I'm working on a composite with a green screen shot I filmed handheld. I hand tracked the background plate to move with the foreground element, but I'm wondering, is there anyway to achieve z-depth in Compositelab so I can get a little parallax action happening?

CompositeLab is a 2D compositor so doesn't have depth, but you can simulate parallax by moving elements at different speeds.

Thanks. So for an object farther away from camera, would I stop the move frame or two after an object closer to camera?

They should start and stop moving on the same frames, but the object farther from the camera won't move as far.

Figured I might as well resurrect this thread, rather than start a new one. If I were just doing a simple pan, it'd be no problem to simulate parallax. But if I'm tracking elements to match something shot handheld, how would I simulate parallax on that? The only ways I can think of seem to be grotesquely complicated.

I'm pretty sure the only ways to do it would be grotesquely complicated. The best way wound be to use motion tracking software, but even that isn't exactly a simple solution. Without it, yeah, its gonna be tough and time consuming.

theclone56,
Producing parallax manually may not be as hard as it may seem but you need to understand it first and then break it down.

Put simply, parallax (or prospective shift) involves the difference in speed at which a point or object moves in relation to another point or object further away from or closer to the camera !!!!.

The easiest way to see this effect is to film a few seconds with a camera 'fixed/locked off' but moving along a track left or right at 90 degrees to the camera's view (with the camera not rotating on the tripod).

Try another, lock your tripod down and slowly rotate your camera on the tripod. Have objects closer and further away so you can see how they move in relation to each other and the background. This will probibly be more of use to you as a pan is what your intending.

If you look closly at your clips, you will notice that things close to the camera cross the frame quite quickly, but things much further away take longer to cross the frame (or vice versa with a pan !). This is Parallax and doing a simple test shot like this can give you a great deal of understaning on how added layers/elements should move in relation to the scene's prospective.

With your little problem, you will have to take into consideration horizontal movement, vertical movement and of course camera rotation (Scale may play a part too depending on your clip).

With Fxhome software short of 2D tracking tools i can only see you progressing as follows:

Try to see the horizontal, vertical and rotational parts of the handheld movement as seperate entities. Start, let's say, by animating your 'element' in the left/right direction first, once you are happy with this horizontal movement then move on to the vertical element of the motion and then address any rotation you can see, finally, if there is any in and out movement, your elements scale may need a bit of animation too.

Tedious though this will be, it's about the only way to get a reasonably close match to your camera movement manually. Remember also that depending on the depth you want your element to be into the scene will dictate the time it takes to cross the frame in relation to other objects in there.

I dont envy you with this task but try it out

Regards
Paul

I just toyed around with it and I can't really get anywhere. I wrote down the two quad origin numbers for every frame and observed their variance. I had the idea of adding/subtracting a number from this amount for every frame. I would add a number if the number grew from frame to frame, I would subtract a number if the number shrank from frame to frame.

Obviously if I just added the same number every frame, the effect would be the same, I would've just moved the object. So I thought maybe some sort of a cycle would work. The numbers in the cycle would depend on how close to the camera the object was. The numbers I add would be higher for something closer to the camera, and decrease the further back things got, since the element it's all tracked off of is the farthest from the camera.

So I picked an element that was close to the frame, and decided my cycle would be 0,5,10,5,0,5,10, etc. So first frame I did nothing. Second frame both numbers shrank, so I subtracted five. Third frame both numbers shrank again, so I subtracted ten. Fourth frame they shrank again, so I subtracted five once more. Fifth frame I did nothing. And on from there.

The result was unsatisfactory. The object ended up just jostling about. So I'm thinking what I may need to do is to take the quad numbers for each frame, subtract the quad numbers from the next frame, multiply my results by whatever amount I want the movement to change, five, ten, whatever, and then add THAT number. Or subtract that number if my result is negative. Does that make any sort of sense?

This is a pain in the ass... thank goodness most of the rest of the shots are locked off...