Open Source Color Management

OpenColorIO v1.0.8 documentation


The easy way

While prebuilt binaries are not yet available for all paltforms, OCIO is available via several platform’s package managers.

Fedora and RHEL

In Fedora Core 15 and above, the following command will install OpenColorIO:

yum install OpenColorIO

Providing you are using the Fedora EPEL repository (see the FAQ for instructions), this same command will work for RedHat Enterprise Linux 6 and higher (including RHEL derivatives such as CentOS 6 and Scientific Linux 6)

OS X using Homebrew

You can use the Homebrew package manager to install OpenColorIO on OS X.

First install Homebrew as per the instructions on the Homebrew homepage (or see the Homebrew wiki for more detailed instructions)

Then simply run the following command to install:

brew install opencolorio

Building from source

While there is a huge range of possible setups, the following steps should work on OS X and most Linux distros.

The basic requirements are:

  • cmake >= 2.8
  • (optional) Python 2.x (for the Python bindings)
  • (optional) Nuke 6.x (for the Nuke nodes)
  • (optional) OpenImageIO (for apps including ocioconvert)
  • (optional) Truelight SDK (for TruelightTransform)

To keep things simple, this guide will use the following example paths - these will almost definitely be different for you:

  • source code: /source/ocio
  • the temporary build location: /tmp/ociobuild
  • the final install directory: /software/ocio

First make the build directory and cd to it:

$ mkdir /tmp/ociobuild
$ cd /tmp/ociobuild

Next step is to run cmake, which looks for things such as the compiler’s required arguments, optional requirements like Python, Nuke, OpenImageIO etc

As we want to install OCIO to a custom location (instead of the default /usr/local), we will run cmake with CMAKE_INSTALL_PREFIX

Still in /tmp/ociobuild, run:

$ cmake -D CMAKE_INSTALL_PREFIX=/software/ocio /source/ocio

The last argument is the location of the OCIO source code (containing the main CMakeLists.txt file). You should see something along the lines of:

-- Configuring done
-- Generating done
-- Build files have been written to: /tmp/ociobuild

Next, build everything (with the -j flag to build using 8 threads):

$ make -j8

This should complete in a few minutes. Finally, install the files into the specified location:

$ make install

If nothing went wrong, /software/ocio should look something like this:

$ cd /software/ocio
$ ls
bin/     include/ lib/
$ ls bin/
ocio2icc    ociobakelut ociocheck
$ ls include/
OpenColorIO/   PyOpenColorIO/ pkgconfig/
$ ls lib/
libOpenColorIO.a      libOpenColorIO.dylib

Enabling optional components

The OpenColorIO library is probably not all you want - the Python libraries bindings, the Nuke nodes and several applications are only built if their dependencies are found.

In the case of the Python bindings, the dependencies are the Python headers for the version you wish to use. These may be picked up by default - if so, when you run cmake you would see:

-- Python 2.6 okay, will build the Python bindings against .../include/python2.6

If not, you can point cmake to correct Python executable using the -D PYTHON=... cmake flag:

$ cmake -D PYTHON=/my/custom/python2.6 /source/ocio

Same process with Nuke (although it less likely to be picked up automatically). Point cmake to your Nuke install directory by adding -D NUKE_INSTALL_PATH:

$ cmake -D PYTHON=/my/custom/python2.6 -D NUKE_INSTALL_PATH=/Applications/Nuke6.2v1/ /source/ocio

The NUKE_INSTALL_PATH directory should contain the Nuke executable (e.g Nuke6.2v1), and a include/ directory containing DDImage/ and others.

If set correctly, you will see something similar to:

-- Found Nuke: /Applications/Nuke6.2v1/
-- Nuke_API_VERSION: --6.2--

The Nuke plugins are installed into lib/nuke$MAJOR.$MINOR/, e.g lib/nuke6.2/


If you are using the Nuke plugins, you should compile the Python bindings for the same version of Python that Nuke uses internally. For Nuke 6.0 and 6.1 this is Python 2.5, and for 6.2 it is Python 2.6

The applications included with OCIO have various dependencies - to determine these, look at the CMake output when first run:

-- Not building ocioconvert. Requirement(s) found: OIIO:FALSE

Quick environment configuration

The quickest way to set the required Environment variables is to source the share/ocio/ script installed with OCIO.

