Coronas dynamic range

[rendermax]

Hello, can some one pls explain me, what dynamic range of colours use corana? If I save image in 16 bit it has really 16 bit colour depth? Or Iam all wrong?
Thank You.

[Ondra]

I dont understand the question - of course corona uses HDR internally, and allows you to save both HDR and LDR.

By any chance, are you asking because you saw a particular blender video about tone mapping and sRGB?

[pokoy]

Hello, can some one pls explain me, what dynamic range of colours use corana? If I save image in 16 bit it has really 16 bit colour depth? Or Iam all wrong?
Thank You.

That Blender video is so crazily wrong about the dynamic range of Blender and 3d programs in general. I can't understand how anyone would come up with a range of 6,5 EV, I guess someone rendered a white plane and changed the f-stop until the render was black. Anyways, that led to the misconception of limited EVs in 3d programs and it's wrong. If you want, a 3d software can have any dynamic range you wish. You can arbitrarily define a dynamic range and it could surpass any practical limits if you wanted.

Max uses 32 bits internally and did so since its first days (maybe even in DOS times but at least from R1), it's just the VFB that displays colors in 8 bits as a last step.  Even the Scanline renderer uses 32 bits internally.

[rendermax]

LoL, yes that is the reason I asked.
So if I want to have a best control over the colors in my render, its best way is to save the file in HDR or LDR ? I dont know LDR.
I usually save files in 16 bit TIF.

Thank You

[Ondra]

just for my internal bookkeeping since I am not currently at the office and I cannot start another counter.

Current number of people who asked about the incorrect video spreading FUD: 3

[pokoy]

LoL, yes that is the reason I asked.
So if I want to have a best control over the colors in my render, its best way is to save the file in HDR or LDR ? I dont know LDR.
I usually save files in 16 bit TIF.

Thank You

16bit would be LDR, 32bits HDR.

It depends. If you need to tone down highlights or brighten up dark areas, introduce extreme contrast, or if you want to use plugins that produce bloom, glare etc then 32bits is the correct depth. Keep in mind that PS is not a good choice to work with 32bit images. For anything else, 16 bits will be enough.

Generally, 8 and 16 bits are clamped, colors go from 0 - 255, and 16 bits has many more steps in between and will produce less banding in postproduction (65536 values per channel instead of 255). With both, if you want to reduce highlights it'll not work correctly and instead darken the highlights turning them grey.

32 bits have color information well beyond the 0 - 255 space (that space is called 0 - 1 then), so if a highlight is white there's still information that within that highlight there are areas brighter than white. If you tone them down, they'll recover the tones that were previously brighter than 1 and clamped before. Since you're always looking at a LDR representation of a HDR image colors brighter than 1 will not be visible to a LDR display and therefore clamp, but the information is still there.

Best is to make a simple scene, set the lights or VFBs exposure to high values so the image is white. Now save from VFB as TIF with 8/16 bits and again as EXR/HDR and open both in PS. Add an Exposure adjustment layer to both images and tone down exposure. See how in the 8/16 bits image the white color just turns grey (color were clamped at 255, no information beyond this point), while in the 32 bits image you can tone down the exposure like in Corona's VFB which will 'reveal' the image since all the information is there even if it showed you a white image.

The Blender video suggests that there's no information beyond 0-255 space out of the box and 'solves' it by applying tone mapping, which maps the existing colors to the 0-255 space in a more natural way. It's not really about LDR or HDR since Blender internally uses HDR, too. Otherwise that plugin wouldn't have any information to process.

[rendermax]

Thank You pokoy, I will definitely try it. I want to understand this colour thing.

[ac3dartist]

Hi guys, i'm really courious about the theory behind corona color mapping, my core focus is try to simulate as much as possible a real dslr camera behaviuor and for do that the first step was rebuild my own camera LUT using 3DL and a colorchecker.
The first thing i noticed was that, the curve extracted from my canon, compress a bit the highlights. I tried to use that LUT inside corona but it seems that doesn't compress well as the highlight compression value in the vfb.
I decided to modify that LUT (removing the shoulder from the curve to avoid the highlight compression) and using a value of 2 in the vfb highlight compression.
My questions are:
Does the highlight compression inside the LUT file works in 32 bit the same way as the higlight compression value in the vfb?
Is the theory behind highlight compress value inspired to a real camera (and if yes, there's a suggested range value?) or is just a generic value not linked in any way to reality?

Thanks guys