Bit of a leechy sorta question…what the hell worth a shot.

I'm wondering who's had experience with file format bit depths out of mantra into Discreet compositors, and is willing to share any practical information. Things I know:

Discreet remaps everything to 12 bit linear integer internally(tied to hardware - bleh).
Discreet doesn't support OExr.
Discreet input bit depth of choice *used* to be cineon…don't know now.


We'd be rendering images out of mantra, but we obviously need the extra bit depth for film work. What I'm thinking now is using OExr for rendering and pre-comps in Houdini, then converting the layers to something else(?) for final assembly in Inferno. Unfortunately we're over-rendering, but then there's really no choice since the oddball depth of 12 isn't supported anywhere else anyway. With 16 bit float we'll get more “play” than we'll need, which is good, but I'm curious about how that can be remapped into Discreet.

If anyone's been through this mill and knows a good balance between capability and render times, I'd appreciate any tips/horror stories…

Cheers,

J.C.

OK, let me rephrase the question a little more specifically…

If I'm rendering out OExr from mantra(16 bit float) and I want to be able to import it into Discreet as a cineon (10 bit int, I believe), how would I go about making up a LUT to handle this?

But yes, any other thoughts are welcome…

Cheers,

J.C.

Have you tried converting the exr file to a tiff using icp? It'll convert the 16 bit float to 32 bit float. I'm assuming that Discreet can load tiffs, and hopefully supports loading 32 bit float tiffs (though it may immediately convert it to 12bit upon load).

The downside is the file size pretty much doubling.

Yes, I had sort of tried that, but there seemed to be some issues with the mapping, thus I thought perhaps a LUT was needed. More specifically, I took the standard kodak reference image in cineon format, read it into Houdini, then saved it out as a 32 bit image, and when reading into Discreet there was corruption in the hot spots. I keep thinking LUTs are related to this - but perhaps that's only needed in a non-linear colourspace? I'm still reading around about this…but any thoughts are appreciated.

I keep thinking there needs to be a 16to12.lut file that's needed to properly remap this - but let me try the space-unfriendly method you suggest precisely. I'm not particularly worried about the filesize, but I'll admit the bandwidth is troublesome.

Make no mistake, though, I think this is all Discreet's fault. Bloody 12 bit.

Cheers,

J.C.

Why not ‘icp’ from EXR to Cineon? You can then at least set white/black points (white being the likely culprit) and gamma and experiment pretty easily…

Cheers,

Peter B

The problem with exr->cin is that the cineon conversion doesn't play well with FP images. It assumes that the data is in 16 bit fixed point, between 0 - 1. You can generate a LUT using

icineon -i 65535 out.lut

for use in

icp -O out.lut file.exr file.cin

which will work for FP. However, it cuts off the HDR information (values > 1). Perhaps this is okay if it's going into a fixed linear 12 bit system, but if you want the HDR info, it's a no go.

Now that HDR images are common, I'll have to fix the cineon exporter to properly support FP images.

Excellent guys, thanks. This is precisely the stuff I'm looking for. It's hard to tell which issues are mine and which is the software's…

Cheers,

J.C.