PDA

View Full Version : Volumetrics, Viper and Rendering in Progress.



RedBull
08-08-2006, 11:36 PM
Has anyone noticed volumetrics not working with 'rendering in progress' and the perspective camera?

Basically if you use Viper it will stop working, turning Viper off and Enable Viper OFF will not restore it to the default working mode. Which would be good at least.

Anyone else noticed this?

Add Null
Add Hypervoxels and Activate Null object.
Hit F9 (make sure "rendering in progress" is on and set to 320 or 640)
(Notice that you can see the volumetrics render in the preview window)

Enable Viper:
Hit F9 for render: Now you will not see the volumetrics in the preview.
(Turning Viper off in both panels does not restore it to the previous state.)

PS: Just noticed it's all the new camera types that this occurs on.
PPS: I also notice Advanced Camera seems to break Viper completely too.

Dave Jerrard
08-10-2006, 10:59 AM
They're definitely weird. Sometimes they show up & sometimes they don't. I noticed that if I start off with them without transparency, they sho wup in Render In Progress fine, but if I tweak them with transparency, that stops them from appearing during rendering again. They are rendering though. You just don't see them until the frame is done.

The good news is, they DO respect the new cameras, so if you're using some advanced camera settings, HVs and other volumetrics will render correctly. Now if we could just see them as they render, all the time....


He Who Often Feels Like He's Rendering Blind When This Happens.

RedBull
08-10-2006, 05:10 PM
Yeah it basically means i need to use Classic Camera, for volumetrics or to at least use Viper with volumetrics. I've noticed that if you have geometry, the geometry shows in the rendering in progress but you may see spurts of volumetric passes during AA passes..

But it's a bit crappy, as i rely on this feature particularly on processer intensive tasks like volumetrics. FPrime does most things but obviously not volumetrics so without some kind of feedback it's really limiting.

I'm starting to use XSI 's particles and FX over LW's, which is the opposite to how i'm used to doing things, but Mental Rays region tool is far better for giving me fast feedback these days. Viper is almost a joke, as i can do faster renders at small res using multiple processers faster than Viper can draw a single threaded thumbnail.. So having this work with rendering in progress would be nice.

PS: Here's a more severe problem.
I notice that you can break Viper in LW9 easily with the Camera Panel open.

Load Layout:

1) FXBrowser - Add HV Emitter (close FX Panel)
2) Open Viper Panel
3) Add Volumetric Plugin - Pixie Dust: (You may need to change the particle size to 2.0 in PD Panel to be able to see the particles in Viper)
4)Hit Render on Viper - You should now see a white particle cloud in Vipers preview.
5)Open Camera Properties:

Notice that Viper stops responding now that it's open, and will not respond.
THIS SUX! It means everytime the Camera Panel opens Viper is screwed until you restart Layout.

You can press render or update, but Viper doesn't respond anymore.
There is obviously some volumetric/camera/viper conflicts and problems that need to be addressed over the next update.

Can someone confirm it please!
It's nice to know i'm not crazy(ier) :)

It took a lot of testing time to narrow down both these problems...
And i have still have to work out some renderproblems with my 6hour Taiki test which has render errors the smaller res version didn't. Grrrrrr!

Note: I believe that Hypervoxels is the odd one out, as Viper seems to work without this problem... All other volumetric plugins will break Viper when opening the camera panel... Taiki, TBinfinite Plane and PixieDust all seem to exhibit this problem.

Lightwolf
08-10-2006, 05:24 PM
Check your VIPER title bar.
If you switch to the camera panel, VIPER will switch contexts to the Camera, which also allows you to save camera presets.
To get VIPER to render HVs again, click on the HV panel to change the VIPER context, and then render in VIPER or tweak a HV setting to trigger a referesh.

Cheers,
Mike

stevecullum
08-10-2006, 05:29 PM
But it's a bit crappy, as i rely on this feature particularly on processer intensive tasks like volumetrics. FPrime does most things but obviously not volumetrics so without some kind of feedback it's really limiting.

