Reverse tracking help…

[Stuart Elith]

For some reason i’m now having trouble doing this too, even without the rotation!
I did a quick setup just to test what you had done (which is how I remembered to do it as well) and i can’t get it to work, yet I have another project using the same expressions that de-stabilized well!

Hmm, more investigation needed 🙂 We must have forgotten a step.

[Stuart Elith]

OK, well i have realised what *I* did wrong, hopefully it helps you too…
The inner comp (with the stabilization) needs to be scaled up so that the footage has plenty of room to move without getting cropped at all. Then, when it gets re-stabilized, it worked for me, even with rotation.

[Michael Mccarthy]

Ok, I understand why we use reverse motion tracking, I understand the technique in AE, pickwhipping rotation with an expression and swapping anchor and position.

What I don’t understand is why / how the technique does what it does. What am I actually doing? The tutorials I’ve seen go through the process, but I still haven’t reached that AHA moment where I’d be able to replicate the process in another app or be able to recognize a problem when something goes wrong.

Can someone point me to a faq or discussion on this? I’m running into a brick wall with google.

Thanks.

[Stuart Elith]

Michael, it seems there isn’t much out there on it, as you’ve found through Google.

As for the procedure, I don’t entirely understand it either. I only use it occasionally, and as long as I follow the procedure I have learnt, it works, and that’s good enough 🙂 There aren’t many variables as far as I can see, so for any situation I can conceive of, it is fine.

But here’s what I have tried to nut out :

The rotation * -1 is basically just inverting the rotation values that the stabilising put in. just like adding a negative in front of a Scale value on footage will flip it.

Swapping the position and anchor points :

Well, on the original footage, the position value isn’t actually changing, it’s the anchor point sliding that moves it. If you pickwhipped the anchor point to anchor point, it wouldn’t be resetting the anchor point to the original spot, it would be copying the same movement I guess. You can’t just do anchorPoint * -1 either, or it will be giving values like -900, -500, which will throw it off the stage.

Now, anchor point and position behave in opposite ways – increasing the position value for X will move it to the right, and increasing the anchorPoint X will move it to the left. Try it… put anything in a comp, look at the position and AP values (which will both be half of the comp width/height) and then add 5 to each… the result will cancel each other out.

So this is all you’re doing by pickwhipping them to each other – the position value takes on the anchorPoint values from the precomp, effectively cancelling them out. And the anchor point taking on the position values – well… i’m not sure if this is actually needed, as it doesn’t change throughout the stabilization. Perhaps in some cases, if the 2 comps aren’t the same size, it might matter, not sure.

Explaining that has helped me work it out in my own head, so that’s cool 🙂 hope it helps.

[Chris Buttacoli]

An easier way to de-stabalize footage is make your stabalized layer a 3d layer, and add a camera to the comp and parent it to the stabalized layer. Simple.