00:00:00
you now have access to one of the
00:00:02
world's best AI video models there is
00:00:05
huge news for 3D artist and the CEO of
00:00:09
Netflix has some bold predictions about
00:00:12
the future of AI in Hollywood this is
00:00:16
your AI film news of the week I want to
00:00:18
let you know that you can stick around
00:00:20
to the end of this video to get a bonus
00:00:23
tip on how to win our game of the week
00:00:26
now let's get to the news we're going to
00:00:28
kick things off with some new from one
00:00:30
of the most powerful people in Hollywood
00:00:33
and I am of course talking about the
00:00:35
co-ceo of Netflix Ted Sandos Ted was
00:00:40
recently interviewed about his thoughts
00:00:42
on the future of AI in Hollywood and he
00:00:45
has some interesting predictions first
00:00:48
Ted said that AI isn't going to reduce
00:00:51
cost but really be helpful for improving
00:00:54
the overall quality of the films that
00:00:57
are created so rather than trying to
00:00:59
make a film for 50% less he really feels
00:01:02
like it's a better use of AI to try to
00:01:05
make a film 10% better with more
00:01:08
iterations coming from artificial
00:01:10
intelligence and this is a view we
00:01:12
definitely share here at curious Refuge
00:01:14
at the end of the interview Ted says AI
00:01:18
is not going to take your job the person
00:01:20
who uses AI might well take your job and
00:01:24
that is a really important Point
00:01:26
artificial intelligence is not able to
00:01:29
craft interesting stories by itself but
00:01:32
people who use artificial intelligence
00:01:34
are able to create projects that are
00:01:36
even better than if they wouldn't have
00:01:38
used AI in the first place so I highly
00:01:41
recommend checking out AI tools that are
00:01:43
available to you as a creative artist
00:01:45
there are a lot of AI tools out there
00:01:48
that are helpful for filmmakers but if
00:01:50
you want we actually put together a list
00:01:52
of some of our favorite tools over on
00:01:54
the Curious Refuge website you'll find a
00:01:57
link below this video we also have some
00:01:59
really exciting news the team at cing
00:02:02
has released their video model publicly
00:02:05
to the entire world so this is great
00:02:08
because previously you had to have a
00:02:10
Chinese phone number in order to have
00:02:12
access to cling and most people do not
00:02:15
have a Chinese phone number there have
00:02:16
been some really awesome examples of
00:02:19
cling online for example we have this
00:02:21
shot here of George Washington eating a
00:02:24
cheeseburger which is uh just Majestic
00:02:26
we have the shot here of a Viking Chief
00:02:29
drinking from a cup looks pretty darn
00:02:31
good the people in the background look
00:02:33
very realistic and the motion looks
00:02:35
really good we have this shot of this
00:02:37
alien robot character that looks pretty
00:02:40
darn great the bokeh in the background
00:02:42
and The Parallax looks very realistic
00:02:45
and we have the shot of kind of this
00:02:46
World War II scene I think the airplane
00:02:49
in the background is moving pretty slow
00:02:51
but other than that it looks pretty good
00:02:52
you can see there's a lot of like
00:02:54
particles uh in the background it kind
00:02:56
of look like paper leaflets there's also
00:02:58
kind of Fire coming in front of that
00:03:01
building there and it looks pretty darn
00:03:02
good and lastly we have the modern-day
00:03:05
version of piano cat wearing a fedora
00:03:08
which is just delightful now you may be
00:03:10
wondering how cing compares to some of
00:03:12
the other tools that are out there like
00:03:15
Luma dream machine and we went ahead and
00:03:18
did a quick comparison so first I want
00:03:21
to show you a video that was created
00:03:23
using the lumo
00:03:29
[Music]
00:03:41
okay so you can see that the results
00:03:42
from Luma are pretty good but let's take
00:03:44
a look at cling
00:03:51
[Music]
00:04:02
okay so as you can see they're pretty
00:04:05
comparable I do think that there's one
00:04:07
specific shot that I really liked from
00:04:09
cing the shot where the woman is kind of
00:04:11
looking off the camera and then she
00:04:13
looks down and looks back up there's a
00:04:15
lot of dynamic movement in like the fur
00:04:17
in the jacket that she's wearing the
00:04:20
background has some fire elements that
00:04:22
are blinking in a very realistic way
00:04:24
with the bokeh being natural to what
00:04:26
camera lens you would expect from that
00:04:29
type of of shot and I just think it's a
00:04:31
really really Dynamic and interesting
00:04:33
shot now is cing perfect no I think when
