Create Cinematic AI Video using Kling | AI Filmmaking Tutorial
By Curious Refuge
Summary
## Key takeaways - **Cling's Image Generator Limitations**: While Cling does have an image generator, it's not as intelligent as other tools like Nano Banana. It's recommended to create your initial images in external tools and then bring them into Cling for animation. [02:24] - **Upscale Images with Topaz Gigapixel**: For upresing generated images, Topaz Gigapixel is considered the best tool available. Use the "highfidelity" setting, turn advanced settings like denoising to zero, and set the upscale factor to x4 for realistic details. [05:49] - **Iteration is Key for AI Animation**: Cling does not get the animation right every time, so expect to generate multiple versions for most shots in your films. It typically takes around five generations on average to achieve the desired outcome. [08:30] - **Upscale Videos with Topaz Video AI**: After animating, it's essential to upres your videos from 1080p to 4K or beyond for professional projects. Topaz Video AI is recommended, using the default "Proteus" model and setting "recover detail" to zero. [09:47] - **Add Film Grain for Realism**: To make AI video footage feel more realistic, apply a simple film grain to your timeline in post-production. Set the blend mode to overlay and keep the opacity below 50% (e.g., 42%), as less is more with grain. [11:00]
Topics Covered
- Unlock Cinematic AI: A Multi-Tool Workflow Outperforms All-in-One.
- Master Consistent AI Characters and Styles with Image References.
- Craft Cohesive AI Scenes: Advanced Prompting for Reverse Shots.
- AI Doesn't Get It Right: Expect Iteration in Video Generation.
- Add Film Grain to AI Videos for Instant Realism.
Full Transcript
In this video, we're going to take a
look at how to create cinematic videos
using Cling. To create this video, I
collaborated with Caven Cardoza, also
known as Caven the Kid. He's one of the
most talented AI filmmakers in the
world, and he's also the instructor of
our advanced AI filmmaking course. So,
over the next few minutes, we're going
to take a look at how to get maximum
cinematic qualities from your AI videos
using Clink. Let's get started. So, as
an illustration for this tutorial, I put
together a quick little scene. Let's
take a look.
man. We've all been there. Okay, so
that's the example that we're working
towards. Let's hop into the tutorial.
So, of course, it all starts with our
character. And for our example, I'm
going to use this image of this woman.
Of course, you can use any character you
want. And the cool thing is with the
techniques that I'm going to show you
today, you'll get consistent characters,
clothing, and environments. So, we have
our character. Now, we want to focus on
the cinematic color grading and style
from the images that we want to create.
Now, of course, you could go in and
curate a board and work inside of tools
like Midjourney to fine-tune the images
to be exactly what you're looking for,
but for our example, I'm going to make
it super easy. Essentially, I came
across this image, and I really like the
color grading style. So, next up, it's
time to create the images. Now, there is
an image generator directly inside of
Cling, but it's not quite as intelligent
as the image generator that I'm going to
show you inside of this video. So, I
like making the images in other tools
and then bringing them into Cling to
animate them and ultimately create your
videos. So, the tool that I'm going to
be using is called Nano Banana. But if
you want to learn more about creating
images or getting consistent styles from
your images, be sure to click the link
below this video to check out our video
on consistent characters. So, I'm going
to use the online platform Free Pick to
use Nano Banana and we'll go ahead and
go to generate image. So, first things
first, make sure under model you're on
Google and then Nano Banana. Let's go
ahead and bring in both of those images.
So again, we have our character. And
then we have our color grading style.
We'll drag and drop them into the image
reference section. And for our prompt,
we'll say a cinematic closeup of at
image one. That's the image that has our
character wearing a linen dress sitting
outside an Italian cafe in Positano. Use
the same color grading and filmic style
as image number two. And we'll go down
here and make sure that we're going to
create four images in 16 by9. And go
ahead and click generate. So after just
a few seconds, we have quite a few
images here. So let's take a look. So
these are already looking pretty darn
good. You can see each one is uniquely
different. I don't like her looking at
the camera quite as much, but I do think
there's a lot of things that I do like
about some of these images. We'll work
our way through here. And I really like
this image. I like the details on the
cup. The composition's interesting. And
I like that she's looking forward. So,
we'll go ahead and go to the export
button. Make sure we're set to PNG and
go ahead and select download. So, now
it's time to create the reverse shot.
So, what is she looking at? So, the cool
thing is inside of Freepic, you can of
course upload the image like we did
earlier, but if you created your image
directly inside of Freepic, you can just
drag and drop it directly in the image
references section, which is a pretty
fun hack. Okay, so now let's type in our
prompt. This one's a bit longer. So,
what I said is show the reverse shot
from the woman's perspective of a guy
drinking coffee. Use the same color
grading as the image one, which is the
only image that we uploaded. The guy
should have the reverse angle of
Positano with the homes on the cliff on
the left side of the frame as this is of
course a reverse shot. So, let's go
ahead and click generate. So this is our
first image here and yeah, it really
knocked it out of the park. That's
exactly what we're looking for. Let's
take a look at a few other images here.
So we have this one which basically
replaced the woman. So not really what
we're looking for. We have this shot
here. Not exactly what we're looking
for. And then we have this shot here.
Again, it replaced her. So not exactly
what we're looking for. So you can see
it's sometimes about iteration, but we
actually got lucky with the very first
generation being exactly what we're
looking for. So, let's go ahead and
download that to our computer. Okay. So,
I'm going to spare you the back and
forth. Essentially, we used the exact
same process to create each one of the
shots for our scene. So, now that we
have our images, it's now time to upres
them. So, there are a few different
tools out there that can help you with
the upresing process. We've tested all
of them, and I really feel like Topaz
Gigapixel is still the best tool on the
market. To use the tool, all you have to
do is take your image, drag and drop it
into Gigapixel, and from here, there are
a few different options. You can see we
have standard, highfidelity, and low
res. I'm going to go ahead and select
highfidelity. And from here, there's a
bunch of advanced settings. I like
turning them all the way down because I
don't want to have any extra denoising
or artifacting happening to our images.
