The idea is to capture video from several cameras simultaneously and play these back later at a gallery show. I want to take the original video streams and play each back in a quad positioned more or less where it was in reality. Then you will be able to move around the scene while the videos simultaneously play back.
What’s the practical limit for number of videos you can play back simultaneously? Does all the video need to be in video or system memory? I will probably have about 5 clips, 30 seconds total run time.
If you want to see some video we shot of the test run, you can see the movie
For example, I have a shot from in front of the operators and another one behind. I want to make it so that when you move around the scene you see the appropriate view.
vvvv basically uses mswindows for video playback. so if you can fluently play 5 clips at the same time in the windows mediaplayer, you can assume that vvvv will play them as well - at least if your patch doesnt get to complicated.
video is always streamed from disk. so getting the data rate low, having superfast processors and disks, playing from different disks etc. will help.
there is a video faq elsewhere, but video playback efficiency increses when squishing the video source to a square format with a power of two as height and width. something like 1024x1024 or 512x512 or 256x256. stretch the quad with a transform node to go back to your 4:3 or 16:9 aspect ratio later.
as i suppose, windows is optimized for playing back exactly one video at a time,
use something like after effects to combine different source files into one clip and use a texture transform later on to select the right area of the texture.
i´d suggest trying if you can nicely play back a 1024x1024 30fps video on your machine and then using that area to mount 6 or 9 videos into that square. should be possible…
interestingly we recently had 7 (seven) dv-pal streams playing on an ordinary shuttle pc with an nvidia 6600 card. if you’d ask me if that is possible, i’d say no. but i saw it with my own eyes…
in another project we used mp2 encoded videos, because those (together with .wmvs) are the ones that have decoders that work directly on the graphics card (DXVA is the keyword for the decoders). i think the wmv-decoder is included with a mediaplayer download. dxva-capable mp2 decoders are available from different vendors (including nvidia and moonlight, or was it elecard). those decoders decrease the use of cpu while doing their job on the graphiccards.
from what i’ve learned i’d say playing 5 videos should not be a problem with filestream>videotexture. at least with the correct decoder.
what should also help is different a raid-system for your harddisk
actually i’d say the power-of-2 side format for the video is a historic thing and no longer gives an advantage with current graphic hardware. only beware that pixelshaders don’t cope with non-power-of-two textures.
Thanks for the feedback. That’s a good idea to combine my clips together at say 1024x1024 and grab different parts of the texture. Then I don’t need to worry about synchronisation.
@joreg: I also found that DV-pal tends to play back really smoothly. But did you notice any quality degradations? As you know, Quicktime differentiates between low and high quality on dv playback. I believe so does AVI, but i dont know the switch.
It might be interesting to add that we also managed to get 2 simultaneous & fairly jitter-free streams of 1920x1024x24 .WMV on the same machine.
= HD is a reality. Only skipping is a pest. Place key frames carefully.