How to use Auto Tracking in Final Cut Pro

Atti Bear
120 followers
32 Video Views·Dec 7, 2022
Yes, you read right. Final Cut now finally supports autotrack and I'm going to show
you how to use it, how to tweak it and also where its limitations are.
What up guys, my name is Final Cut 10.6 and welcome back to another Eddy Bear.
In today's episode we are going to talk about some of the new features that Apple released
And we finally did receive autotracking capability in Final Cut. This is amazing
because for using something like this you had to resort back to third party plugins in Final Cut
and that sometimes tended to be messy because it either was super slow or it broke whenever
Apple released a new update. But since we now have it inside of Final Cut and it is using
machine learning, it is unbelievably fast. So I'm going to show you some use cases and how you can
actually make good use of it. But there are some caveats which Apple hopefully fixes in the future.
But let's head into Final Cut and let me show you what you can do with it.
I prepared this sample footage where we have the van on the side and we are going to drop a call
out that is going to track the car here. You can use all sorts of items to drop onto your footage.
You can use generators, you can use titles, whatever you have at your disposal you can throw onto your
footage so it's being tracked. The easiest way to do this is you take your designated item you want
to apply to your footage and drop it onto the object you want to be tracked. And you can see
here that Final Cut automatically recognizes certain objects if they're big enough. However
you can also just drop it somewhere in the middle of your footage and then you can adjust your
tracking box accordingly to your subject. However we will do the easier way and just drop it onto
our object. So to make it more precise I'm going to zoom into my footage and move the tracker box
over to the car itself. Now I can adjust the form of the tracker box so I can make it more of a
cylinder and I can also adjust the length and the height of the whole box. So I make it a little bit
bigger so it's more onto the car and I will also adjust the edges a little bit so we don't have so
much unnecessary trackable space here. Okay now here on the top left corner you have the tracking
controls. You can either track left from the playhead, right from the playhead or go into
both directions. We are going to do both but before we do that I'm going to apply my callout
over the whole snippet. Whenever you're triggering the analyzing process it will do it for the whole
clip no matter how long your effect you're applying is. So let's make it easy for sake being
and select analyze. By the way the tracking you're witnessing right now happens in real time so this
was not sped up and this process is the slowest of the two available ones. Now you can see here
that it says combined and I will explain to you what that means in a moment. Also you get a nice
preview here what the tracking does for you and where it is tracking. So let's zoom out and adjust
our callout effect here for a moment. Let's move the text a little bit more down here and maybe
like this so it's a little bit more appealing. Alright let's go to the beginning of the scene
and just let it run. This is how easy it is to track an object. Now let's talk about the analysis
methods for a moment. If you select the underlying clip and you go all the way down to trackers
you can have this drop down menu where you can choose between combined, machine learning or
point cloud. Depending on the scene you're tracking sometimes it helps to switch the
tracking method. In this case we leave it to automatic, select the tracker and hit analyze.
In this instance it's using combined in order to track the subject which means it's using
machine learning and point cloud at the same time to even out some imperfections. Let's see how it
looks. You can see that it is a little bit jumpy and that comes down to the point cloud settings
but you can change these if you go to the drop down menu, select machine learning and hit analyze
again and now you see the faster auto tracking feature of the two because this is real time.
Let's go to the beginning of the clip. It's still jumpy but it evens out the imperfections a little
bit and this is thanks to the machine learning because machine learning pre-predicts where the
movement could be and therefore evens out such little jumpy movements in there. Did you see that?
Did you actually see that? I did not speed up the footage. This is how fast auto track works
in Final Cut 10.6 thanks to machine learning. It's unbelievable what they can do with something
like this. Third-party auto tracking apps for now solely rely on keyframing and pixels. So essentially
you're selecting a bunch of pixels, the app will go through your footage frame by frame, look for
those mentioned pixels and adjust the keyframing accordingly. Depending on how long your footage is
this is tedious and takes a huge amount of time. So the time you're saving with Apple's way with
machine learning is mind-blowing. It's just a portion of the time you needed with
third-party traditional apps. Let's say you're done with your tracking but you're not fully
satisfied the way how it turned out and the subject you want to track is not properly tracked.
Well you can adjust this later with the help of additional keyframes. So move the playhead to a
where your tracker moves off space. In this instance we have this little dot here that is
not perfectly sticking to the car. For this to fix I'm selecting the layer I want to adjust
and just add one additional keyframe here. Now I'm repositioning my frame, continue the playback
and once again I'm repositioning it. The second time I'm repositioning the layer you can see that
there are little arrows showing up. This shows me the way how my layer will move over the time span
here. So as I continue to playback and once again readjust my layer you can see that it will extend
with every additional keyframe I add. And once again here's the last one and I'm moving it right
there. So let's see how this looks now. It starts a little bit off so we are going to fix that too.
Let's see there it comes into frame and boom place it and you can see it adds another keyframe.
Once again playback and now it pretty much sticks perfectly to my trackable subject.
Same with the other layers. Make sure to set one initial keyframe. First select the layer,
add a keyframe, start the playback, pause where you want another keyframe, adjust the framing,
it once again adds another keyframe here, once again pause, adjust your layer, play,
adjust your layer and done. Third layer, same procedure, starting the playback,
adding a keyframe, continue to playback to the point where it does not align anymore,
adjust the layer, playback, adjust the layer, playback, once again adjusting the layer.
Let's go back for a sec and once again all quick and dirty but you get the point.
Let's see how it now looks. Every tracker is now sticking to the individual car and this way you
can even out easily any imperfections within the tracking itself. Another thing we prior needed
third-party apps for is the ability to pixelate people within our footage. For this enter the
effects browser, enter pixel, like pixelate, take it, drop it and you see once it recognizes a face
it will automatically recommend doing face tracking, say both directions. So the auto
tracking that you can see right here is happening in real time, I did not speed it up, however I
will not let you sit through it all the way but I will show you in a moment how fast it is if you
set it to machine learning only. Okay let's see how well it was able to pixelate me.
It was kind of okay. We can also adjust the effect of course. For this go into your effects,
increase the amount and you can have a proper pixelated person within your footage and remove
them later without any hustle whatsoever. So this is really convenient. But now let's see how we can
do this if we set it to machine learning. Once again hit analyze. I'm not kidding you,
this is real time. Boy this is quick. You cannot do this with any kind of other software in this
kind of speed. This is blazing fast. If you do this with whatever else software you have at home
you will not be able to do any kind of recognition or tracking with this kind of speed.
No way. Once again let's check the result. Here you go. Boom. We have a pixelated person
within seconds. So what do you think about Final Cut's new features? Are there any features you
were looking for and still did not receive? What would you wish for to have in Final Cut?
Let me know down in the comments below. And if you liked today's episode don't forget to like,
share, subscribe. Thank you for watching and I see you on the flip side.
