any particular reason you are using a camera? if you want to stabilize your footage for roto, use your tracker of choice to stabilize it, do your roto, then introduce the motion back. it all depends on the shot. if you show it to us, we could offer some more help.
I am guessing this is fast movement? you can stabilize it manually using ruler guides as a reference, then roto, then re-introduce the motion. no need for a camera. or you could roto it frame by frame. if the motion is abrupt and fast it will have to be manual. still, looking at the shot moving and not a frame will give us (me in this point) more information. by the looks of it I am guessing this is only manual work.
hey I got icecream screenrecorder too
Yes it is very fast. The way i'm doing this is that i tracked the motion in the right window, and then once that was finished i go into the left window and animate the motion blur ( you know how a lightsaber does that smear look, when it's in motion, well that's what the left is for). But as i go through it frame by frame i have to manually update the right window so i can see where the bottom and tip of the lightsaber are. It would be ideal if i could just parent the camera to that layer (which is already tracked , painfully), so i don't have to go back and forth, and could just work solely in the left window.
I looked at creating a new camera to do this, but you have to enable it for 3d. Which i don't need, 2d for the win. Or do you? I'm not sure how it works.
icecream bums me out since you can't determine the frame rate. they said they will add this feature a long time ago. never did it.
as for you workflow, I am not sure but this looks insane. why not using rotoscoping paths on a solid or shapes? looks like you found points to null useful for rotoscoping? I am not sure how you are approaching this but if this works for you and it won't take months - keep at it.
in my opinion using traditional roto, separating to different masks is the way to go. not using transform parameters or nulls, just plain old vertices.
I'm with Mylenium in not really understanding your approach. What is your end goal?
Personally I'd be doing the roto directly in the main comp, using masks on a solid, unless there's a reason not to.
But to answer your original question directly, I can't see why you wouldn't be able to Parent a camera to the position of the lightsaber layers while you're in the design phase. You'll need to convert all your layers to 3D layers, but so long as you don't change their 3D parameters you can simply turn them back to 2D when you're done.
Rotoscoping is not the right word. Right now I'm just tracking 4 points to meet up with the ends of the lightsaber, then i'll comp everything and then roto it out with masks.
As for parenting the camera in position, it does something silly with the saber and moves it. Not sure why, i just want the position to influence the camera, so i tried connecting the two in expressions: ice video 20171030 163621 - YouTube
oh, maybe it's the z value? How do you connect just the x,y values?
And yeah, i wish iceream could have some framerate options. That would be nice
If you have actual footage and there is camera movement then the way to approach this is to stabilize motion so that the fixed geometry in the scene does not move. Then you do your roto, then you add the motion to the roto layer and unfreeze the motion stabilized layer.
I know that sounds confusing. Here's how to to it in a step by step.
- Select your original footage layer in the timeline and then change to the motion tracking workspace
- Select Stabilize Motion and pick something in the shot that is not moving like the corner of a doorway or a mark on a wall
- Make sure that you have enough detail area selected and then run Stabilize Motion
- When the tracker finishes its work select apply
- The camera movement should be removed and the fixed geometry should not move (note: if there is roll or the camera moves in or out of the scene you will need to stabilize rotation and scale and the end of this procedure is a little more complicated)
- Now do the roto work on a second layer that you are going to use to create the lightsabers (note: a colored layer with the multiply, overlay or screen blend mode will let you see through the layer and easily do your roto)
- When the roto is complete and you have created your lightsabers add a null to the timeline and reveal the position property by pressing p
- Select the stabilized footage layer and press U to reveal the keyframes generated for this layer. You should see keyframes applied to Anchor Point and the tracking keyframes.
- Hold down the Alt/Option key and click on the stopwatch for Position on your null layer
- Drag the expression pickwhip from the Null's position property to the Anchor Point of the original footage
- Make sure that the CTI is on the first frame and parent the roto light saber layer or layers and the Original footage to the null
There you go. The parenting puts the camera motion back in the original footage and adds that motion to the lightsaber layer so everything stays lined up. If you needed to stabilize rotation and scale then name the footage layer 'stabilized" and use this animation preset to add expressions for scale and rotation. Dropbox - destabilize Rotation Scale.ffx
If it helps at all take a look at this quick Roto tutorial that uses the basic technique of stabilizing and then putting the motion back in the shots. One of these days I will spend a little more time and go through the most efficient way to do this kind of work.
Did you end up solving this issue? How did you end up doing it? Did any of our experts assist you in reaching a solution? Let us know!
yes i did figure it out. Andrew set me on the right track, and i ended up working out the rest on my own. I was having trouble connecting the cameras position, and point of interest, to the targets position; since you cant separate the dimensions of the cameras point of interest, and the z value was causing problems. So i ended up just creating a null, and connecting that to the position of the target. Then i just parented the camera to the null, then presto it worked solid. Here is a video of how i did it, if anyone wants to see: