https://www.youtube.com/watch?v=InIuTtt7W3E&ab_channel=CGMatterhttps://www.youtube.com/watch?v=InIuTtt7W3E&ab_channel=CGMatter

>> Hey guys, welcome back to
another Blender 2.8 tutorial.

And today I'm going to be showing you
how to do 3D camera tracking all inside

of Blender without any add-ons necessary.

Now because 3D camera tracking is such a big
and broad topic, and there's a lot of techniques

in 3D camera tracking, I'm going to take
this tutorial and split it up into two parts.

So, in this first part we're going to be
dealing with footage that has tracking markers,

which makes it much easier to track.

And remember, our goal is to make
a camera object inside Blender

that matches the motion of the shot.

So, we want the camera to move in the same
way, and also have the same perspective.

And by doing a bunch of point tracks on our
footage, Blender can solve for our camera.

So, obviously in this shot,
we're going to choose

to track the markers on our sheet of paper.

In a harder shot, we're going to need to get
more creative with what we pick to track.

But in this case, we just have it really easy.

So, the first step is converting
this video into an image sequence.

And this step isn't entirely necessary, but
an image sequence doesn't have a frame rate

which ultimately gives us a lot more control.

So, when you can, it's always
best to work with image sequences.

But again, you don't have
to if you don't want to.

So, all we have to do is go into the movie clip
editor and then just import in our footage.

And now that it's imported, you see on the
right panel which you can click N to toggle,

you see a whole bunch of information
like the resolution, the frame rate,

and also the duration which is in frames.

And we want the settings in our
Blender project to reflect these.

For the video duration there's
two ways to do this.

We can either change it manually here to 139,

or we can click this button called set scene
frames, which does it for us automatically.

And now these both have the same frame duration.

Next, we need to match the frame rates.

And to do this, go to the output tab and
then look for the frame rate options.

Now in our case, we have a very
weird frame rate, which is 29.87,

and you see that's not an option here.

So, we're going to have to use custom.

And we can't just type in 29.87 because in
this case, it's just going to round up to 30.

Instead what we're going to do
is scale both of these by 100.

So, we have 2,987 over 100.

And you can think of this as division,
which gives us our correct custom framerate.

And also, we have the correct
resolution of 1080p by default.

Now we just go into compositing
and start setting up some nodes.

So, make sure that you have
nodes and auto render enabled,

so you can see what you're doing.

And we're going to take this render
layers node and delete it with X,

because we don't need it for this.

To bring in our footage, just click shift A to
add, and then go to input and then movie clip.

And this gives us our node.

And make sure that this is set to the
correct footage that we imported before.

And now, all we have to do is hook these two up.

And if we want to see what we're doing, we're
just going to also add in a viewer node,

and then just hook it up as well.

Now that we're done with the setup,
we're just going to go to the output tab,

and then set this export to jpeg.

And this isn't really the
highest quality format,

but it doesn't really matter for this tutorial.

And then we're just going to create
a new folder and set an output path.

And while we're still in EV, we can just
hit control F12 to render out our animation.

Now that we've converted into an image
sequence, we can start up a new Blender project.

Again, we're going to go
to the movie clip editor.

But this time we're going to
import in our image sequence.

And to import in an image sequence, just select
all your frames with A and then import them.

And remember that an image
sequence doesn't have a frame rate,

so we can just go to the output tab
and freely choose our frame rate.

So, I'm going to choose 30 frames per second.

And for the project duration, we're just
going to hit set scene frames again.

And that's going to automatically
take care of it for us.

Now notice that when we scrub through the
timeline, the footage is a bit choppy.

And to fix this, what you
have to do is click pre-fetch,

which loads this whole sequence into memory.

And now this is playing much more smoothly.

Remember that we get our camera solved
by just having a bunch of point tracks.

For the Blender camera solver, we
need to make sure that there's always

at least 8 trackers visible
during any frame in the shot.

So, the name of the game is getting as
many high quality trackers as possible.

We want trackers that last for a long
time and have a very low average error.

And to help us with this, we want to make sure

that we're using the best
settings for our trackers.

The first thing we need to
pick is a motion type.

And there's a lot of options.

Like location, for example, which tracks
the location of a pattern between frames.

But in this case, it probably makes
the most sense to use perspective.

Because our markers are kind of getting
perspective warped as the camera is moving.

We can also enable normalize, which is going to
help Blender ignore some subtle lighting changes

that might have thrown off our track otherwise.

This does take some more time to
compute, but generally it is worth it.

Lastly, in the tracking settings extra, we're
going to pick a correlation value of 0.9.

This number is basically just a
threshold that the tracker needs

to meet during every frame, so it doesn't stop.

So, when this value is zero, it's just going
to ignore any error and just keep tracking.

And instead, if this number is 1, it needs
to be 100% confident during every frame

that it's correct, or otherwise
it's immediately going to terminate.

So, 0.9 is a very high value which
demands a very high accuracy,

but it does give us a little bit of wiggle
room, so it's not stopping constantly.

And now we can begin adding in
trackers with control click.

And in the tracking panel, we get
a zoomed in window of our tracker,

which lets us make fine adjustments
and center it.