I'm sure it said there was support for HV's on Worleys site..

Or is that just for rendering to disk?

RedBull
08-10-2006, 05:54 PM
Check your VIPER title bar.
If you switch to the camera panel, VIPER will switch contexts to the Camera, which also allows you to save camera presets.
To get VIPER to render HVs again, click on the HV panel to change the VIPER context, and then render in VIPER or tweak a HV setting to trigger a referesh.

Cheers,
Mike

Yep Mike you are correct, because NT have added camera presets in LW this issue seems to be tied to the camera presets. Obviously 8.x didn't have them to interfere with Viper.....

Obviously this needs to be fixed, (as you said reopening the panel can get it to update it again) But still it needs to be fixed, or work like 8.x...As that's a bad workflow killer, and has caused me hours of chaos between both of these issues. But thanks if i don't have to shutdown LW everytime, at least it's still usable.

And now it's just means everytime i tweak my camera, i have to reopen the volumetrics panel to update viper, back and forth.... And that's all at the cost of having rendering in progress.


I'm sure it said there was support for HV's on Worleys site..

Or is that just for rendering to disk?

I mentioned Volumetrics, not specifically Hypervoxels.
Hypervoxels are not the only volumetrics i use, Taiki, Infinite Plane, HD-Instance, Ground Fog, Pixie Dust, Vlights are often used more these days.

FPrime has support for HV in it's renderer, not in the previewer....
(Although their are builds of FPrime that allow it in the previewer too) ;)
But really it's not faster than LW to render HV, and without the realtime preview is not a huge benefit to me personally.

RedBull
08-10-2006, 09:53 PM
While we are here and without another thread.
Here is a volumetric cloud test shot done in Taiki....
I have two more concerns on this project, that are driving me nuts.

Rendertimes:
-------480x270----960x540----(Resolution)
LW8.5 = 2.45 ---- 23.27 (Classic Camera)
LW9.0 = 5.20 ---- 48.44 (Classic Camera)
LW9.0 = 4.40 ---- 39.55 (Perspective Camera)

As you can see LW8.5 is far faster with no visual difference to LW9.0.
Even with the classic camera. In fact with or without raytracing enabled Taiki is faster with the LW9 Perspective camera, but still almost twice as slow as LW8.5. Anyone know of any reasons why this could be? Volumetrics in general do appear slower to me in LW9.0!

I'm having problems, in that at a test resolution of 480x270 the render looks fine, if i double the resolution and render @ 960x540, rendering errors are introduced, and i can't for the life of me work out why this is (consider my complete lack of sleep as a possible excuse) It seems to be an obvious thing but i can fix it, Adaptive sampling, construction filters and AA are not the cause. I'm still looking into older versions of the plugin to compare.
The problem is evident in 8 and 9, so Taiki perhaps may be the problem on this one. Still investigating...

Here is a [email protected] - and on the right is a 960x540, but downsampled in Photoshop to 480x270. The difference is fairly obvious. Any thoughts on these problems are appreciated...

PS: I also note that with 2x the resolution, i get basically a 10x performance hit on those tests.
And on the flipside i notice I have reached the magic milestone of 666 posts...... Bwahahahahaha...

Karmacop
08-11-2006, 12:21 AM
Scott, if you never post again it'd be quite a fitting number for you ;)

That's a big speed difference, and without knowing how taiki works and what is in your scene there's not much I can say. Are the clouds the only thing in the scene?