00:04:36
you look at a side by-side comparison
00:04:39
the Fidelity that you get from cing is
00:04:42
worse than the Fidelity that you get
00:04:44
from Luma Labs but that's not to say
00:04:46
that it isn't an amazing tool I highly
00:04:49
recommend checking it out let's take a
00:04:51
look at how you can get signed up with
00:04:53
the tool right away it was a little
00:04:55
confusing but I'll show you each step of
00:04:57
the way first things first go to the
00:04:59
cing AI website that's cing ai.com for
00:05:03
some reason there are some websites that
00:05:05
do not link to cing AI that claim to be
00:05:07
kling's website so it's a little strange
00:05:09
you'll find a direct link to this
00:05:11
website below this video and go ahead
00:05:13
and click the sign in for free credits
00:05:16
button and if you don't already have an
00:05:19
account which you probably do not you
00:05:21
can click the sign up for free button
00:05:23
and basically it just takes an email
00:05:25
address and you get to create a password
00:05:27
once you're signed in go ahead and go to
00:05:29
the a i video section and like always
00:05:32
we're going to use image to video not
00:05:34
text to video text to video is really
00:05:36
only helpful for establishing shots and
00:05:39
if you want to get likes on Twitter but
00:05:41
if you want to tell a compelling story
00:05:42
you really need image to video so you
00:05:44
can have consistency in your Creations
00:05:47
so we'll go to image to video and now
00:05:49
all you have to do is drag and drop an
00:05:51
image I have this image of a Viking
00:05:54
woman here and we'll just drag and drop
00:05:56
it into cink now for your prompt you can
00:05:59
prompt whatever you want it's best to be
00:06:01
specific and talk about the camera
00:06:03
movement that you want to see so we'll
00:06:05
say a viking woman walks around her home
00:06:08
handheld camera movement now a lot of
00:06:10
the features inside of cling are not
00:06:12
available to the public for example they
00:06:14
have an ad an inframe button that is
00:06:16
very similar to the inframe feature
00:06:18
inside of bluma labs that's not
00:06:19
currently available they also have
00:06:21
something called a highquality setting
00:06:24
that isn't available just yet either and
00:06:26
so once we get that high quality setting
00:06:28
it's going to be very helpful because
00:06:29
right know the quality is not the best
00:06:31
but it does look pretty good and you
00:06:33
also have the ability to adjust the
00:06:35
creativity or the relevance so relevance
00:06:37
will make the video more in line with
00:06:40
the uploaded image creativity will just
00:06:43
take it further and it can get more
00:06:45
creative with the movements and just
00:06:47
overall uh aspects of the frame that
00:06:49
might change Now the default actually is
00:06:51
pretty darn good I find that the default
00:06:53
is a little better than the default and
00:06:55
some of the other AI video tools that
00:06:57
are out there so we'll keep it there and
00:06:59
and we will keep it at 5 Seconds now you
00:07:02
also have the ability to type in a
00:07:04
negative prompt to say what you don't
00:07:06
want to see in your scene which can be
00:07:08
helpful but we are not going to select
00:07:10
that now because I want to see what we
00:07:12
get by default so let's go ahead and
00:07:15
click generate I should note that when
00:07:17
you generate a video using clling it
00:07:20
uses 10 credits and they're giving
00:07:22
everyone 66 daily credits to create your
00:07:26
video so basically you get 30 seconds of
00:07:28
video for free every single day I should
00:07:31
also note that the videos you get from
00:07:32
cing are not HD quality they're only
00:07:35
1344 and most critically for filmmakers
00:07:39
the frame rate is 30 frames per second
00:07:41
not 24 frames per second so if you're
00:07:43
working on a film project that frame
00:07:46
rate issue might be a serious issue that
00:07:48
you need to work around cling does not
00:07:50
have a paid version available to
00:07:53
International customers but luckily for
00:07:55
us Tim at theoretically media was able
00:07:58
to get access so what the pricing looks
00:08:00
like at least in China right now and we
00:08:03
were curious about how that compares to
00:08:05
other AI video tools that are out there
00:08:07
so right now on King's website they're
00:08:09
basically using a credit system that
00:08:12
whenever you subscribe for a year you
00:08:15
basically get about 25 minutes of AI
00:08:18
generated footage for about $18 a month
00:08:23
which is pretty good actually because if
00:08:25