We can also scale up this tracker with S, so
our whole feature is inside the bounding box.

Now, there's also a second hidden
bounding box, which we can show with alt S,

and this is called the search box.

The search box just defines where we're going
to search for our pattern between each frame.

So, a larger search box means it's
going to search a larger area,

but that's also more computationally expensive.

For this shot, the camera isn't
moving around very quickly,

so we can keep this search box rather small.

Once we're happy with everything,
we can begin tracking manually

by hitting alt and then right arrow.

This is how we track frame by frame, and
you see that it's sticking on very well.

To center this tracker to the view, you can just
hit L and that's going to lock it to the center.

And after doing a few more frames manually, we
can have it go through the rest of the footage

by just hitting control T, and
it tracks all the way to the end.

And of course, we can also check how
good this tracker is throughout the shot.

Once we're happy with this tracker, we can just
lock it by hitting control L. And this means

that we're done with the tracker and
we can't accidentally move it and mess

with the key frames or anything like that.

So, now we go back to the first
frame with shift, left arrow,

and we have to do the rest of the trackers.

Now, this isn't going to take as long
because we can do all of them at once.

So, I'm just going to pick a couple trackers.

We don't need to do every single feature.

Let's do something like 10.

When you set up all your trackers, you
can just select all of them with A.

And then we're just going to hit control
T, and they're going to track all together.

Then we can just hit control
L to lock all of them.

And now it's finally time
to work on our camera solve.

So, just go over to the solve panel, and
you see a whole bunch of options here,

so we're just going to take them one at a time.

For the tripod option, we're
going to keep this disabled

because this shot has a free moving camera.

It's not locked to a tripod,
and it's not a nodal pan.

Now, for key frame and key frame A and B
options, what this basically lets us do is

when we do our camera solve, we pick some
range that we deem to be very accurate,

and then it's just going to extrapolate that.

So, key frame A and B is
basically defining this range.

Now in this case, we don't really know what we
should pick for key frame A and key frame B,

so we're just going to let Blender decide
by enabling this key frame check box.

And now we can click solve camera motion and
Blender is going to solve for our camera.

And we get some number, which
is our solve error.

Which generally, we want to
make as small as possible.

A solve error under 1 is somewhat
okay, but if we can get it

under 0.5, then that's pretty good.

Basically, we want to get this
value as close to zero as possible.

And as a result, our solve is
going to become better and better.

The first thing we can do to lower our solve
error, is go over to the refine options.

And you see that we have
a whole bunch of choices.

We are going to choose the focal length,
and K1 and K2 option, so that Blender tries

to calculate what the focal
length was in the shot.

That's essentially how zoomed in we were.

And also, the lens distortion,
which are these K1 and K2 numbers.

And then we can just click
solve camera motion again.

Another trick for lowering our solve error, is
to go into clip display and then enable info.

And this is going to give us
information about the error

of each individual tracker that we had.

Obviously, what we want to
do is find the tracker

with the highest error, and then just delete it.

We then click solve camera motion again
and we should have a lower solve error.

So, let's say that we're happy with this now,
and we want to incorporate it into our 3D scene.

All we have to do is click setup tracking scene.

And already this is looking very wrong.

But that's because we just didn't
set up our world orientation.

The first thing we do is select three trackers
on this table, and set this to be our floor.

We can also pick one of these trackers in
particular, and set this to be the world origin.

So, everything is going to
be centered from this point.

We can then pick any two trackers and
choose a distance, and then pick set scale

so that the distance between these two points
is defined to be however many units we put in.

Lastly, to set up the X and Y
axis, with respect to the origin,

all we have to do is select a tracker to the
left of the origin, and then click set X axis,

and then we pick a new tracker below
the origin and click set Y axis.

And now you see our camera solve
is behaving how we expected.

We can see our results by going
into the rendered view in cycles.

The plane becomes transparent, because
it's set to be a shadow catcher.

However, there are no shadows.

To fix this, all we have
to do is select our plane

and bring it into the foreground collection.

We can also go to the background
layer that was set

up when we did our camera
solve, and we can delete it.

And now you see that everything
is working out as we expect.

We're now going to go into compositing
and set up our nodes before we render.

And you see, we already have this whole network.

That again, is set up by default when
you selected set up tracking scene.

I'm just going to delete this with X, and
show you how to make a custom node setup.

We're going to click shift data add.

And then we need our movie
clip for the background,

and then our render layers for the foreground.

And we can then overlay these two,
just using an alpha over node.

And then finally, we're just
going to add a composite node,

and also a viewer node so
we can see what we're doing.

And so far, you see that we have the
background but not our 3D objects.

And that's because we didn't
render anything yet.

So, I'm just going to go to the first
frame, and then hit F12 to render.

And then you see that our
render layers node updates,

and then everything works out as it should.

Finally, in the output tab I'm going to
set this to render as a jpeg sequence,

which we can save in a new folder.

And now, all we have to do is hit
control F12 to render out the animation.

And that is the basics of 3D
camera tracking in Blender 2.8.

So, in the next video I'm going to be
going over much more difficult footage.

But also, we're going to get much more in
depth about how the 3D camera tracker works.

So, thank you guys for watching and
I'll see you guys in the next one.