Tweet
If you've seen my other recent thread, you will read that I've been messing about with creating a composite image in Photoshop from the various render elements.
I have discovered one significant issue. You may well be aware of it, and I am probably being slow, but we use Photoshop CS2. This does not allow multi layered 32bit images to be built. CS3 does. When I build an image in a demo of CS3, I get an exact match between the RGB render (as seen in the VFB) and the added (linear dodge) GI, Lighting, Reflection and Specular elements. It is perfect.
In CS2, I must reduce each of the EXRs from 32bit to 16bit before I can begin to layer them up. The result is that the combined layers are quite a bit different from the RGB render. In the simple scene in which I am working, the 16bit version is quite a bit lighter, and the reflections are all brighter.
To test the theory, I converted the layered 32bit file in CS3 down to 16bit and I get the same inconsistency between the layered and the straight RGB version.
This is a pain in the butt as we use CS2 having seen no significant reason to upgrade lately. If we decide to go down the composite image route, it would obviously be a good reason to upgrade, however, I was wondering how many of you out there are working with 32bit compositions. In my limited past experience in working with layered 16bit compositions, I found the overall Photoshop pipeline much slower to work with: bigger files and generally much slower refresh rates when zooming around a 4k+ pixel image.
Is their mileage in the 32bit route, or do you use a more complex pipeline where the 'base composite render' is constructed in one 32bit Photoshop file, but then it is 'flattened' into an 8bit or 16bit file for further PP work (adding people/vegetation etc)?
I have discovered one significant issue. You may well be aware of it, and I am probably being slow, but we use Photoshop CS2. This does not allow multi layered 32bit images to be built. CS3 does. When I build an image in a demo of CS3, I get an exact match between the RGB render (as seen in the VFB) and the added (linear dodge) GI, Lighting, Reflection and Specular elements. It is perfect.
In CS2, I must reduce each of the EXRs from 32bit to 16bit before I can begin to layer them up. The result is that the combined layers are quite a bit different from the RGB render. In the simple scene in which I am working, the 16bit version is quite a bit lighter, and the reflections are all brighter.
To test the theory, I converted the layered 32bit file in CS3 down to 16bit and I get the same inconsistency between the layered and the straight RGB version.
This is a pain in the butt as we use CS2 having seen no significant reason to upgrade lately. If we decide to go down the composite image route, it would obviously be a good reason to upgrade, however, I was wondering how many of you out there are working with 32bit compositions. In my limited past experience in working with layered 16bit compositions, I found the overall Photoshop pipeline much slower to work with: bigger files and generally much slower refresh rates when zooming around a 4k+ pixel image.
Is their mileage in the 32bit route, or do you use a more complex pipeline where the 'base composite render' is constructed in one 32bit Photoshop file, but then it is 'flattened' into an 8bit or 16bit file for further PP work (adding people/vegetation etc)?
Comment