you compare that to Luma labs for about
00:08:28
$23.9 9 a month you get 10 minutes of AI
00:08:32
video and on runway for $28 a month you
00:08:35
get 3 minutes and 45 seconds of video so
00:08:39
clling is by far the most efficient AI
00:08:42
video generator that is out there
00:08:45
however I should note that at curious
00:08:46
Refuge our favorite AI Video
00:08:48
subscription is the runway unlimited
00:08:51
plan because it's literally unlimited
00:08:53
okay let's take a look at our result so
00:08:55
here's the clip from cing you can see we
00:08:58
have our Viking woman
00:09:00
and she's walking around and there's a
00:09:02
weight in the steps her hair is blowing
00:09:05
in the breeze and you can see that the
00:09:08
fur like it's really interesting to see
00:09:10
the fur lifting as if her shoulders are
00:09:12
kind of rising and pushing up the
00:09:15
costume that she's wearing and even when
00:09:16
we zoom in and analyze some of the
00:09:18
deeper parts of this footage like look
00:09:20
at the background you see the way that
00:09:22
the mountain parallaxes with the cabin
00:09:25
and her outfit how they're all in
00:09:27
different planes that's a very advanced
00:09:29
movement that artificial intelligence
00:09:31
was able to recognize and create in just
00:09:34
a matter of minutes and that brings us
00:09:37
to our game of the week for this week's
00:09:39
game I'm going to show you six different
00:09:42
video clips I want you to tell me which
00:09:44
one was created in Luma and which one
00:09:47
was created in cing all right let's get
00:09:49
started here's clip number
00:09:56
one here's clip number two
00:10:02
loving the uh Dynamic Fire movement in
00:10:04
the background there here's clip number
00:10:09
three okay got some pretty good wave
00:10:11
simulations there this is clip number
00:10:14
four not loving the way the water is
00:10:17
rising there I don't think that looks
00:10:18
super realistic and here's clip number
00:10:21
five we have this guy driving around
00:10:23
guessing like an English
00:10:25
Village and clip number six
00:10:30
all right pretty good I like the camera
00:10:31
Shake on that one if you think you know
00:10:33
which clip was created in which tool let
00:10:36
us know in the comments we will randomly
00:10:38
select one of the winners to receive a
00:10:41
free AI film making sweatshirt from
00:10:43
Curious Refuge I highly recommend
00:10:45
checking out cing and of course if you
00:10:48
create something awesome with the tool
00:10:50
be sure to tag Us online so that we can
00:10:52
share it in our AI film Gallery the next
00:10:54
news that's pretty cool is sunno has
00:10:57
released the ability to create stems
00:11:00
with their songs so a stem basically
00:11:02
allows you to isolate certain
00:11:04
instruments or vocal layers in a music
00:11:07
track so with suo right now you can
00:11:08
really just isolate the vocals and the
00:11:11
music but I imagine we will get more
00:11:13
stems very soon let's take a look at how
00:11:15
to use this feature so go over to the
00:11:18
sunno website and log in with your
00:11:21
account and go to the create button okay
00:11:24
so inside of sunno go to the create Tab
00:11:26
and make sure you're in version 3.5
00:11:29
which is the newest model now you can
00:11:32
type in a prompt we'll say a pop song
00:11:34
about an AI film news show from Curious
00:11:37
refuge and go ahead and click create
00:11:39
Okay let's listen to our
00:11:46
[Music]
00:11:57
song okay that's it's a great song So to
00:12:01
create the stems all you have to do is
00:12:02
go to the three dots and ensure that you
00:12:05
have the time code in the little
00:12:07
thumbnail there if you don't have the
00:12:08
time code that means that sunno has not
00:12:10
created the entire song yet so it can't
00:12:12
pull the stems so let's go ahead and
00:12:15
click on the three dots and we will
00:12:17
click get stems so now sunno will go in
00:12:20
and separate the vocal track and the
00:12:22
audio track okay in about 60 seconds it
00:12:25
created the stems let's listen to each
00:12:27
one as
00:12:33
[Music]
00:12:36
okay that distorted AI voice is still
00:12:38
there which is really funny and let's
00:12:40
listen to the instrumental
00:12:48
track and you can see that it is an AI
00:12:51
removal process so clearly there still
00:12:53
is some of the music in the vocals and
00:12:55
some of the vocals in the music but it
00:12:58
is very helpful to have an automatic
00:13:00
tool directly in pseudo so you don't
00:13:02
have to go to a different AI tool to
00:13:04
separate the tracks I'm also very
00:13:06
excited to announce a new course at
00:13:09
curious Refuge we've partnered up with
