartiflette. :)

Mr. Wilde
04-05-2015, 07:29 AM
Doesn't have Flash installed so i can't see the video you linked, but caustics are "just" a side effect of GI with "directional rays" turned on. (you would need to check the "use transparency" button as well)

The good thing is that these caustics actually work, as opposed to the one you get with caustics turned on in the Render Globals panel, which fails miserably almost 100% of the time ! :D
The bad thing is that it takes longer to render a complicated scene (that is, NOT a ring and 2 balls) when "directional rays" is turned on, so that's the biggest trade off.
The other thing to consider is that you will never get the crisp caustics you can get from a bidirectional path tracer as Maxwell, Thea, Indigo or LuxRender (this "list" is non-exhaustive), but it's already good enough in a lot of cases.

And to reply more directly to your question it's NOT a 2015 feature at all, you can get this with LightWave 9.5 if you want. :)
What is new in 2015, though, is the "BG Importance Sampling" feature that permits this kind of caustics to show up even when only using a HDRI as a background image, whereas these caustics were only accessible through light panels before...

Hope it helps.


Cheers,
Laurent aka Tartiflette. :)
Thanks a lot for the explanation. I didn't even know that you could create caustics like that.

wellsichris
04-05-2015, 08:47 AM
Yes you could have done a smilier technique back in the day, but with a light probe it would have taken a very very long time to resolve,

as for the idea of "never getting crips caustics" as explained you can in theory it will just take a very long time. the limit dynamic range hack "with interpolation OFF" will give you crisp caustics, but not nearly as intense that's what I mean when I say not technically correct, the shape is, but the intensity is not. the technically correct caustics, that are intense, just take to long to be practical but are possible. I have maxwell version 2 and have tried it in c4d and both those renderers appear to be doing the same thing, "limiting dynamic range" octane does it more technically correct intensity, just takes many rays to refine it.

one last thing, I'm not saying, this is a perfect system, I have talked to the devs and hope this is something they are working on. the point of the video isn't to say this is the ultimate solution for caustics. the point of the video was. hey here is a way to make it work in case you weren't aware of the technique. I do light probes, I use light probes for dang near everything. and it's nice to have this option. thanks Tartiflette for explaining in more detail the technique.

Chris

Tartiflette
04-05-2015, 10:54 AM
yes you could have done a smilier technique back in the day, but with a light probe it would have taken a very very long time to resolve,

as for the idea of "never getting crips caustics" as explained you can in theory it will just take a very long time. The limit dynamic range hack "with interpolation off" will give you crisp caustics, but not nearly as intense that's what i mean when i say not technically correct, the shape is, but the intensity is not. The technically correct caustics, that are intense, just take to long to be practical but are possible. I have maxwell version 2 and have tried it in c4d and both those renderers appear to be doing the same thing, "limiting dynamic range" octane does it more technically correct intensity, just takes many rays to refine it.

One last thing, i'm not saying, this is a perfect system, i have talked to the devs and hope this is something they are working on. The point of the video isn't to say this is the ultimate solution for caustics. The point of the video was. Hey here is a way to make it work in case you weren't aware of the technique. I do light probes, i use light probes for dang near everything. And it's nice to have this option. Thanks tartiflette for explaining in more detail the technique.

Chris
Yes, the fact that you can get caustics (and, even more important for day to day projects, clean renders !) from background HDR images is a HUGE plus in LightWave 2015, that's for sure.
But, yes, the "caustics from GI" were already possible with LightWave 9.6 (even 9.2 iirc) with light panels, i know i've used it a lot at that time when testing the new GI engine.

As for getting crisp caustics with LightWave, though, i wouldn't be so affirmative, as for me crisp caustics from GI engines can really be achieve with a bi-directional path-tracer, trying to get the same result with a "simple" path-tracer is just out of reach, except if you have a battery of super-calculators ready to render for a century !
But that's the nature of it, not "LightWave's fault", because every simple path-tracer are facing the same problem.

