Need help for 3d stab

Please help me explain more about this tutorial how to do ?


To add the motion back after stabilization you want to duplicate the action that you used for the stabilization. That’s gonna go downstream. You’re gonna do all your comp stuff you need to do on the stabilized plate and then plug that, with it’s matte, into input 1 of the duplicated action you made downstream. The key here is to change the result camera in the downstream action not to the camera that has the keyframe kept, but to the camera that has all the keyframes still. (Be careful when doing this, result camera selection is Keyframeable)

1 Like

If you’re just stabilizing or smoothing the camera movement, and not doing any paint work/reprojection, you can skip the second action node and do it all in one action:

  1. 3D track the scene

  2. add an object (I like using a large sphere that encloses the camera, but a white frame or hunk of geometry placed out in front of the camera usually works too).

  3. add your source image as a diffuse map, set to projection from your 3D tracked camera.

  4. duplicate your 3D tracked camera and either average it (for smoothing the camera move), or “keep” the first keyframe, or your favorite keyframe (for more thorough stabilization).

  5. set your new camera as your output camera.

Hope that helps!


This one is easier to follow

This one was actually included in the post before it was edited and removed. I’m not sure we’re understanding what you’re looking for @gumpmick? Is it the stabilizing that doesn’t make sense? Or the re-introduction of motion?

I find it helps in terms of projection to put action into a physical space in your mind. What is happening when you’re projecting is: you have a tiny perfect invisible short throw projector (your projection camera shooting out your diffuse map) that is bolted on to the object you’re projecting on (you’re surface/card/geometry). The camera you’re recording all that through, that’s moving around, and whose output seeing in your monitors on set is the result camera.

When we’re stabilizing, this relationship changes. Our camera is now completely locked off. We have no movement. BUT the projector is in motion and painting with light over the geometry of the scene. Anywhere the projector has not touched is black. No light. We can see this very clearly when this technique is used on a more complex piece of geo like a landscape that doesn’t enter fame until later in the shot. Because it’s not “entering frame” where our camera that’s just locked off is, our projector has not reached it yet. The texture, light, color, everything is only illuminated when the projector passes over it. So once the projector reaches that with the camera track data we have, it paints it on.

(Edit: I have no idea if this has anything to do with real world physics, this just how I visualize it and may be completely wrong)

This guy might as well upload his footage and we do it for him. In the end that is what he wants without asking for it.