00:13:11
Fred grinstein to create the world's
00:13:14
first AI documentary course this course
00:13:17
is specifically for anyone who wants to
00:13:19
use AI to create compelling
00:13:21
documentaries or unscripted content Fred
00:13:25
has worked on documentary and unscripted
00:13:27
projects for incredible Studios
00:13:29
including HBO A&E and vice now there are
00:13:34
a lot of people that are already on the
00:13:36
wait list but I should note that there
00:13:38
is special beta pricing for the very
00:13:41
first session of this course which is
00:13:43
awesome it will go on sale on July 31st
00:13:46
at 11:00 a.m. Pacific time and speaking
00:13:49
of July 31st at 11:00 a.m. Pacific our
00:13:53
other courses at curious Refuge will
00:13:55
also go on sale this includes AI film
00:13:58
making and AI advertising we've updated
00:14:01
the course with the latest tools and
00:14:03
techniques that you need to know be sure
00:14:05
to check out our course pages over on
00:14:08
Curious refuge and join the wait list to
00:14:10
be the first to know when the course is
00:14:11
dropped the team at Adobe also came out
00:14:14
with some new features inside of
00:14:16
Photoshop and Adobe Illustrator let me
00:14:19
show you how to use them the first tool
00:14:21
that we're going to check out is inside
00:14:23
Adobe Photoshop I should note that you
00:14:25
need to update to the most recent
00:14:27
version of Adobe Photoshop to use this
00:14:29
tool at the time of recording that's
00:14:31
Photoshop
00:14:33
2.11 so in order to use this new tool
00:14:36
called The Selection brush tool all you
00:14:39
have to do is go to the little icon on
00:14:41
the left usually by default it's the
00:14:43
lasso tool so if you just hold down your
00:14:46
mouse you can skip over to the selection
00:14:48
brush tool now the thing that makes this
00:14:50
tool unique is the fact that it's
00:14:53
designed to select a subject but with
00:14:56
varying levels of opacity so usually
00:14:59
when you select the subject you're
00:15:00
basically getting each and every pixel
00:15:03
selected and you can't select the amount
00:15:05
of opacity in that selection so this is
00:15:08
really helpful especially when you begin
00:15:10
to pair it with generative fill because
00:15:12
you can bring in certain number of
00:15:13
pixels from the previous image into your
00:15:17
selection so it basically Smooths out
00:15:20
the edges so in order to select your
00:15:22
subject all you have to do is go here
00:15:24
and change the size of your brush you
00:15:27
can also change the hardness my favorite
00:15:29
way to do this is to hold down control
00:15:31
and option on a Mac and you can click to
00:15:35
the right and left to drag for the size
00:15:38
so let's say we wanted to select this
00:15:39
tree here basically we can just
00:15:42
drag and just kind of lightly
00:15:46
select the
00:15:47
frame just like this and we can now go
00:15:51
to gener fill and type in whatever we
00:15:53
want so let's say we want just a
00:15:55
completely different plant here I'm
00:15:56
going to say out of focus
00:15:59
and go ahead and click generate okay so
00:16:02
let's take a look at what we got here so
00:16:03
the first one here we have this
00:16:05
generation I don't like what it's doing
00:16:07
to our leg down here and I don't know
00:16:09
why it put a gigantic a quarterback
00:16:10
there so that's not great we have this
00:16:12
second generation which I like I like
00:16:14
the plants don't love the wrinkles in
00:16:17
the leg but it's not bad and the last
00:16:19
one here we have her standing here with
00:16:21
these very basic pots I think that one
00:16:23
looks pretty good there's also a new
00:16:25
feature inside Adobe Illustrator that
00:16:28
allows you to generate vectors inside of
00:16:31
shapes so this is really helpful for
00:16:33
creating patterns and branding again you
00:16:35
need to have the most upto-date version
00:16:37
of Illustrator to get access to this
00:16:39
tool so we have this shape selected here
00:16:41
and we can click the generate shape fill
00:16:44
button here and we can type in whatever
00:16:46
we want for our prompt we'll say flowers
00:16:49
and ice cream and go ahead and click
00:16:51
generate okay let's take a look at our
00:16:53
results here so we have number one
00:16:55
number two and number three so I think
00:16:59
this one did pretty good I like number
00:17:01
three a lot because look here it created
00:17:03
an orange slice for the wing which uh is
00:17:05
pretty hilarious now the cool thing is
00:17:07
because we're in Adobe Illustrator all
00:17:09
of these objects are shape layers and
00:17:11