Here is a render from maxwell from i would call "crisp caustics", i don't think you can render that kind of thing within LightWave, even if let the render cook for 2 or 3 months.
But i could be wrong, of course, in which case i'd be eager to see that kind of quality render coming out of my beloved render engine. :)

127739


Cheers,
Laurent aka Tartiflette. :)

Mr. Wilde
04-05-2015, 02:43 PM
Does Importance Sampling have any effect on non image based lighting renders? The way I understand it, it only affects the ray generation from a HDR image. The brighter a spot on the image is, the more rays are generated for that spot. Or does it also affect the ray generation from extremely bright surfaces and light sources?

wellsichris
04-05-2015, 06:59 PM
in it's current state it's just background. that can be a light probe. or light studio plugin but it doesn't help with bright surfaces. hopefully some time they will extend it further. I believe there are some other renders that do it so it is possible just not sure how hard it is to implement

madno
04-06-2015, 04:04 AM
I tried caustics with interpolated GI (took 45 minutes to render).

127748

Because I used an inside pointing sphere, I could not use importance sampling.
Is there a way to use a node setup like the one on my sphere, within the textured environment?

127747

If yes, I could try a similar render with importance sampliing and see if that helps to reduce the render time.

Mr. Wilde
04-06-2015, 04:37 AM
I tried caustics with interpolated GI (took 45 minutes to render).

127748

Because I used an inside pointing sphere, I could not use importance sampling.
Is there a way to use a node setup like the one on my sphere, within the textured environment?

127747

If yes, I could try a similar render with importance sampliing and see if that helps to reduce the render time.

You could try to use the Texture Environment, then use Procedural Texture and select Node Editor.

S0nny
04-14-2015, 11:25 AM
I saw the video yesterday thanks to the newsletter and I found this thread becase I can't get it work even with the same image as backdrop. Anybody tried yet? Same settings as the video.


EDIT: solved

Dan Ritchie
04-14-2015, 12:36 PM
What I want to know is, how'd you get your embedded video to be a thumbnail. It even plays!

You can embed video from several different places, such as youtube, facebook, or Vimeo if you use the "go advanced" option when posting.

spherical
04-14-2015, 01:51 PM
EDIT: solved

How, please?

jboudreau
04-14-2015, 05:46 PM
I saw the video yesterday thanks to the newsletter and I found this thread becase I can't get it work even with the same image as backdrop. Anybody tried yet? Same settings as the video.


EDIT: solved


Hi

I was wondering how you solved your problem? I tried the same settings as the video and couldn't get as good a results as chris did in his video.

Chris is there anyway you can provide us with the scene file. I'd really like to test this out again.

Thanks,
Jason

Snosrap
04-14-2015, 08:22 PM
How, please?

Be sure and turn Draft mode OFF in VPR.

jboudreau
04-15-2015, 07:21 AM
Be sure and turn Draft mode OFF in VPR.

Hi

I made sure I had Draft mode off at the time. Still can't get the same results. Wondering if it's the materials I'm using, It's hard to tell without having the same scene file Chris used

Thanks,
Jason

mummyman
04-15-2015, 07:42 AM
I couldn't get the "same" results, but I was using different EXR / HDRI images. They gave some similar results. I turned all my lights off and just used the texture image as a backdrop like the movie tutorial. The caustics seem to show. Depends on the angle of the image. Play with rotations maybe? My sphere was just using a simple incidence gradient falloff in the transparency channel. Give it some refractive index (1.3) or something... and add some color filter. I used a red sphere...so nice red caustics!

Add some reflection and make sure the sphere / glassy object is double sided too!

wellsichris
04-15-2015, 10:03 AM
Here is the scene, Go to my website and get a couple of the free light probes, or better yet buy some :) then in the texture Environment assign them . hit render. Hopefully this will help those that are still having difficulties. I've also attached a couple of renders I did with a couple of the free probes. Each took about 1 min on my machine.

mummyman
04-15-2015, 10:06 AM
Here is the scene, Go to my website and get a couple of the free light probes, or better yet buy some :) then in the texture Environment assign them . hit render. Hopefully this will help those that are still having difficulties. I've also attached a couple of renders I did with a couple of the free probes. Each took about 1 min on my machine.

