exporting .exr 16-bit instead of 32-bit

[Robert P]

Hey!

We are compositing vray render elements in photoshop using the proexr plugin. Since we are using photoshop cs3 we can not edit the layers since 32 bit is not really supported. So we need exr files in half float, 16 bit.

We are already saving our images as exr directly from vrayrawimg, by adding the .exr extension.

I already read something on the forum about it being possible to save the exr as 16 bit instead of 32 by changing some parameters, but i don't really understand howto do this.

thanx in advance.

[Preisler]

Hi,

I don´t know if you can change the settings for the .exr file when saving directly from vrayrawimg but you can easily change the bit depth to 16 in Photoshop under Image - Mode.

You may know this already but if you save the .exr file manualy from VFB then you can change the exr settings under Setup.

Preisler

[vlado]

You can set this option from MaxScript:

renderers.current.output_rawExrUseHalf=true

This is the default value too - by default, raw .exr files are written in 16-bit mode.

If you use the vrimg2exr tool to convert a .vrimg file to an .exr file, use the -half option on the command line.

Best regard,
Vlado

[vlado]

You can set this option from MaxScript:

renderers.current.output_rawExrUseHalf=true

This is the default value too - by default, raw .exr files are written in 16-bit mode.

If you use the vrimg2exr tool to convert a .vrimg file to an .exr file, use the -half option on the command line.

Best regard,
Vlado

[RErender]

hmm, just checked one of my exr's in PS and it says that it is 32bit, renderers.current.output_rawExrUseHalf is set to true in the file, maybe the switch is not working?

Another oddity I noticed today is when I changed the bucket size on one of my frames from 64 to 6 to getthe displacement to stop hanging and that EXR ended up being 26.5 megs where all the others are around 19 megs, is this expected?

[Shimakaze]

Your problem isn't actually the EXRs being exported but the way Photoshop handles them. 16 bit "half-float" isn't the same as 16 bit color images.
Whenever you import any floating point image into Photoshop it will always treat it as a 32bit image. No matter if it's full or half-float. That's the only way Photoshop can handle the full color range. If you want to do something with the image that you can't do in 32bit then you can downsample in Photoshop easilly.

[RErender]

Your problem isn't actually the EXRs being exported but the way Photoshop handles them. 16 bit "half-float" isn't the same as 16 bit color images.
Whenever you import any floating point image into Photoshop it will always treat it as a 32bit image. No matter if it's full or half-float. That's the only way Photoshop can handle the full color range. If you want to do something with the image that you can't do in 32bit then you can downsample in Photoshop easilly.

ah, makes sense

[Robert P]

Thank you for all the advice!

My problem was when i changed the file from 32bit to 16 in photoshop the linear add mode didn't work correctly anymore. Changing my photoshop colorsettings for my working space from the standard profile to one with gamma set to 1, it solved this problem. Makes sense since photoshop is not linear by default.

Still photoshop is far from the best program to use for comping. But i do really like all the extra options comping gives you in postprocessing images, we are now looking into other programs like Nuke and combustion.

Which one do you use? Any advice?

thanx

[instinct]

Depends a bit on what you do on a daily basis. Nuke is great, and a full linear pipeline is sure fun, but if you're retouching (say painting) a lot then nuke prolly isnt for you at the moment.

For CC/postPro/EXR workflow it sure is the only tool out there to properly handle multichannel EXRs.

Regards,
Thorsten

[Robert P]

That's the thing, there's a 'slight' difference between what i do on a daily basis and what i want to do every day. And since business is not as good as last year i've got some left over time, giving me the opertunity to go from Robert 1.4 RC3 to Robert 2.0!

So what's the big advantage of 32-bit over 16-bit half float?

[instinct]

Higher Precision. We usually go with Halffloat as it seems sufficient. It's a floating point datatype using 16 bits per channel. Wich is NOT the same as 16bit per channel integers!

You can read up on the technical details (if you are savy with floating point maths) in the openEXR doc's Technical Introduction.

Regards,
Thorsten