Grade changes when Edit sizing bypassed

When grading commercials we normally match the sizing of each shot to the cutting copy, we then do the session with clients. At the end we bypass sizing and render off at source resolution for the online team.

A couple things I have noticed. If you have a power window with a grade in it and change the transform size either in the edit page or with a edit sizing node the window size tracks as you would expect. What does not track is the "Soft 1" control in a hand draw "Curve" window. So if you have a few windows on a shot, maybe on someone's face or a product and it has a very specific "Soft 1" amount, the grade will be different when you "Render at source resolution". You can test this by simply putting up a "Curve" window with a grade in it and add "soft 1" amount, then go to the edit page and zoom in out with the transform and notice the change.

The Lens flare softness also behaves this way. So if you are using a lens flare the grade will be different from what is seen during the session to what is rendered out at source resolution.

The last thing I can see that is affected by this is the old colourist blemish fix, the Over-under. In the colour page if you want to get rid of a blemish you can draw a window around it , then use a node sizing with Keylock on to replace what's inside the window with something else. When rendering at Source resolution or making changes in the Edit sizing transform the window will be in the same place but what is in that window will change based on the transforms in the edit tab -- specifically using rotation, flip and flop.

I had hoped that the fix for all of this would simply be adding a "Node Sizing" node at the end of the tree, so that the sizing of the image happens last, after all blurs, flares and over-unders. Then before the render simple bypass this node and render at "Source Resolution" Unfortunately, for this to work the input sizing would have to be set to "Scale entire image to fit" so you are not losing footage top or bottom. The problem with this is that it seems to remember the aspect ratio. Meaning, if my source footage is 4x3 Alexa footage and I try to push in to a 16x9 frame with a "Node sizing" in the colour page it leaves black pillar box left and right. If you set the scaling to "Scale Full frame with crop" there will be the same discrepancy when rendering to full frame.

So I'm left wondering what exactly the best way to handle this is, I just want to be sure what the clients see during the session is exactly what is being rendered out.

In Baselight I just put my Transform strip at the bottom of my stack which allows me to bypass for render and everything matches perfectly.

Does the Soft Inside / Outside sizing hold any better/differently? I honestly never use the base "soft" setting, because I need to see what the window is actually doing (in the GUI).

That's a great question Marc, The inside/outside mask blur does hold better, perfect actually.
I actually prefer that one but I have talked to a couple people that prefer the "Soft1", for whatever reason.
Makes me a bit nervous that somethings track and some don't.

And just to follow up on the other way of adding a blowup, you can use "Output sizing" to push a 4x3 image to fit the 16x9 frame then add a "Node Sizing" node at the end of the tree to do the actual resize. One issue I have found with this is that the window outlines don't track properly. So the window will be different than what the grade inside is. You would just have to bypass the "Node Sizing", track the window, then add the resize back on.

Anyway, sounds like a Me problem, to get used to. Haha