Excellent work! Very nice tip on this video. Thanks again

wellsichris
04-15-2015, 10:19 AM
So I have been playing with "sharp" defined caustics, and yes it is way harder to get realistic render times. It seems the difference between focusing light down, which can be fast and spreading light out. which takes way more samples. The deal with maxwell is at fast render times the caustics are very subtle but if you let it bake for a long time they come out and are accurate, but you have to let it bake a while. That being said, You have to let LW bake a long time as well.

so here are the renders. with appropriate render time burned into jpg. Also my specs are listed, I have a fast machine, both CPU and GPU. so it's maybe hard to compare. But at least it will give you an idea.
for reference, my 670 is about 6x slower than my current titans setup. so on a 670 the octane render would have been 6 hours.

So conclusions. Octane is pretty great at caustics. If my life consisted of doing caustics renders. It would be the way to go. LW and maxwell seem fairly comparable to me. both can get sharp defined caustics if you let them bake long enough. But both are painful. LW needs many samples. and by many I mean, GI was 1 sample with 1 bounce, and AA was 65536. yes 65k samples to clean it up. No Limit Dynamic Range. I did some PS color adjustments to try and get them all closer in appearance to better compare.

pretty interesting.
Chris

jboudreau
04-15-2015, 01:47 PM
Here is the scene, Go to my website and get a couple of the free light probes, or better yet buy some :) then in the texture Environment assign them . hit render. Hopefully this will help those that are still having difficulties. I've also attached a couple of renders I did with a couple of the free probes. Each took about 1 min on my machine.

Thanks Chris for sharing the scene

I figured out the problem. I had all the setting correct but there were 2 issues:

1 - My Ring texture was not using nodes, I just made a standard reflective material with nodes exactly how you did with nodes except I noticed that with the nodes you can tell it to receive Caustics, Where I didn't have that option using the standard material editor.

2- Somehow I clicked on the jpg verson of the Light Probe image instead of the EXR. I only realized it when I comparded your scene with mine.

On another note does anyone know how Chris got his scene to show the wireframe geometry when selected while in bounding box mode? See image below

127873

Thanks,
Jason

jboudreau
04-15-2015, 01:51 PM
So I have been playing with "sharp" defined caustics, and yes it is way harder to get realistic render times. It seems the difference between focusing light down, which can be fast and spreading light out. which takes way more samples. The deal with maxwell is at fast render times the caustics are very subtle but if you let it bake for a long time they come out and are accurate, but you have to let it bake a while. That being said, You have to let LW bake a long time as well.

so here are the renders. with appropriate render time burned into jpg. Also my specs are listed, I have a fast machine, both CPU and GPU. so it's maybe hard to compare. But at least it will give you an idea.
for reference, my 670 is about 6x slower than my current titans setup. so on a 670 the octane render would have been 6 hours.

So conclusions. Octane is pretty great at caustics. If my life consisted of doing caustics renders. It would be the way to go. LW and maxwell seem fairly comparable to me. both can get sharp defined caustics if you let them bake long enough. But both are painful. LW needs many samples. and by many I mean, GI was 1 sample with 1 bounce, and AA was 65536. yes 65k samples to clean it up. No Limit Dynamic Range. I did some PS color adjustments to try and get them all closer in appearance to better compare.

pretty interesting.
Chris

Hi Chris


I was setting up a similar scene with a water glass for testing. Are you able to share your glass scene. I'd love to see how you set it up.

Thanks,
Jason

wellsichris
04-15-2015, 02:26 PM
"On another note does anyone know how Chris got his scene to show the wireframe geometry when selected while in bounding box mode? See image below"
Bullet Dynamics are enabled. the coin falls and hits the balls. was going to do a video of it for fun, but in bullet it's set to draw bodies.

jboudreau
04-15-2015, 02:58 PM
"On another note does anyone know how Chris got his scene to show the wireframe geometry when selected while in bounding box mode? See image below"
Bullet Dynamics are enabled. the coin falls and hits the balls. was going to do a video of it for fun, but in bullet it's set to draw bodies.

Haha, I thought it was some sort of new display feature. Cool trick. Even when you disable the Bullet Dynamics it still shows the wireframe geometry while in bounding box mode which is pretty cool

Thanks
Jason

spherical
04-15-2015, 03:06 PM
and make sure the sphere / glassy object is double sided too!

