Recap:
-
@jzp presented an example of a production that is using a non-ACES colorspace for grading, and we spent a bit talking through using the gamut compression algorithm in that situation:
- @matthias.scharfenber theorized that it would need to sit at the end of the LMT, right before the OT, as the colorist might want to take advantage of the wider gamut of the other grading space
- However, for VFX, it’s often needed before any compositing work is done (right after the AP0 > AP1 conversion) so how is that handled if it’s also part of the LMT/OT stack?
- Is it okay if it’s applied twice? Or more?
- @nick pointed out that one of the reasons for grading in an alternate space was to avoid these OOG issues, so theoretically, it should become less of a need.
- A question was posed: “What are the feelings of the group if pulls to VFX are done in AP0, but VFX does not invert the gamut compression before delivering back to DI?”
- @jzp: no real concerns, but VFX does have to match dailies.
- @matthias.scharfenber & @mario pointed out that it is a bit antithetical to the VFX dogma of “send pixels back the way you got them + VFX”
- @nick used the example of despill where pixel colors are definitely changing in VFX, but it is accepted
- The transition period might be a bit bumpy, and really well documented workflow considerations and recommendations from ACES are essential
- We talked a bit about how we track whether or not the compression has been applied:
- AMF is the obvious choice, it could be listed as an LMT in the stack, and order of operations would be clear there too.
- @joseph brought up work that is being done to standardize the EXR Reference Implementation which could also help
- @matthias.scharfenber showed a nuke script with examples of applying the compression to images multiple times - we’ll show it next meeting too
- We ended with a talk about inverse transforms, and revisited the possibility of folding the gamut compression into a colorspace transform (either AP0 > AP1 or AP0 > some new space). @carolalynn will pull up the relevant notes from the Architecture group as to why this was not the ultimate recommendation for Implementation to review.