davinci shaper luts, ocio and nuke

[Gresham Lochner]

Hi all,

Hoping that someone here will be able to point me in the right direction.

Recently I've had DI using Davinci Resolve export a cubed lut from their
color grade. When applying this in nuke, I'm getting a slight
colorshift. After reading through some OCIO documentation as well as
searching online and a post on the nuke list, it looks like Davinci is
using a pre-lut (shaper lut) which is similar to a linear to cineon
conversion.

When looking at the shaper lut from Davinci, it looks like a
proprietary, binary file.

I noticed in OCIO documentation that OCIO only supports linear
interpolation for a shaper LUT application.

Has anyone successfully brought in a colorgrade from Davinci into Nuke?

One hack could be to export the CMS test pattern, bring it into davinci,
apply the grade to that and re-export it to nuke and generate the lut in
nuke....skirting around the Davinci shaper lut issue. But I'd prefer to
do it in a less....hackish manor.

Any ideas would be greatly appreciated.

Thanks,

- Gresham

[Jeremy Selan]

Gresham,

I'm happy to take a look at it.

Would you send me (off list) the example LUT? Please include, if
possible, a small example image (prelut), and then the same example
with the reference color processing baked in. Bonus points if your
example lut/images are not proprietary (i.e., suitable for posting to
our git repo as a format example).

The 'linear interpolation only' for shaper luts is a bit out of date.
We do support spline interpolation (currently internal to the .csp
format) but it could be generalized as needed.

Our goal in OCIO is to have 100% LUT support for all common industry formats.

Also, keep in mind that sometimes errors of the kind your seeing are
inherent to the creation / sampling of the initial lut (i.e., the
export phase), not their usage. If this is the case, you'd see the
same quality issues when using the CMS test pattern approach. Probably
worth testing.

-- Jeremy

[t.spe...@googlemail.com]

Hi Jeremy, Gresham,

I appreciate it's some time since this question was posted!

We've just bumped into the same problem.  Did anything come of this?  I looked at the binary file (extension .cube_shaperlut) and it seems to be a 28-byte header (with the cube size as an integer at offset 16) followed by the correct number of single-precision floats.  I wrote a short conversion tool that assumed the floats were in the same order as a .cube, but I've no idea what to do with the resulting LUT, if it's even correct (my colour science knowledge is next to nil).

If the slight colour-shift is not a result of the initial LUT export, we'd love to correct for it.

Many thanks,
Tom

[Gresham Lochner]

Hi Tom,

Haven't heard anything from Jeremy yet, I know he's been really busy. Admittedly I haven't had time to follow up.

I've heard rumor that Davinci 9 has resolved the shaper lut issue and able to export usable .cube's.....

However I'd still love if there was a solution to the shaperlut issue that didn't involve Davinci 9.

Cheers,

- G

[t.spe...@googlemail.com]

Hi Gresham,

Thanks for the reply!

We're actually using DaVinci Resolve 9 at the moment.  Unfortunately we'd not thought to try this when using Resolve 8, so I don't know if it's the same or better.  All I know for sure is that the colour shift is slight enough for us to live with it (though we'd prefer to adjust) and that I can't imagine a use for the .cube_shaperlut.

If you'd like us to export a .cube of a supplied DRX from Resolve 9 for comparison it might answer a few questions?

Cheers,
Tom