RenderElements to Exr

[tbny65]

All,

It says that one should be able to save render elements to .exr using the Vray Frame Buffer by using .exr instead of .vrimg. I've tried but it doesnt seem to be working for me. Can anyone help or point to a thread where it might have been discussed?

Thanks

T

Edit: OK it works in Photoshop but only with a third party plugin. I now have the exr elements in photoshop but is the plugin necessary?

[vlado]

As far as I could test, the .exr works fine, but you need to have the right program to open multichannel EXR files (e.g Nuke). What are you using now?

Best regards,
Vlado

[tbny65]

Photoshop CS3. So its Nuke or AE that works? thanks Vlado!

[instinct]

I havent seen any App out there that is able to handle EXRs properly besides nuke. And even nuke doesnt handle all possible cases (tho it handles most, and surely any needed usually in a GREAT way). This is mostly due to the fact that the design of most comp and paint app dont allow for multichannel fileinput properly.

Anyways, nuke does a great job, AE does a better then PS if i recall right, and for PS you do need the plugin or split the renderelements to seperate images.

Regards,
Thorsten

[rivoli]

Anyways, nuke does a great job, AE does a better then PS if i recall right, and for PS you do need the plugin or split the renderelements to seperate images.

ae does indeed a much better job than ps (not only in handling exrs), but you still need a plugin to read any other channel than rgba.

[panthon]

I still wonder why photoshop is so slow when loading EXRs, while fusion is blazing fast.

[gunny263]

http://www.fnordware.com/ProEXR/

Proper Exr importing for CS3
$95 and works like a champ.

[savio alvares]

could one post the web site for "Nuke" please....

[dlparisi]

http://www.thefoundry.co.uk/

[cpnichols]

The big issue to realize about the EXR coming directly out of Vray Rendring is the way they are written. Vrimg files are written as the buckets drop, and so are the EXRs, since EXRs support that. This is really useful as you can use this to render with your framebuffer being saved to disk instead of holding it in ram until the render is done. Becomes a key feature when dealing with 2k images with 20 layers (that is a lot of ram).

The issue with that, is that the image (vrimg or EXR) is written on blocks, which makes it harder to read in Nuke since Nuke renders images scanline. So it makes Nuke very slow when reading the EXR files as it has to read the whole image before displaying it. For this reason we always convert our EXR to scanline versions before using them in comp. You can do this by writing out an vrimg instead and converting it with vrimg2exr, or you can do this by taking your EXR into nuke, and writing it back out as a scanline.

[jreynolds]

Fusion had some problems with 4.x but 5.2 seems to be handling the extra data in EXRs very well. I haven't used this feature in a large comp but for quick LnR slopcomps and tests I'm having no problems. Fusion 5 also uses an OpenGL based viewport which can be a disadvantage in some cases with a texture heavy Max scene open or while rendering with an OGL based engine in the background but it may be an advantage to what Chris is describing though I can't be 100% sure as Fusion is all I use other than CS2.

[instinct]

just to clarify what i meant above. Nule isnt based around RGBA but around channels organised in layers (1024 is the limit for 4.8 if i recall right).

So basically you have one Read node that has all the masks etc in it. Then you add a grade and use any of the elements as mask for example. This is what i meant with "handling exrs properly"

As for tiled vs. scanline, it is for sure an issue and we too re-convert all vray exrs, but tho it is more appearent in nuke (due to the scanlinebased renderer wich is plain brillient for other reasons) it is a general problem. Reading buckets from arbitrary physical positions in the file based on adresses simply takes more time then blindly reading line by line as needed.

And by "rare cases" for exrs that arent handled well either in nuke i meant something different. Like it doesnt handle metadata from scratch (naturally as it can be just about any data you wish) and i am not sure it handles multires correctly as i never tried, but it just might as well heh

Kind regards,
Thorsten

[sv]

1024 is the limit for 4.8 if i recall right).

So basically you have one Read node that has all the masks etc in it. Then you add a grade and use any of the elements as mask for example. This is what i meant with "handling exrs properly"

or rendering all the different render elements into a single layered exr file (diffuse, GI, Lighting, Reflections etc) then shuffling them into there own parts to have separate control and reconstructing again afterwards

It's amazing you can do all this without splitting into different files - saves so much space and organising....

[cpnichols]

or rendering all the different render elements into a single layered exr file (diffuse, GI, Lighting, Reflections etc) then shuffling them into there own parts to have separate control and reconstructing again afterwards

It's amazing you can do all this without splitting into different files - saves so much space and organising....

Conceptually, a lot of this is based on a pipeline has been used at Digital Domain. So naturally it would work well with Nuke. The reason Vray supports such a system so well, is that many ex-DD people (wink wink) were interested making Vray work within such a pipeline.

It needs compositing software that supports more than 4 channels (RGBA), which Nuke is pretty much the only one that does.... or at least does it well. Older versions of nuke supported 64, and now are nearly unlimited.

[pg1]

We are playing around with using layered exr. files at the moment and I was hoping for some hints and tips on working with layered exrs in Afterfxs.

Our current process is to output layered exrs. directly (i.e no vrimg file) import to afterfx and use proexr to read the extra data which works but is damn slow.

One of the guys here did output some vrimg files yesterday and converted it to exr using the vrimg2exr convertor and loaded those files into afterfx but said that he saw no real speed improvement.
Does that sound correct? follow from this thread the vrimg convertor outputs a scanline exr. which I thought should be faster.

So I suppose my first question would be does vrimg convertor convert vrimg file to a scanline exr by default or do you have to add something to the cmdline to state you want a scanline exr?

Is the Proexr. plugin the only way in Afterfx to read the extra data in an layered exr. file?

Anyone who is using Afterfx with layered exrs. what is your typical workflow, can you resave the exrs from Afterfx as a scanline exr. (as you can do with nuke) and do you get a significant speed increase by doing this?