Photoshop - 32bit - Exposure vs Camera Raw Exposure

[CiroC]

I am just trying to understand why this difference happens

Normal Exposure Tool in Photohsop - Exp1.jpg

and the

Camera Raw Exposure - Exp2.jpg

I saved the file as OpenExr 32bits Gamma 1.0. Does anyone know why this happens?

[grafichissimo]

Yes, I know the problem, Camera Raw is working differently, I am not sure CRaw is managing  32 files very well.
Instead use After Effects if you can.
You can always change the exposure than convert the image to 16bit and then use Camera Raw.

[CiroC]

Exactly.

Initial I thought that it was assigning another profile to the image, but it doesn't make sense. I end up by using the tone mapping curve to achieve something closer. If someone has a solution, please let me know.

[CiroC]

And I now, I just updated Photoshop and I can no longer open the image in Camera Raw. Great

[Juraj]

The exposure in ACR was never linear, regardless of in which mode you opened your image, so think of it more as "brightness".

Regarding the recent Cloud update :- D Oh Adobe, instead of fixing the ACR so it could work linearly, they just put restriction there.
This is btw just consistency for them, Lightroom(using same internal ACR engine) had always this restriction.

If it's any consolation, ACR can still access a LOT of dynamic range when you convert your PS mode to 16bit (after all, photos use it to shuffle 14 +/- dynamic stops these days).
You don't loose much since ACR didn't have linear tools anyway.

[CiroC]

Thanks Juraj

What I really liked was that tonemapping curve, but I could see that something was right. I really don't understand why Photoshop can't work with 32bits images as After Effects does.
I used to convert the render into a smart object, which allowed me to go back and forward with any adjustments. I think I will have to redo my workflow, but the CoronaImage Editor is actually a good tool.

[totinguis]

Have you tried recomposing the beauty by adding linearly your render elements? You will have a lot more control than just correcting the beauty in exr. Then you can apply any tone map curve you want. I mean, if you are already working with exr... what I do is correct channel by channel basically by means of exposure adjustments and then, just as Juraj said, correct with camera raw the result of the 32bit psd applied and duplicated to a 16bit PSD. You can have smartobjects and everything...

[CiroC]

I actually do that. I create a smart object with all the passes I want (Linear Dodge(Add)), then I apply the Camera Raw and tweak the exposure and most important the Tone Mapping Curve. I need to test the results with a 16bit file to see if I can tweak the highlights and shadows. I was trying to create a workflow that allows me more flexibility, but this move from Adobe is just annoying. Maybe I will just move to Affinity

[Juraj]

You absolutely can use 16bit mode to tweak highlights, it might just behave differently. I don't know what the exact ratio is (somewhere I read that 16bit can hold 16bit of dynamic stops... not sure if just coincidence or what since 32bit. hdr has 256 while .exr just 32), but 16 stops of dynamic range is from fully black only sun, to fully overbright.
You would be working exactly as .raw photography.

BTW, important thing to know about smart layers, you probably know but generally, they apply in current mode. Even if your smart object is 32bit, the layer will be applied in 16bit mode if that's your current environment. Even if you make preset in 32bit, the result will be different.
I also keep smart layers (and smart layers within smart layers :- D ), but it's for flexibility, it cannot cheat the PS rules about 32bit.  (just writting this because there was numerous people on cgarchitect who thought it can)

[romullus]

Unless you're talking about exr, i don't think that saving in 16 bit will give you any advantage on highlight recovery, than saving in 8 bits. It still clamps dynamic range at 0-1. Yes, it has a lot more information in that range, but anything that is beyond it, is unrecoverably lost. It just no match to 32 bit imagery at all.

[CiroC]

BTW, important thing to know about smart layers, you probably know but generally, they apply in current mode. Even if your smart object is 32bit, the layer will be applied in 16bit mode if that's your current environment. Even if you make preset in 32bit, the result will be different.
I also keep smart layers (and smart layers within smart layers :- D ), but it's for flexibility, it cannot cheat the PS rules about 32bit.  (just writting this because there was numerous people on cgarchitect who thought it can)

Yes I know. My work flow is like this. Tweak the image in 32 bit, add render passes, etc. Save the file. Create another file as 16bit and link the previous one.This is one of the reasons why started using Fusion, but isn't the same think. I don't find as flexible as Photoshop.

[Juraj]

Unless you're talking about exr, i don't think that saving in 16 bit will give you any advantage on highlight recovery, than saving in 8 bits. It still clamps dynamic range at 0-1. Yes, it has a lot more information in that range, but anything that is beyond it, is unrecoverably lost. It just no match to 32 bit imagery at all.

No, I don't mean half-floating ('16bit') .exr, but that is the mode I save to save harddrive space :- ) I meant changing the mode in PS from 32bit to 16bit. Here is what I believe happens from my experiments:

Everything inside regular Photoshop environment get's clamped (0-1), the tools (exposure,etc..) can't access any dynamic range but unless you used the tonemapping mode the information shouldn't be lost. It's should be similar as saving .raw files (.nef,etc..) as 16bit tiffs and using that format for post.

But I don't argue it's anything similar to linear compositing and post. But it does allow for highlight recovery through its tonemapping tools (ACR only).

Edit: Honestly now I am not sure anymore. But one interesting solution to our problem could be corona .crx format that behaved like .raw files. Raw files are just 12-14bit mostly, Hasselblad recently did 16bit, but it's linear format internally with embedded curve (could be regular sRGB gamma curve like we use in rendering). But AdobeCameraRaw can read it normally.
16bit .exr isn't the same as Photoshop and other applications see it as regular 32bit file.

Or even better, saving to .dng directly.