I posted this in the new users FaceBook forum which seemed to have been a waste of my time. So I’ll post it here. I would be curious from experienced users (looking in your direction Mike Parsons) if I’m on dope or not. My argument is sometimes I find it easier to not always use Action if I don’t want a 3D composite. That I would rather have a CFX node or two (2D Transform notably to be fleshed out a bit) to work like Normal node compositing. Take Blend&Comp node for example. AD added that so you can do simple comps without going into Action. I feel that 2D transform should be made better to make it the “Blend&Comp” equivalent of all DVE/Tracking/Matchmoving solutions in CFX. Or you may say, “Quite whining and learn Action better”
I have a small rant about CFX match moving workflow. I’ve hinted at some of it before, but my thoughts are clearer about it now:
In the small time I’ve worked with Smoke, a lot of the VFX work I’ve done so far is freezing a frame using Timewarp node, then using a paint node to do clone work on it to build a clean plate. Then need to track that clean plate back to the original shot to fill in gaps I make with GMasks. What I have been doing is piping the original shot into a 2D transform node, doing a 2 point track on it, then re-piping my clean plate into the 2D to replace the FRONT input then set the stabilization to INVERT to get a matchmove. It was suggested to me by Grant to use Action to do Matchmoves. Although you CAN do that, I want to say that I find it easier to do what I’m doing with the switching out inputs in 2Dtransform node because of 2 things. One because GMasks are not in Action which is again super odd to me. SO, I tend to avoid Action until the end and want to just work in CFX like a Nuke setup as much as possible where everything is integrated. The other reason I prefer 2D transform to match move is because it has a matte output built in ready to go whereas in Action you have to go “make” one. So, I have a feeling we may never see GMasks in Action (I know they are in media effects in the keyer but I mean actually IN Action to share Matrix data between nodes). I would like to have a STRONG request to “finish out” the 2D transform node.
First, add a BACK input so the motion plate can be tracked from there. Then the FRONT input is what gets matchmoved. If in stabilize mode it can do it from the front or the back. But if in matchmove, it can get the data from BACK to apply to FRONT without having to do a re-piping. I also would want to enable 4 point tracking for a good old fashioned Corner Pin without messing with BiLinears in Action. So there would be a
relationship between the points on Back and Front - source points on BACK link to destination points on FRONT. Nice and simple and stays in the land of CFX where the GMasks live (for now).
Don’t look at me im as disappointed as anyone the direction this box is heading!
The scenario you describe was a DREAM in smoke 2012 using desktop tools. Look at some of my old posts challenging people to do this exact kind of retouch work. Used to be simple in AD and now simpler to do in nuke with a few time offset nodes. Welcome to my nightmare.
Agreed. The more I commit to Smoke the more gotchas I find in the vfx workflow. I’m still trying to see this is a 1.0 release. So hopefully AD will take a look at how people work outside the AD universe. Having said that, even if I staded in the AD universe, it seems I would need flame to do some basic vfx tasks. Anyway, I don’t want to be negative as AD have done huge amounts of work to bring us this release. I’m sure they’ll keep on with the hard work and bring us the tools we need.
I try to be that open minded too - but you know what happens if I say to my clients “hey sorry we can’t quite get there, bear with me it will be better next time”? No one gives a &&$@ and I never see them again. It’s been 8 months of excuses, my patience has gone.
All's well that ends. That's why its called finishing.