I want them to look as natural as
possible. And for the upscale factor,
I'm going to set it to x times 4, you
can of course multiply it to whatever
number you want. But of course, once you
start getting too high, the images are
just simply too large. So you can see
here with the before and after, if you
see the before, super pixelated, and
then after, it has a lot more realistic
details. So there's just a lot more
individual details with the hair
strands. The eye looks much more
realistic, and you can even see some
texturing in the hat. So, it's doing a
really, really good job. So, when you're
ready to save your image on your
computer, go ahead and click export
image. So, you'll do the exact same
thing for all of the images inside of
your sequence. Okay, now comes the fun
part. It's time to animate our images.
Of course, you can go over to the Cling
platform and animate your images
directly inside of Cling, which is
awesome. I'm actually going to continue
to use freepic here simply because we
already have credits to the platform and
there's an API that connects clink
directly inside of freepic. So it's a
great back and forth process directly
here on the platform. So in order to
create a video inside of freepic, we're
going to go to the generate video button
here. Let's go ahead and bring in one of
our images. I really like this image
here of her making this like grossed out
face. So, we'll go ahead and drag and
drop that into the start image section.
And for our prompt, we'll say, "A woman
looks at the camera confused and grossed
out, looking around, looking back at the
camera." And we'll also say handheld
camera shake. I like defining what
camera movements I want to see if I have
specific things that I'm looking for.
Now, for the duration, we're going to
set this to 10 seconds. basically just
giving us more time and it allows us to
pick and choose the exact few seconds
that work for the shot that we're
working with. And go ahead and click
generate. And after a few minutes, we
have this shot here, which honestly
looks pretty good. You can kind of feel
the sense that she's really grossed out
from the shot. So, I think that Clling
did a really good job. Now, does CL get
it right every single time? Absolutely
not. you definitely will have to
generate multiple times for most of the
shots inside of your films. Let me just
show you a quick example of that. So,
basically, I wanted this guy to rub his
hand through his hair and then lick his
hand and be grossed out. And uh we tried
getting that a few different times. And
so, we have this shot here. He rubs his
hand and then he licks it, but he kind
of just like bites his hand. So, that's
definitely not what we're looking for.
Uh, we have this shot here. Rubs his
hair. He also looks like an Oompa-
Loompa. I think the color grading
changed. And then I don't know, the
dynamics on that tongue is just like
really off. And then we have this shot
here. He touches his head and then he
like licks his fingers and he's like
grossed out. I don't know. It's just
very offputting and weird. So, a lot of
the shots are not exactly what we're
looking for. It's a back and forth
process. You'll create a video, adjust
your prompt. It's going to take you
probably five generations on average to
get exactly what you're looking for. So,
now that you have your shots, it's time
to upres them. So, by default, they're
going to be in 1080p, but you need to
upres them into 4K or beyond if you're
working on a more professional project.
And the tool that I like to use to upres
my videos is Topaz Video AI. So, let's
go ahead and take that shot here of the
woman. She just really grossed out here.
and we'll drag and drop that into Topaz
Video AI. And under the output
resolution here, we'll set it to 4K. And
under the model, we'll set it to
Proteus, which is the default model.
Honestly, it does a pretty good job. And
we'll turn down recover detail to zero.
We're not going to adjust any of the
other settings, and you can go ahead and
go to the quick export button in the
bottom right to export it to your
machine. So, after a few minutes, you'll
have all of the shots on your computer,
and now it's time to edit them all
together inside a video editing
application. I like using Adobe Premiere
Pro, but of course, Da Vinci Resolve is
another fantastic option if you're
looking for a professional tool to edit
all of the video clips. So, I have an
incredibly simple timeline here. You can
see we have our shots. We also have a
music track that I downloaded online.
And then we also have some sound
effects. Now, one tip that I like to use
whenever I am working with AI video
footage is to apply a very simple film
grain to my timeline. This helps
everything to just feel a bit more
realistic. You don't want to go
overboard, but just a little bit goes a
long way. So, I've went ahead and pasted
our film grain into the timeline. You
can see we have the film grain over our
video footage. So, I have the grain
selected in track two here, but you'll
see that it's gray. Basically, when we
play it back, there's the grain, but
there's, you know, nothing that you can
see underneath. So, what we're going to
do is change the blend mode to overlay.
And then I'm going to turn down the
opacity below 50%. So, for this one, I'm
going to do 42%. I recommend playing
around with it to dial it in to the
exact grain amount that you're looking
for. I really think less is more when
you're working with grain. A lot of
times people will just set it to 100%.
And it really comes across as a little
too grungy in my opinion. So now that
you have everything in your timeline,
you can go in and finesse the shots.
This is your shot to, of course, adjust
the sound design. You can add in
different music. you have the ability to
match the color grading, which can be
very helpful for helping to sell the
overall realism of your scene. And once
you're done, you can export your
project. So, there you go. That's how to
create realistic and cinematic AI video
footage inside of Cling. Of course, if
you want to learn the latest AI film
making techniques, be sure to check out
our AI filmm and advanced AI filmmaking
courses over at Curious Refuge. We train
artists at every major studio and we
would love to have you inside our
community. And of course, you can always
like and subscribe here on YouTube to
get the latest AI film news and
tutorials directly here on the platform.
Thank you so much for watching and if
you have any questions at all, be sure
to ask in the comments. Best of luck on
all of your creative projects.
Loading video analysis...