736.ocio_r
Color management framework for VFX/Animation
Jeremy Selan, original author
Various Contributors, via the Academy Software Foundation
The Academy of Motion Picture Arts and Sciences recognizes the importance of open source software in the post-production flows of modern digital cinema. To that end, the Academy Software Foundation (ASWF) was created to increase the quality and contributions to the content creation industry's open source software base, and to provide a neutral forum to standardize and coordinate cross-project and cross-company efforts in visual effects pipeline tools.
OpenColorIO, one of the first Academy sponsored open source projects, is a complete color management solution designed for use in motion picture post-production. It ensures that visual effects artists, animators and others see the same image across a range of software applications, hardware systems and post-production processes. OCIO provides a straightforward and consistent user experience across all supporting applications while allowing for sophisticated back-end configuration options suitable for high-end production usage. OCIO is compatible with the Academy Color Encoding System (ACES) and supports many popular formats of LUTs (look-up-tables). OCIO is integrated and ready to use in many major software packages readily used throughout the industry, including Autodesk 3ds Max, Adobe After Effects, Blender, Autodesk Maya, Foundry Nuke, Adobe Photoshop, Epic Games' Unreal Engine, and Magix Vegas Pro.
In SPEC CPU, the program-under-test is the OCIO bundled tool called ocioperf. This tool is a program which applies a specified color transformation to an internally-generated RGBA image that is either 3840x2160 (4K UHD) or 7680x4320 (8K UHD). The benchmark consists of multiple transformations (which represent colorimetry calculations, gamma adjustments, application of LUTs, color conversions, etc), which are passed into ocioperf to be applied to the reference image.
In addition to the procedurally generated RGBA image that is the input to all workloads, the ocioperf tool has these command line options:
ocioperf -- apply and measure a color transformation processing
usage: ocioperf [options] --transform /path/to/file.clf
--h Display the help and exit
--help Display the help and exit
--verbose Display some general information
--test %d Define the type of processing to measure: 0 means on the complete image (the default),
1 is line-by-line, 2 is pixel-per-pixel and -1 performs all the test types
--transform %s Provide the transform file to apply on the image
--colorspaces %s %s Provide the input and output color spaces to apply on the image
--view %s %s %s Provide the input color space and (display, view) pair to apply on the image
--displayview %s %s %s ( Deprecated) Provide the input and (display, view) pair to apply on the image
--invertview %s %s %s Provide the (display, view) pair and output color space to apply on the image
--iter %d Provide the number of iterations on the processing. Default is 50
--bitdepths %s %s Provide input and output bit-depths (i.e. ui16, f32). Default is f32
--nocache Bypass all caches. Default is false
--nooptim Disable the processor optimizations. Default is false
--8kres Set internal image size to 7680x4320. Default is 3840x2160
--spec-validation-offset %d Additional offset into the image to begin printout
--spec-validation-stride %d Additional stride adder between pixel printout
--spec-validation-pixels %d Number of pixels to printout
CTF is the Autodesk Color Transform Format, and CLF is the Academy's Common LUT Format. There are various .ctf and .clf tranform files available in data/all/input which were sampled from the public git repository and recommended by the community for benchmarking a breadth of behaviors. The files are written in a rich language capable of describing many colorspace transformations. A savvy user can craft their own command lines using the alternative transforms.
OpenColorIO has a rich set of documentation available at opencolorio.readthedocs.io/en/latest.
The output specifies which tests were run, and how many iterations each ran for. The internal image that was transformed is not saved to disk. Instead, a sample of pixel color values are printed out at the end of each transform. The "spec-validation" cmdline options specify which pixels are printed out for validation purposes. This output is used to verify work, and that the transformations were applied in the same manner.
C++
OCIO is single-threaded.
OCIO relies on correct floating-point math, and in particular the way it handles NaN. You can enable correct math evaluations using -fno-finite-math-only, or the more conservative -fno-fast-math flag for the GNU compiler. For other compilers please find the appropriate equivalents for using correct IEEE math. Some information on this can be found at the community's discussion on github: issue 1774.
OpenColorIO is available for download at github.com/AcademySoftwareFoundation/OpenColorIO. The SPEC CPU benchmark started with commit hash 360e66d7 snapshotted on July 17, 2024.
OpenColorIO is distributed under the BSD-3 license.
Individual component packages with different (and compatible) license terms are:
Copyright © 2026 Standard Performance Evaluation Corporation (SPEC®)