Compositing VERY LARGE images???

[The Bezier kid]

Your image files don’t need to be that big. Find out how close you are going to zoom into each photo and scale them down in photoshop before you import them into after effects. The extra pixels will only eat up your ram. I hope I’m not being too vague.

[Jonathan Miller]

I agree, only use images as large as they NEED to be. As well as bogging down AE to the point where it may not be fast enough to do even previews, if it does work, the extra resolution beyond what you need will cause the image to flicker.

Besides, AE can’t access more than 2 gigs of your 8 gigs of RAM.

Good luck!

Jon
TreeLine Productions
Fort Collins, CO USA

[Joey Foreman]

I recently worked with 4 large files of an atlas that I comped together to make one large world map. The comp size was about 15000 x 10000. I have 1.5 gigs of RAM and found that I could barely scrape by without getting out of RAM errors. If i added an effect to one of these layers it maxed out and i couldn’t render. But if it’s true that AE can only access 2 gigs of RAM then you’ll need to just carefully assess how much of each picture you’ll actually be seeing in the final output comp, and crop the images accordingly. In any case it will feel like swimming in wet cement with files that large. BTW, where’d you get that 100 megapixel camera?

[Accountneedsrealnameupdate]

I would recommend importing the full size images initially and then setting much smaller proxies for them, that way it’s easy to replace the small proxies with larger proxies when required and still keep all your positions and scale ok. You could even turn off proxies in your render if feeling brave and not in a rush (maybe last thing friday night you could queue one render using proxies and another without later in the queue so if it crashes you still have one render.
Glenn Stewart
1K studios

[Otheronenorehto]

Thank you all for your input. These are not Digital photographs thes are computer generated fractals which are only limited in size by the amount of calculating power you have. I am exploring alternate methods for animating fractals. A standards fractal animation can take anywhere from 3 days to 3 months to render so believe me I am not concerned about how slow AE will be rendering the image.

One thing that concerns me is what someone said about flickering. If I was zoomed out all the way with a 1280×720 frame wouldn’t AE resample the 43kx24k pixels to 1280×720? The aspect ration is exactly the same so it should be really easy for AE to do this…

I will have to do a little more research on Proxies. This sounds more like what I was thinking about. For instance as I zoom in AE can forget much of the original information, I think if I understand you correctly that Proxies would accomplish that? The truth is that AE would only have to deal with the whole 42kx24k image for only a short amount of time. I could remove tiles as I zoom in but I definately need them all near the begining.

The frame of the animation will only ever be 1280×720 so only at the very deapest parts of the fractal will the original image be a 1:1 pixel ratio.

Of course I am fairly sure that how the images are compressed will make a difference with how easily AE can handle such a large picture. Can anyone make recomendations or comments about how different compression codecs might effect AE’s ability to handle the picture? (lossy compression is out of the question it needs to be as lossless as possible)

Thank you again for your responses!

Jeremiah

[Otheronenorehto]

EDIT: sorry only each tile will be 43kx24k pixels big! The whole image would be 174,080×97,920 pixels big. Remember I am only ever rendering it at 1280×720 which is exactly the same aspect ratio ats 174,080×97,920! As long as AE can even display the whole image (16 43kx24k pixel images) it should only be a matter of resampling the image.

[Andrew Yoole]

AE has a hard limit of 30000 x 30000 pixels for comp sizes and footage sizes. You are not able to work beyond this scale, regardless of hardware.

[Otheronenorehto]

That is a shame. I was hoping that it was only for rendering not for composition/footage. Well I guess I will be checking out whether Shake has such a limitation If so then as my last resort I will be getting down and dirty and writing code. I don’t think it should be too hard since I am not interested in any effects or transitions, only movement of the camera on an x,y and z (for zooming only no tilting).

I wonder if there would have been artifacts of AE’s resampling alogarythm in a downsized image or if the large image reduced would have been indistiguishable from a Fractal rendered at the smaller resolution:)

Oh well thanks for your help!