Only if you are doing "old school" transparent materials. Any texture that can take advantage of the Volume Stack does not need double-sided polys. They will mess up the refraction and/or just slow the render down. Haven't needed to go through that hack for years, now.

wellsichris
04-15-2015, 03:11 PM
Decided to render the anim with one of my probes :) I really love the light going through the red ball and bouncing off the ring. I did a little comp work, degrain, bloom, that type of thing.

Mr. Wilde
04-15-2015, 03:12 PM
So I have been playing with "sharp" defined caustics, and yes it is way harder to get realistic render times. It seems the difference between focusing light down, which can be fast and spreading light out. which takes way more samples. The deal with maxwell is at fast render times the caustics are very subtle but if you let it bake for a long time they come out and are accurate, but you have to let it bake a while. That being said, You have to let LW bake a long time as well.

so here are the renders. with appropriate render time burned into jpg. Also my specs are listed, I have a fast machine, both CPU and GPU. so it's maybe hard to compare. But at least it will give you an idea.
for reference, my 670 is about 6x slower than my current titans setup. so on a 670 the octane render would have been 6 hours.

So conclusions. Octane is pretty great at caustics. If my life consisted of doing caustics renders. It would be the way to go. LW and maxwell seem fairly comparable to me. both can get sharp defined caustics if you let them bake long enough. But both are painful. LW needs many samples. and by many I mean, GI was 1 sample with 1 bounce, and AA was 65536. yes 65k samples to clean it up. No Limit Dynamic Range. I did some PS color adjustments to try and get them all closer in appearance to better compare.

pretty interesting.
Chris
The "Maxwell 7hrs" and the "Octane 1hr" image are identical. Did you perhaps mix up the images?

spherical
04-15-2015, 03:15 PM
So conclusions. Octane is pretty great at caustics.

Interesting that Octane and Maxwell at 7 hrs. are effectively identical. LightWave's glass is darker/more absorbent. What is the Dielectric Absorption set to and what color is it?

wellsichris
04-15-2015, 03:20 PM
yes I did mix it up they shouldn't be identical. had them all stacked as layers in PS. forgot to change it. anyway here is octane 1 hr. and maxwell 7.

wellsichris
04-15-2015, 03:28 PM
here's the scene, Minus light probe, if anyone want to see, but it's very basic and takes long time to render. so not to fun to play with.

spherical
04-15-2015, 03:28 PM
OK. Now Octane's glass looks more like LightWave's.

wellsichris
04-15-2015, 03:39 PM
they all have the same IOR 1.45 and consequently the caustics look dang near identical which is cool, but it was hard to get the shaders to match exactly, I just got it as close as I could.

spherical
04-15-2015, 03:45 PM
Last year I did a comparison of sorts between Thea Render and LightWave on caustics. Starts here:
http://forums.newtek.com/showthread.php?136117-Thea-Render-1-2-released-introductory-offer&p=1366296&viewfull=1#post1366296
and goes for a number of posts to where I got a workaround to get LightWave closer to Thea's output. Some tradeoffs in order to get render times down. Was an interesting and educational exercise.

jboudreau
04-15-2015, 06:01 PM
here's the scene, Minus light probe, if anyone want to see, but it's very basic and takes long time to render. so not to fun to play with.

Hi Chris

I downloaded your glass scene and put in one of your free Light Probe images and downloaded the Octane Render demo. When I go to render your scene I don't have any caustics at all and the glass does not match the octane render you have here on the forum

Does the Demo not render Caustics or is there something else I'm missing here?

Thanks,
Jason

wellsichris
04-15-2015, 09:15 PM
Jason. just get a probe that has CSP at the end of it. or in other words has a bright sun. I think there are two in the free ones. I think the demo works with caustics. but not sure if it keeps the settings. watch my other video about setting things up in octane for LW.

madno
04-16-2015, 02:58 PM
After 11h 29min I got this using a dual Xeon.
All non interpolated GI without LW lights.
For some reason the bump on the floor does not look like expected (even though "use bump" in GI was active).

127886

S0nny
04-17-2015, 11:15 AM
I apologize I didn't answer before, I was busy with work and I totally forgot the post. I see Chris explained everything. Anyway for me It was vpr in draft mode and too low sampling. Also you need to wait a lot of time, my ws is pretty fast but not faster as the video so it's better to have patience!