- Posts: 456
- Joined: Thu Sep 16, 2021 7:05 am
- Real Name: Cody Predum
I must be missing something about corner pinning with the planar tracker, because this seems like an issue that would be both extremely common and long since solved. The question is essentially: assuming the footage you're working with wasn't shot with markers, how can you get the image/footage in your composite to have the correct aspect ratio in "real-world" space?
I'll clarify this some more just in case: Often, the points used for defining the corner pin will be different than where the corners of the composited image will be. This can be for several reasons, but a very common one is because you need to use lines known to be real-world parallel to define the corner pin so that the perspective is correct, and these are often not where the edges of the comped image need to end up. For a simple example, imagine a shot that includes a blank whiteboard (at an angle, but with all corners visible). You can use the corners of the board to easily get the corner pinned image to be at the correct perspective angle; but if your image is, for example, a square and the whiteboard isn't, how do you make the image appear to be a square at an angle if your corner pin doesn't represent a square in the real world?
I think the main problem for me is that corner pin stretches the pinned image to fit within it. On its own, this isn't necessarily the problem, since adding a transform node between the media and the tracker will cause size and position/center adjustments to happen in the perspective defined by the corner pin. But since the corner pin stretches the image to fit, you need to adjust the image's x and y size separately to un-distort the image back to normal, and you can only do this by eyeballing it, which is almost guaranteed to be at least slightly wrong.
What really makes me think I'm missing something is that the existence of the "show grid" feature means that Fusion already knows the aspect ratio of the world-space rectangle defined by the screen-space corner pin. The idea of having to imperfectly eyeball something that I know the software already has the data for really makes me think there's a simple feature I'm not aware of, especially since the need to composite something with the correct aspect ratio into footage that wasn't shot with markers seems like an extremely common task.
Someone please help me out! There's no way this isn't already a solved problem. Thanks!
I'll clarify this some more just in case: Often, the points used for defining the corner pin will be different than where the corners of the composited image will be. This can be for several reasons, but a very common one is because you need to use lines known to be real-world parallel to define the corner pin so that the perspective is correct, and these are often not where the edges of the comped image need to end up. For a simple example, imagine a shot that includes a blank whiteboard (at an angle, but with all corners visible). You can use the corners of the board to easily get the corner pinned image to be at the correct perspective angle; but if your image is, for example, a square and the whiteboard isn't, how do you make the image appear to be a square at an angle if your corner pin doesn't represent a square in the real world?
I think the main problem for me is that corner pin stretches the pinned image to fit within it. On its own, this isn't necessarily the problem, since adding a transform node between the media and the tracker will cause size and position/center adjustments to happen in the perspective defined by the corner pin. But since the corner pin stretches the image to fit, you need to adjust the image's x and y size separately to un-distort the image back to normal, and you can only do this by eyeballing it, which is almost guaranteed to be at least slightly wrong.
What really makes me think I'm missing something is that the existence of the "show grid" feature means that Fusion already knows the aspect ratio of the world-space rectangle defined by the screen-space corner pin. The idea of having to imperfectly eyeball something that I know the software already has the data for really makes me think there's a simple feature I'm not aware of, especially since the need to composite something with the correct aspect ratio into footage that wasn't shot with markers seems like an extremely common task.
Someone please help me out! There's no way this isn't already a solved problem. Thanks!