Screamernet Light Projection

[joecygan]

Looking for any insight with this situation:

I have a rather lenghty animation that incorporates 10 Quicktime clips projected from lights. The lights are also volumetric, so even with minimal anti-aliasing settings, the render was going to take about 48 hours.

In an attempt to reduce that time, I used ScreamernetII on three machines. The setup worked fairly well, but after some time I realized that about every 3 frames or so had NO QUICKTIME clips being projected from the lights. My guess is just one of the Screamernet boxes wasn't handling the QT's, but why? All machines had access to the same files. All 3D elements and textures rendered beautifully, but for the QT's.

I have since found out that one of the boxes did NOT have QT loaded on it... but would that necessarily cause Screamernet to not know how to process the animation file?

I noticed that Screamernet did not seem to like .psd texture files. I converted all mine to .tgas and then they worked. Could the QT issue be something similar?

Unfortunately, I don't have time to investigate which box was not processing the QT's... I have to get the render done w/o Screamernet to make sure I meet my deadline. But this would be nice to know for future Screamernet efforts.

Anybody have any ideas?

Thanks,
joe

[Mylenium]

Whatever else may be the problem, one thing is for sure: If QT is not installed on your OS, LW can't use it. The LW plugins do not contain any native QT functions but will access Quicktime components on your system. If everything is installed properly, there might still be a minimum chance that at a time all 3 computers want to access the exact same frame in your QT file which also would produce bad results, especially if you are using a compressed CoDec where certain frames are calculated based upon preceding frames.

Mylenium

[joecygan]

Well, then it would seem most likely that the box w/o Quicktime installed was the one not producing the video on the final frames.

I don't think it was a Screamernet issue... or that any of the computers were trying to access the same frames of animation, but I could be wrong.

After my (now very long) render completes sometime tomorrow, I will install QT on the one machine that didn't have it, retry the Screamernet render and post the results here for anyone who cares.

Thanks for the input, Mylenium.

[toby]

using image sequences is much more reliable than video files, LW has had other issues with QT also, like frame rates

[trentonia]

Man, if you have a crushing deadline, go to http://www.respower.com. These guys will help you out. If your job is taking 48 hours on your machine, you can probably render it at respower in about 20 minutes at an approximate cost of $80.00. Check it out.

[ResPower]

Yeesh! Trentonia plugs us better than we do... his numbers are about right through...

Anyways I concur, it's a QT issue. Without QT then the screamer won't know how to decode the video file.

PSD's can be tricky. If there are any hidden channels, paths, or effects in the file, that can cause issues with screamer. If it's a completely flattened PSD file, then yeah, it should be good to use.

When people ask us what file format to use on our system, I always recommend using JPG with minimum compression. The files are relatively small, reducing overhead time on the loads. When you're loading a lightwave scene on 250 computers simultaneously, you want your network overhead as small as possible.

Even with video I recommend using JPEG image sequences. Makes it easier on the servers, the network, and the render nodes themselves.

Hope that helps!

[joecygan]

Interesting. I'd never heard of such a thing. Thanks for the info.