For a simple single-user setup, add the following to ~/.bashrc (assuming you are using bash, and the example install directory of /software/ocio):

source /software/ocio/share/ocio/

The only environment variable you must configure manually is OCIO, which points to the configuration file you wish to use. For prebuilt config files, see the Downloads section

To do this, you would add a line to ~/.bashrc (or a per-project configuration script etc), for example:

export OCIO="/path/to/my/config.ocio"

Nuke Configuration

If you specified the NUKE_INSTALL_PATH option when running cmake, you should have a /software/ocio/lib/nuke6.2 directory containing various files.

If you have followed Quick environment configuration, the plugins should be functional. However, one common additional configuration step is to register an OCIODisplay node for each display device/view specified in the config.

To do this, in a on NUKE_PATH (e.g ~/.nuke/ for a single user setup), add the following:

import ocionuke.viewer
ocionuke.viewer.populate_viewer(also_remove = "default")

The also_remove argument can be set to either “default” to remove the default sRGB/rec709 options, “all” to remove everything, or “none” to leave existing viewer processes untouched.

Alternatively, if your workflow has different requirements, you can copy the function and modify it as required, or use it as reference to write your own, better viewer setup function!

import nuke

def register_viewers(also_remove = "default"):
    """Registers the a viewer process for each display device/view, and
    sets the default viewer process.

    ``also_remove`` can be set to either:
    - "default" to remove the default sRGB/rec709 viewer processes
    - "all" to remove all processes
    - "none" to leave existing viewer processes untouched

    if also_remove not in ("default", "none", "all"):
        raise ValueError("also_remove should be set to 'default', 'none' or 'all'")

    if also_remove == "default":
    elif also_remove == "all":
        # Unregister all processes, including None, which should be defined in config.ocio
        for curname in nuke.ViewerProcess.registeredNames():

    # Formats the display and transform, e.g "Film1D (sRGB)"
    DISPLAY_UI_FORMAT = "%(view)s (%(display)s)"

    import PyOpenColorIO as OCIO
    config = OCIO.GetCurrentConfig()

    # For every display, loop over every view
    for display in config.getDisplays():
        for view in config.getViews(display):
            # Register the node
                name = DISPLAY_UI_FORMAT % {'view': view, "display": display},
                call = nuke.nodes.OCIODisplay,
                args = (),
                kwargs = {"display": display, "view": view, "layer": "all"})

    # Get the default display and view, set it as the default used on Nuke startup
    defaultDisplay = config.getDefaultDisplay()
    defaultView = config.getDefaultView(defaultDisplay)
        DISPLAY_UI_FORMAT % {'view': defaultView, "display": defaultDisplay})

Environment variables


This variable needs to point to the global OCIO config file, e.g config.ocio


The lib/ folder (containing libOpenColorIO.dylib) must be on the DYLD_LIBRARY_PATH search path, or you will get an error similar to:

dlopen(.../, 2): Library not loaded: libOpenColorIO.dylib
Referenced from: .../
Reason: image not found

This applies to anything that links against OCIO, including the Nuke nodes, and the PyOpenColorIO Python bindings.


Equivalent to the DYLD_LIBRARY_PATH on Linux


Python’s module search path. If you are using the PyOpenColorIO module, you must add lib/python2.x to this search path (e.g python/2.5), or importing the module will fail:

>>> import PyOpenColorIO
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
ImportError: No module named PyOpenColorIO

Note that DYLD_LIBRARY_PATH or LD_LIBRARY_PATH must be set correctly for the module to work.


Nuke’s customisation search path, where it will look for plugins, gizmos, and scripts and other customisations.

This should point to both lib/nuke6.2/ (or whatever version the plugins are built against), and share/nuke/