Focus Breathing - Lens Grid with "Animate by Frames"

Shows a rack focus shot of a lens grid being analyzed using "animaate by frames" parameter solving mode to examine the horizontal and vertical scaling changes produced by lens focus breathing. This process can be used on simple but frequently encountered shots such as a hand‐held but static shot of an actor, with a change in focus.

Script for Search Engines:

Hi this is Russ Andersson. This is the first of two tutorials on lens focus breathing, starting out with a shot with an anamorphic Cooke lens viewing a lens grid. So the shot starts out with the focus pretty close, where the lens grid is, and then the focus goes out to infinity, and you can see how the grid is moving as that happens. Now, I found that it's helpful sometimes to add some additional filtering to the shot, so that it's more consistently blurry and that allows you to get more consistent trackers over the length of the shot. We'll go to the Features panel and we want to have plenty of trackers potentially, but we want ones that are long, and to do that we're going to use the checkerboard neural net tracker. So let's go let that run. And now we need to do some cleanup. You'll see that there's some issues around the edges here. There's also a faint blue line that's a framing rectangle so we'll turn on the Delete mode and the Trackers panel and you need to take out anything that's right on the edge. Those get affected by that edge. And trackers that are right along the blue line also tend to get thrown off as that blue line sweeps on by. It's kind of a whole row of them over here, they get affected. So with that cleanup done, let's move on to the solver, and obviously there's very little motion in this shot, so we're not going to be able to get any kind of real solve. We put the solver into tripod mode. And, there is a zoom here and that's really what you're looking at. The focus breathing is a change in zoom. We're going to change the zoom mode. Sometimes, people like to have it be eliminated. We'll see how to do that later. See, the argument being "oh, it's a fixed lens, let's not have a zoom," and you can move that to scaling instead. So we'll do that later. And, the next thing that we're going to do, we're going to need to set up some constraints here. And to do that, this part here at the end of the shot, it's really sitting out at infinity, and that's kind of the most nominal part of the lens behavior, and we're going to use that as our reference. We'll open up the Solver locking panel, and just turn on the locks right at the end of the shot. We're going to set it to 75 degrees field of view. It's probably even a bit larger than that. That's going to lock that last frame at that point, to kind of set a reference point for the rest of the shot. Now we're just going to tighten up the range of frames that we're going to solve, just to that final section to do our initial solve. So here's our initial solve, we got a grid of tracker sitting out there. And the thing that we need to guard against also, as we start looking at the vertical scaling, is that the entire shot can crush down to a single horizontal line, or extend to some infinitely high or very large high vertical thing, if there's nothing else that says that it can't do that. So we're going to give it something that puts an overall constraint on where things go. And to do that, I'm just going to pick out two of the trackers here diagonally. Maybe we won't use that one there, that's right by the blue line. And we'll go to the Coordinates room. We'll set those guys up as locks. So we're locking them up so that they're going to stay aimed in the right direction the entire time, even as the solver does anything that it it's doing with the vertical scaling and so on. Now we're ready to go back to solve our entire shot again. I'll shift drag the beginning of the shot back to the actual beginning. And I'm going to use a standard anamorphic lens model here. Now you could use a radial plus squash mode that has the vertical scaling parameter that we might need, but the advantage of the anamorphic model is that it has both the horizontal and vertical scaling and that's what's necessary to eliminate the zoom later. Now we're not going to be able to compute any of the distortion coefficients here, because there just isn't enough information in this kind of a limited shot to do that. If you do turn those on, what you see is just the change in some distortion. You can do that it's interesting, but it's not really useful for anything. We're going to take the vertical scale, and I just right click that button and it it shifts through the different modes. So we're going to just keep it at "animate on frame" so we to get a different vertical scaling value for each individual frame. Now we do need to have our constraints be effective. We'll do that, and we'll switch to refine tripod mode as we extend our solve to the entire shot and add all this vertical scaling. So we'll do that and soon enough now we can go look at our results. Here's our field of view track and down here we've got our horizontal scaling. Turn off the trackers there. So this is the solve; you can see we're kind of hanging out doing nothing for a bit, and then we zoom down. There's a little spot, maybe the camera operator had to move a hand on an actual focus ring, or took his hand off the button or whatever, so there's a little pause in it. But here you get to see the actual shape of the zoom motion induced by that focus change. So this is what we're getting if you go and look at the shot now, you'll see that the trackers stay with the grid the entire time. So that's what we're looking for. Now if we want to eliminate that zoom, as I said, and have just a fixed zoom, what we need to do is go back, and we want to do the same thing for the horizontal scale. Now the horizontal scale is is redundant with the zoom; that's it's kind of they're playing in the same pool. So to do that we want to switch the field of view mode to be Known but rather than using that field of view that we just computed, we're going to say "ah, we don't want to use that." We're just going to go and it's at 75 at the end. There's a key down at zero that we don't want to have any effect. So we just want to change this to be 75 degrees. You can scrub through and see that the field of view is staying at 75 degrees. So that's what we want, now we've got that, that's our fixed field of view value, and we're set up to go and compute the horizontal scale. So we'll just continue to refine it, and again we can go back to the graph editor. And now we can look and see both of those two values, and you can see that there's a lot bigger change in this horizontal scaling value than there is in the vertical scaling value. And and you saw that as I was scrubbing through the shot. And in fact the horizontal is here changing by about four times maybe what the vertical is doing, and that relationship is different for different lenses. In some cases they scale mostly vertically, some mostly horizontally. Your mileage may vary. That's why we need to do this. So this is an example of what you get from a lens grid where there are a lot of trackers. Something like this might also be typical for an essentially static shot that has just a rack focus if there's a particularly busy background. So in this shot because of that, all those trackers, we were able to use the full frame by frame parameter solving, because there were a lot of trackers and we wind up with some pretty smooth data. Now the second tutorial shows how to use the "by keys" parameter solving mode on a more realistic shot to eliminate jitter that can occur when there aren't so many trackers and the data isn't as clean. Thanks for watching, have a great day.