The double thing again is strange. If you set the resultion to 960x540 and don't use the 200% multiplier do you get the same problem?

stevecullum
08-11-2006, 03:07 AM
I've seen massive speed ups with Lightwave native HV's compared to 8.5. It could be that some of the code that was re-designed, has had a negative impact on third party volumetrics. :confused:

RedBull
08-11-2006, 03:35 AM
Scott, if you never post again it'd be quite a fitting number for you ;)

LOL, yeah it may just very well be..... I have freakishly awoken speaking tounges on more than one occassion, so anything is possible. :)


That's a big speed difference, and without knowing how taiki works and what is in your scene there's not much I can say. Are the clouds the only thing in the scene?

It's a simple scene which i can't post.. But i will post another one similar, later.
But i'll have a poke around and see if i can find the anomaly first.
The render artifacts that occur are of more concern than the rendetimes at the moment :cursin: I still have to try the previous Taiki.


The double thing again is strange. If you set the resultion to 960x540 and don't use the 200% multiplier do you get the same problem?

My default resolution is 960x540, for tests on Taiki i usually use 50% multiplier
so 480x270 that render worked fine. Setting the multiplier back to 100% (960x540) causing the seemingly aliasing/ghosting edge problems.

But it's even stranger if you look at the clouds closely, they are different.
There's a small gap noticable in the right image, that's closed in the smaller render. I don't know how this is possible, or if it's a volumeteric thing.

Just for fun i did a PAL res test, and the artifacts are also present at 720x576.
So i guess i need to slowley increase res, from 480x270 until artifacts are introduced... It's a real head scratcher.

I did think and test the multiplier setting, actually due to the seemingly strange exponential rendertime increase that shows around a 10x slowing
in rendertimes, with just a doubling of the resolution. Which i find strange.
I'm not sure if this is a Taiki or LWVolumetrics rule, but i'm sure geometry would scale in a more linear fashion, if memory would allow.

I've done continual test renders for 24hours, without any luck to solve this artifact issue. I'll forward some more screenshots to Akihiro tomorrow, in the hope he can solve the issue or tell me how i can. Thanks for the reply.

Karmacop
08-11-2006, 08:15 AM
It's a simple scene which i can't post.. But i will post another one similar, later.
No need to post the scene, I don't have taiki anyway, I just thought that if there's some geometry maybe there's a slow down because of the KD-tree as taiki isn't part of that tree. I'm not sure, it's just weird because as Steve said HVs are much faster in 9, but maybe that's a change in HVs and not volumetrics.



Just for fun i did a PAL res test, and the artifacts are also present at 720x576.
So i guess i need to slowley increase res, from 480x270 until artifacts are introduced... It's a real head scratcher.
I'm at a loss as to why this would happen though ... have you tried using similar resolutions and see if you get the same problem? Or does this happen over a certain resolution? My only thought is that he's hardcoded some kind of sampling grid in and by chance your rays are passing through the volume on the "thinner" areas.

RedBull
08-11-2006, 08:26 PM
No need to post the scene, I don't have taiki anyway, I just thought that if there's some geometry maybe there's a slow down because of the KD-tree as taiki isn't part of that tree. I'm not sure, it's just weird because as Steve said HVs are much faster in 9, but maybe that's a change in HVs and not volumetrics.

I have not checked with HV, but i just did a Vlight test, with the same kind
of problems. The difference is staggering, but i came across a thread talking about shadow opacity, and that likely makes a fair bit of a difference, regardless if it's vlights, hv, or taiki. I'm guessing thier will be a shadow opacity setting in Taiki that is working in 9, that doesn't work in 8 that doesn't show any discernable visual difference. I don't believe Volumeterics have seen any speed increases as such.


I'm at a loss as to why this would happen though ... have you tried using similar resolutions and see if you get the same problem? Or does this happen over a certain resolution? My only thought is that he's hardcoded some kind of sampling grid in and by chance your rays are passing through the volume on the "thinner" areas.

Yeah i need to go up in scale in small increments until i can narrow down a few things. This version of Taiki is signifigantly faster than the last and is a Beta, so possibly it's just a Taiki error. So i wouldn't dismiss your opinion, on that one. However only evident at certain resolutions is whack.