Testing Nuke Plugins
Jeremy Selan <jeremy...@...>
An updated nuke-default profile has been posted that should resolve
all your issues:
Sorry for the trouble!
(I'll be submitting a patch later today to avoid the issue in the future.)
Jeremy Selan <jeremy...@...>
Blake,toggle quoted messageShow quoted text
Thanks for the testing!
* Glad to hear OCIOFileTransform is working. My goal is to support
(at a minimum) all of Nuke's lut formats in the near future.
* We're lucky that the nuke cineon transform happens to match Log2Lin.
The nuke-default OCIO profile is desogned to match all the lut
transforms used in the read-write nodes. (see
src/testbed/nuke_to_ocio.py for how we generated the nuke-default OCIO
profile). Nuke's log2lin node does an analytical transform based on
its configurable parameters, so it's only a fortunate circumstance
they match by default.
* OCIODisplay has it's default display transform set to 'None' by
default in the nuke-default profile. If you switch this to srgb you'll
see the node is working. (This is also why switching input colorspace
has no effect. If you select 'no processing', no matter what the
input is no work will be done).
* OCIOLogConvert doesn't work because the config.ocio file has a typo
in it. You can hand edit it to fix the issue. At the top of the
config is the compositing_log space, which is set to "cineon". But
down below the colorspace is listed as "Cineon". (Note the case
difference). If you fix the role's capitalization everything will
What's happening is that the 'cineon' (lower case) role is not found,
so it's relying on the default role, "data". By design, any conversion
to or from a 'data' space is a no-op, so no processing occurs.
This 'silent error' is a critical mistake; I'll get in the appropriate
library fixes ASAP making sure this doesn't bite any other people.
I'll also update the Nuke profile today to fix the errors as well.
On Mon, Dec 20, 2010 at 2:04 PM, bsloan <bsl...@...> wrote:
Gah. Correction:toggle quoted messageShow quoted text
The OCIODisplay node seems to ignore the 'input colorspace' knob.
Display transform and exposure work as expected.
As for the LogConvert node, switching the operation from loglin to
linlog seems not to change anything.
On Dec 20, 1:42 pm, bsloan <bsl...@...> wrote:
I've tested the OCIO 0.7.3 Nuke suite in Nuke 6.1v3 and have noticed
some possible bugs (or user errors ).
First of all, I downloaded the configs for 0.7 and pointed my OCIO env
var to the nuke-default/configs.ocio.
Then I start Nuke and create an instance of each of the four plugins:
OCIOFile_Transform worked well with the .csp I fed it (Bravo)
OCIOColorspace did the right thing with my cineon image (identical
result to Nuke's vanilla LogLin (Yippee!))
The other two, OCIODisplay and OCIOLogConvert seem to have no effect
(regardless of the values I set in the controls). Could I be doing
BTW, thanks Jeremy et al for all the hard work.