2021-04-27

Owned by Michael Dolan
Last updated: May 10, 2021
3 min read

April 27, 2021

Host: Michael Dolan

Attendees: 

Michael Dolan (TSC Chair) - Epic Games
Carol Payne (TSC) - Netflix
Doug Walker (TSC Chief Architect) - Autodesk
Christophe Brejon
Kevin Wheatley (TSC) - Framestore
Thomas Mansencal - Weta Digital
Mark Boorer (TSC) - Industrial Light & Magic
Alex Forsythe - A.M.P.A.S.
Zach Lewis - Method
Scott Dyer - A.M.P.A.S.

OCIO Config Working Group Meeting Notes

  • CG config proposal:

    • Spreadsheet: https://docs.google.com/spreadsheets/d/1DqxmtZpnhL_9N1wayvcW0y3bZoHRom7A1c58YLlr89g/edit?usp=sharing

    • Christophe: Kept four main ACES spaces. Removed all camera spaces. Kept output for P3D65, Rec. 709, and sRGB.

    • Summary of output color space discussion: Questions arose about whether P3DCI, apple displayP3, P3D65, and HDR outputs should be present. Apple displayP3 could be of particular interest for art directors who use apple devices. Further discussion is needed on non-HDR P3 spaces, but group agrees to support 1000nit ST2084 HDR outputs. Lighters likely don't have HDR display currently, but HDR art direction will grow, and HDR support will be important for game engine usage. Currently both P3D65 and Rec. 2020 ST2084 outputs are enabled. 

      • TODO: Finalize output color space choices

    • Christophe: For config, propose keeping ACES, Raw, Log, and Un-tone-mapped views, and displays for P3D65, Rec. 709, sRGB. Used shorter color space names to make easier to work with.

      • TODO: Add HDR displays

    • Roles discussion summary:

      • Group discussed changing texture primaries to AP1 for consistency with rendering space. Some present have used gamma encoded AP1 spaces for this purpose which is an option, though not an official ACES space. Texture pipelines should generally convert textures to the rendering space so that textures and lights are related. Further discussion is needed to agree on a recommended workflow for this. Should the texture paint role drive texture space in renderer, or the texture painting space which may differ (to be converted to rendering space)?

      • Group debated on whether color picking role should be linear or gamma encoded space. There are tradeoffs either way, and the decision can affect application color sliders due to perceptual uniformity. Further discussion is needed on this point to agree on a recommended approach. 

      • TODO: Finalize decision for texture, color picking, and other roles. Measure pros and cons of each.

    • Christophe: sRGB is input, but is utility in ACES

      • Michael: Could be renamed for this config. Proposed at least.

  • Studio config proposal:

    • Spreadsheet: https://docs.google.com/spreadsheets/d/1nE95DEVtxtEkcIEaJk0WekyEH0Rcs8z_3fdwUtqP8V4/edit?usp=sharing

    • Summary: Next steps are to write email to send to camera vendors to discuss IDT contribution and permissions. We will reach out to all vendors, including those with spaces already in ACES. For some this will be more to provide information on the project and its background and goals, and for others to start a discussion around what can be used. We will also aim to discuss how many of each vendor's formats should be supported, since there are many in some cases. We will ask vendors which formats are most important and choose a limited subset  as a group if no recommendation is given.

    • TODO: Michael will help document background of this config for letter and Carol will finalize and reach out to vendors. Copy Alex on requests for ACES involvement.

    • Doug: Could include Josh Pines from IDT implementation group also.

  • Library next steps

    • Michael: Thoughts on modifying config content via generator API?

    • Kevin: Like for providing additional color spaces, or removing some?

    • Michael: Yes, exactly. A layer or interface for modifying the reference config into other needed configs.

    • Thomas: I'm doing things like that already on my end. Can start working on it.

    • Michael: If we can figure out the API, we can empower community to contribute more. CG will just be removal.

    • Thomas: Removal will be easy, addition will be trickier. Config dictionary is nice abstraction layer right now.

    • Michael: Could use JSON/YAML to have configs for configs.

  • Reference config changes:

    • Carol: Not sure if talked about this in UX group, but should we add family to display color spaces in ACES reference config? Could use more organization in nuke. All of the display color spaces are top level right now. 

    • Michael: Could be a PR to the dictionary driving config generation.

    • TODO: Carol will open PR to add family to reference config display color spaces.

    • Zach: There's also an ACES family with single color space, for CSC.

    • Carol: Probably for Venice.

    • Thomas: Need to pull down latest for aces-dev and regenerate config.

    • Michael: Can also add CI for testing builds against aces-dev

    • Thomas: And CTL image comparison as well. We should probably add these as issues in GH.

    • TODO: Create issues for remaining OpenColorIO-Config-ACES work.