Multipass psd-manager and openexr

[3D Betrieb]

Hello together,
we are getting crazy with all the multipass-stuff.
All our tests we tried in LWF and traditional way.
What we want is a 16bit photoshop-image, that can be adjusted easily in ps with some layers and object masks as diffuse, shadow, gi, and so on..
We tried to figure out the management of psd-manager with the demo-version, but the output is too bright in ps. The final image is not matching the frame-buffer. The difference is very extrem and the work is not comfortable with that problem.
In a second test we tried the openexr demo and its great, but all the stuff is 32 bit. The result is matching very good in ps, but all that stuff is 32 bit so we couldnt use most of the ps tools.
When we try to convert it in 16 bit, the result is to bright too.
What would be the best and comfortable way to manage multipass-still-images in vray?
Thank you for your advice,
best regards,
Stefan

[3D Betrieb]

okay, we figured out, that it depends on the 32bit!
If we use a 32bit image everything is okay in ps (with openexr and psd-manger)!
If we convert it in ps (cs4) in a 16 bit image the layer mode "linear dodge (add)" (linear abwedeln hinzufügen) doesnt work correct anymore.
So it seems to be a ps-issue.
Has anyone an idea how we could get the correct result in 16bit?
best regards,
Stefan

[DanielBrew]

Without knowing a little more about your workflow it is difficult to say exactly.

First, are you using 'linear multiply' in color mapping and do you have 'Don't affect colors' checked when you render the image.

[3D Betrieb]

hello Daniel,
I think it doesnt depend on the workflow in max, couse in the meantime the image seems to be okay, but only in 32bit. When I reduce the colors to 16bit mode in ps all the layers wont work as in 32bit.
Then the image gets to bright..
thank you,
Stefan

[DanielBrew]

The reason I asked is that some applications automatically display some gamma 1 images in sRGB space. I think it dependeds on the image type and the colour depth. I don't know whether photoshop behaves like this.
I know that Combustion displays RPFs differently to EXRs for example.

I was just trying to work out whether the images you were compositing with were in linear space or not.

[simmsimaging]

The layers do work differently in 8/16bit from 32bit - especially Linear Dodge. If you want it to work "properly" you need to put your layers together in 32 bit and rough out the look and then merge them before going to 8 bit.

If you do not want to do that then you will have to try and adjust the layers by hand to make them work more like 32 bit (use curves or levels on linear dodge layers etc.)

A 'better' way, and my usual method these days, is to comp the layers in 32bit, then select all the layers and convert them into a single Smart Object. You end up with a single SO layer in your file, and then convert that file to 8bit. This lets you embed your full 32bit images in your 8bit one, and you can always go back and tweak the layer blending in 32bit later. It works well, but the files can get big if you work in hi-res.

b

[3D Betrieb]

Hello Daniel,
hello Brett,
thank you for your advice.

@daniel: yes, we are using the linear multiply. The "dont affect colors" wasnt checked, this was a very good tip! Now we get a 16 bit file in PS that works. The only problem is now, that the image only looks correct in vray frame buffer. The image output only works with the max framebuffer. So we have to switch the framebuffer for output.

@brett: yes, youre right! Espacilly this problem i was wondering about!
Thank you for sharing your workflow!

I think our problem is solved thank you!
best regards Stefan