Hi all
What is the difference between saving an EXR from Max versus saving an EXR using the VRimg? It seems to be 2 different kinds of OpenEXR?
The ones written from 3ds Max are “scanline” exr files (the information is written in the file scanline by scanline); the ones produced with the “raw image output” of V-Ray are “tiled” exr files (the information is written bucket by bucket).
Best regards,
Vlado
When using them in Nuke, is there any difference ? Nuke is scanline based isn’t it ?
I’m told that nuke handles scanline .exr files faster.
Also, I forgot to mention that exr files produced from the vrimg2exr tool are also scanline based.
Best regards,
Vlado
Reading tiled EXRs is generally slower due to the inherent characteristics of the file format.
Each tile has to be read entirely (all channels) before displaying any of the channels for that tile.
In scanline images, each channel can be read independently, making them a much faster alternative for post in any application.
Including nuke.
And even worse, as tiles are in arbitrary order there’s a lot of jumping back and forth whereas scanlinebased ones can be blindly read.
This is not only valid for nuke, but due to nukes great scanline architecture it’s even more appearent.
Regards,
Thorsten
and tiff’s and png’s under what kind of type are these falling?
does that mean that by converting the tiled based exrs output by the vray fb, via the vrimg2exr tool, to scanline based exrs, that they’ll be quicker to load, change channels, etc… in post?
that depends on the aplication you use for post, nuke prefers scanline and fusion prefers buckets if I’m not wrong, what works best with after effects I don’t know
isn’t it possible to decide in whych EXR format you save the images ? (scanline or bucket) ? because savings lets say 1000 frames of an animation in vrimg and then converting all of them to .exr scanline and then starting to work on them, …well it kind of slow down the workflow and uses a lot of HD space 
Well that would defeat the whole idea. Saving to vrimg was introduced to allow writing buckets in arbitrary order. Hence you dont need the whole image in memory for writing it wich enables you to save a lot of memory when rendering big images without memoryVFB.
And EXR also allows for tile based images and hence was added as an option.
This is especially true when using a lot of renderelements. Memory usage can go through the roof pretty quickly.
Regards,
Thorsten
the fact that one may have N renderElements into an EXR makes it extremely slow (and very memory intensive) to read the EXR as tiled, in ANY application, fusion included.
It’s a case of file format, not of application: in a scanline image, if you want to read the RGB channel, you can do so directly: your app will ONLY read the RGB channels out of whatever element you may have in the EXR.
If it was tiled, to extract the RGB the app would need to read ALL tiles and ALL the embedded channels (i have seen exrs with well over 100 channels, go figure…),reconstruct the image in a humanly readable format, then discard the lot and finally present the RGB pass.
Just by writing it it’s longer 
How do they handle this at ILM or DD ?
I sure as hell do not know, but i would suppose like everyone else does: post process the tiled images into scanline, and get on with the post 
Thorsten, how do you guys do it at Mackevision? You save directly to a multichannel EXR from Vray raw image for your animations? (in other words instead of *.vrimg, you just change the extension to *.exr?) Then after that, post process like Lele said?
How do you do that ? to convert tiled based exr to scanline exr ?
read the entire post. i think it was mentioned twice. vrayimg2exr
yeah but if i allready rendered the animation sequence and saved it as .exr from vray. does it still work?
that would be interesting for someone to create an exr-re-order script. you can specify if your going to buckets and what bucket size or if your going to scanline from buckets. what does pdplayer prefer? bucket exr’s?
I’ll bump this one up, cause i just got aware of this bucket/scanline thing with .exr’s.
I’ve always rendered through the framebuffer and just substituting the .vrimg with .exr in the filename. And then using ProExr plug-in in AfterEffects to open up the renderelements.
Should i have done some kind of converting of the .exr’s before i bring them in to AfterEffects?