exr in photoshop

[Nekrobul]

Just testd HF\FF absolutley no difference (atleast noticible)

[Juraj]

Regarding 32 vs 16 bit floating-point difference worth it for 3D:

Short answer: No.

Long answer:

30 f-stops (25 - 2), with an additional 10 f-stops with reduced precision at the low end (denormals).

http://www.openexr.com/about.html

If it wasn't for Ondra writting that half-float might clamp up sun, which is analytical light thus it might create the intensity but almost everything else probably can't. Even then, 30 stops is a lot,
compare to 16 which most high-end aftermarket HDRi like CG-Source have.


Deadclown's reponse intrigued me, I didn't even think it mattered that much outside of color post-production. Will test on Z-depth myself

[racoonart]

There are situations where you need 32 bit for zdepth and or World position passes. Since those have tiny value differences it's crucial to have the full precision (sometimes even depending on the scale and world units of your shot/scene).
The average color passes don't need 32 bit (diffuse, reflection, lighting, ...), for "technical" passes however it's sometimes better to chose higher precision.

[racoonart]

Answering on Juraj's edit ;)

You won't probably see a difference if you don't know how to force it. In "normal" zdepth compositing workflows there is practically no problem. There maaay be a minor difference in dof but that's not what I meant.
I, for example, use world position data to displace cards in 3d space and do some kind of 3d compositing inside Fusion (you can either render out a wpp or generate the position data using a camera matrix and a high precision zdepth pass).
If you have a small scene (some centimeres in size) and still have a system unit of centimetes, value differences in world position data can be extremely small. In such cases you see the typical low-precision artifacts in displacements.
It can also occur on larger scenes where values exceed the 16 bit float maximum (had this a number of times) and your world position data becomes clamped and thus useless.

Those cases may rather be rare and pretty predictable but I depends on the workflow and stuff you work on.

[Juraj]

Interesting, completely out of my world but good to know :- ) VFX compositing is very much magic to me.

But reassuring that Photoshop simpletons can keep on conserving harddisk space.

[romullus]

Thanks for explanation, now i can forget about full float whatsoever :]

[pokoy]

Thanks from me as well :)

[Rimas]

I personally save as 32-bit EXR/Tiff just to have those originals in case of a high dynamic range render (exterior+interior, etc). Load up the file in PS, do any exposure corrections or highlight/shadow recovery I need and then use 16BPC for final compositing. The end result is saved at 8BPC JPEG or PNG.

That's how I do it personally.

[Image Box Studios]

Thank you everyone for their supportive comments on this topic.