1 person found this helpful
To smooth a shot you don't need timeremapping, but you need tracking as in your idea.
Add a nul object, and a camera. Link the camera to the nul object.
Make your tracking and apply it on the nul object. Now add the smooth() expression on it's position value (you can find the smooth() expression with it's default values on the expression menu.
Now you can enable the 3D switch on your background layer and you will have a smoother pan.
Thanks for your input!
Does the method you described work also in a situation in which the camera is first panning, then it literally stops and then starts panning again? I mean is simply smoothing out the keyframes able to make that kind of a pan constant in speed?
Well with this technique you can smooth a not so smooth pan. But if you have a long stop in the middle of your shot, a second or two for exemple, there is nothing much more you can do sadly.
The problem is that whatever you'll do, you'll still miss the panning data for it. Even with the smoothing methodology, you'll have to zoom in order to avoid gaps around the borders.
Sadly, unlike some big producers and directors might think, you cannot always fix everything in post.
Not that what you are asking is impossible, depending on your shot, but it would involve so much hand work, cloning, replacing, masking & rotoing, that you'll better have to reshoot it.
Thanks for helping, I appreciate it!
But in the case we are talking about, couldn't it be possible to first track the footage and then use expressions to drive time remapping so that time remapping would basically try to have the track points moving at constant speed? Isn't this possible in theory (I have no idea about the actual math/expressions though)?
Well I don't see how time remapping can help in trying to move the point at a constent rate without causing problems related to time. I guess it will be easier to adjust the speed with the graph editor, but the longer your pause in the middle, the more you'll have to zoom to avoid gaps in your final comp, and you loose a lot of data.
Be as far as i'm concerned, i would definitly not use timeremapping.
I saw someone else thinking about the same solution on another forum so I decided to bump this a little bit.
The way I see it, this should be doable (unless there is something that I really dont understand).
The way it would work is this:
You have a panning shot of a stationary scene (no moving objects). You choose a track point that is visible during the pan and motion track that point. Then you look at the speed of that track point (how fast it is moving ie. from left to right) as pixels per seconds. And then you tie it to the time remapping property to make the speed (pixels per seconds) move in a constant manner.
The way I see it, there should be no black gaps when using a method like this, which kind of is why I'm interested in the consept. Because you are not moving the frame around spatially, you are simply changing the time it takes for a point to get from A to B.
What do you guys think?