Well, many filters don't work at 32-bpc in PS. There's also a misunderstanding about how the alpha channel (in the transparency sense) is used in EXR workflows by the CG/VFX industry. Fortunately, ProEXR solves the EXR issue.
My reference is the image I have in the Vray frame buffer with srgb checked. (gamma 2.2, don’t affect color, input 2.2, ouput 1.0, bla bla bla…)
My guess is we have to have the same look as the image reference inside photoshop. Right?
Have no idea if/how Vray manages color profiles, but ideally, yes. Though this is not always the case - mostly if formats, standards or color engines are different. But let's suppose this is the case.
When I save in TGA no matter what I choose in bpc, it says me it is 8 bis inside photoshop. Wierd…
Tga is 8-bits per channel (RGBA = 8-bits R + 8-bits G + 8-bits B + 8-bits A = 32 bits - but not per channel!). The correct term would be n-bpc, not just n-bits...
When I save in Tiff 16 bit I have to add a gamma correction but I don’t match the image reference exactly.
16-bpc formats are not considered full floating point (FP), and it's assumed that their data have been stored non-linearly, then no LUT (display gamma correction) will be applied by PS. If you apply a simple gamma correction (Levels filter), you won't match your image reference exactly, because the 2.2 gamma exponent is similar - but not the same - as the sRGB formula. There are several ways to solve this, but guess the simpler one is just saving (from your render package) the 16-bpc images with the sRGB profile tagged in the output settings. We save with gamma 1.0 only when the output format is full FP (32-bpc).
When I save in Tiff 32, it looks just wrong.
Guess what is happening there is that this Tiff 32 is 8-bpc. If so, you should be seeing posterizations in the dark areas of your image after gamma correction due to quantizations. It's not efficient storing linear data in 8-bpc files because there's not much bits to store the data where human vision is more sensitive (dark areas).
The ONLY way to match is to save in 32bit EXR and when I load inside photoshop it tells me assign profile:SRGB (DNW it does that for me) and from then it looks ok. I switch to 16 bits ( like TGA whatever I choose in max, it save my EXRfiles in 32 bits) and I can start tweaking my image.
There are some things happening behind the scenes there. When you load a 32-bpc image, PS will assign - on the fly - a linear version of the working color space that it's using. Then, it applies a LUT to show you the image accordingly with the non-linear version of the working color space. Then, when you switch to 16-bits (and don't change anything in the conversion panel), PS automatically will convert from the linear version of the working color space to the non-linear version, and it will disable the LUT. This operation is not noticed because we are always seeing a gamma corrected image.
IS THAT CORRECT? IS THAT LWF?
Nope. As we have seen, you are working in 16-bits with the non-linear version of the working color space. In order to work in linear light you need force a linear space in 16-bits.
To do so you can create a linear profile of your working color space. There are several ways to do this, but the simpler one is probably within PS.
Go to Edit=>Color Settings and in Working Spaces (RGB), choose the color space you want to linearize from the color spaces list:
Then, in the same list (in the top of the list) select Custom RGB...
[img]http://imagic.ddgenvivo.tv/forums/SGCCTools/linPS2.png%20[/img] A window will pop up with some settings to adjust for a given color space:
Change the name of the color space to something like Linear_sRGB or any other color space you have chosen and change the gamma value to 1.0 (the gamma value that we see there for some color spaces are only approximations).
We have there also different white points settings and their description in xy coordinates. We have also the primaries described in some color models. But you'll be ok if you leave this part as it is.
Then, in the same list, below Custom RGB, you'll find an option to save the color profile called Save RGB:
Choose Adobe Profiles folder and now you'll see the lin version of your profile listed in the working color spaces list.
You can then convert to this linear profile when working with 16-bits images.
Idea is that when you set the working color space, let's say sRGB and open the EXR image (without alpha), the color appearance should be sRGB - but remember, PS is using a linear version of your working color space and applying a LUT on the fly. So when you switch to 16-bits (no exposure and gamma adjustments), your working color space is non-linear. Then, in order to work in linear space, you can convert (Asolute Colorimetric) from the non-linear profile of your working color space to the linear version that you have previously created. That's all.
You won't see any difference because when we work with a linear profile, PS applies a LUT to properly display the images. But if lets say, you apply a lens blur, you'll notice the difference in the bokeh effects and such.
Later for saving, convert back to the non-linear profile.
Guess there are several ways to solve this, but if you ask me, and barely you have the chance, try to switch to a compositing package instead. i.e. you might use After Effects for all the RGB work. Its linear workflow is the most automatic thing and the most (if not more) of the filters/FX available in PS are also available (or substitutable) in AFX. Later you could switch to PS or Lightroom for the CMYK phase.
P.D. Photoshop alpha issue with OpenEXR images has finally been solved with the new OpenERX Alpha plug-in.