Recap:
- On OCIO 2.1 Implementation:
- @KevinJW: definitely want a user-driven version, unsure how to do that with the proposed OCIO implementation
- @matthias.scharfenber: The parameterized version should be provided by the prescribed application, as the ACES-labeled implementation in OCIO should be standardized
- @nick: There should likely be a discussion between OCIO and Foundry etc - on whether or not to expose FixedFunctions to the user
- @KevinJW : We’ll need some tests for the parameterized version too. What’s the spec? How do we define? What happens if I pass a smooth gradient, do I still have it?
- @matthias.scharfenber: Quick unit tests should be enough for the standardized ACES version: forward, inverse, CPU and GPU.
- How does an application meet requirements for the logo program for this algorithm?
- @sdyer : reference images already exist for the gamut compress algo, the still life and the synthetic chart. Not sure if the inverse exists, will check. We can include other reference images if the group wants.
- @michaelmaloney333 - everything eventually gets concatenated into a 3D LUT - both in onset box and as a viewing transform for VFX
- @nick: two different tiers of validation in CLF - preview and finishing, could be valid here for implementation of the gamut compression algorithm as well
- @jzp: Yes, differences between CPU/GPU small form errors might not mean an exact 0 diff match, and we’ll need to account for thresholds. Have we tried it through the CLF strawman? Might be interesting to try it.
- @carolalynn: @ptr couldn’t make it today, but he’s working through an implementation in Livegrade, currently it’s a checkbox on the input. More discussion to be had on placement, checkbox vs layer/node in a stack.
- @jzp : is LMT the right catch-all for all of this? Should there be another slot? Input LMT and output LMT?
- @carolalynn: is there anything in the ACES specification that defines where an LMT must be applied now? As in post-most-work etc? I will find the notes from where this was discussed in the Architecture group. There we ultimately landed on not rocking the boat and sticking with LMT.
- @Giardiello - if Livegrade exports an LMT in AMF based on a checkbox, how does downstream software handle that?
- @KevinJW : how do you distinguish between baked in and not baked in to the exr?
- @Alexander_Forsythe : AMF has an “APPLIED” flag in the xml per element - should work perfectly for this use case. This flag is in active pipelines, and different from archived pipelines.