they can be infinitely scaled and
00:17:13
manipulated which makes them different
00:17:15
from using tools like Adobe Firefly
00:17:18
inside of Photoshop in event news we'd
00:17:20
like to extend a huge congratulation to
00:17:23
the team at the beon International Film
00:17:25
Festival they were kind enough to invite
00:17:28
Dave on behalf of curious Refuge to
00:17:31
share his thoughts and insights on the
00:17:34
future of artificial intelligence in
00:17:36
film making and in advertising the event
00:17:39
was the most popular version of the beon
00:17:43
International Film Festival in history
00:17:45
and it featured everything from AI
00:17:47
competitions to AI workshops we hope
00:17:50
everyone who was able to attend had an
00:17:52
amazing time we would love to be there
00:17:55
next year now on the other side of the
00:17:57
world we'd like to give you a quick
00:17:59
update on the reply AI Film Festival
00:18:01
that's happening at the same time as the
00:18:03
Venice International Film Festival in
00:18:06
early September so as you know we
00:18:08
partnered up to get the word out that
00:18:10
the top 10 finalists will be flown to
00:18:13
Venice to Premiere their film well we
00:18:15
just heard that actually they extended
00:18:17
it to the top 12 finalists so 12 people
00:18:20
will be notified and flown out to that
00:18:23
event we'll let you know more about the
00:18:25
finalist and the projects that made it
00:18:27
into that competition in the coming
00:18:29
weeks in other AI news the team at hyper
00:18:32
came out with the ability to create 8C
00:18:34
long videos which is an added
00:18:36
Improvement they also allow you to
00:18:38
extend videos now and you can upscale
00:18:41
videos directly in their platform hyper
00:18:44
is a really interesting tool so if you
00:18:45
haven't had a chance to check it out be
00:18:47
sure to click the link below this video
00:18:49
you probably couldn't get away from the
00:18:52
big announcement from meta they released
00:18:55
their brand new llama 3.1 model model
00:18:59
and it's completely open source and
00:19:01
available to everyone to use now the
00:19:04
thing that makes this model really
00:19:05
unique is the fact that it has a context
00:19:08
window of
00:19:09
128,000 tokens you can basically upload
00:19:12
any sized script you want to llama and
00:19:15
be able to ask it questions about the
00:19:17
document which is really interesting
00:19:19
inside their terms of service basically
00:19:21
you can do anything you want with the
00:19:23
model but you must attribute llama and
00:19:26
include the words built with llama
00:19:29
inside any project that utilizes the
00:19:31
tool I thought that was really funny
00:19:33
just the idea of like watching a film
00:19:34
and it saying built with llama at the
00:19:36
end and so I created this icon using mid
00:19:38
Journey that is our built with llama
00:19:41
icon and then I animated it inside of
00:19:44
luma and so we have this looping AI
00:19:46
video of this llama just kind of head
00:19:48
bobbing here and I think this is just
00:19:50
delightful if you use llama on a film
00:19:52
Project you are free to use this clip
00:19:54
I'll include a link in the description
00:19:55
of this video there are also a lot of
00:19:57
really cool research papers that dropped
00:19:59
this week let's take a look at some
00:20:01
futuristic film Tech the first research
00:20:04
paper comes from the team at stability
00:20:06
AI they released a paper on stable video
00:20:10
4D basically you have the ability to
00:20:13
upload a video and it will generate
00:20:15
eight different angles of that footage
00:20:18
so it creates a 3D model representation
00:20:21
of the video and it basically allows you
00:20:23
to change the camera movement now it is
00:20:26
very specific it's only for basic
00:20:28
basically 3D model videos at this point
00:20:31
but it's very easy to see that in the
00:20:33
near future you might have the ability
00:20:35
to literally change the framing of your
00:20:38
video using a technology like stable
00:20:41
video 4D not to be outdone a team from
00:20:43
China also released code for a very
00:20:46
similar feature that's available to you
00:20:48
right away basically you can upload a 3D
00:20:51
object and this tool will automatically
00:20:54
animate that object and give you
00:20:56
multiple camera angles so it takes it
00:20:57
even one step step further they have
00:20:59
some pretty cool examples over on their
00:21:01
website for example we have this robot
00:21:03
arm kind of like animates down which is
00:21:05
pretty cool we have this shot of this
00:21:07
bald eagle flying now the animation is
00:21:10
not perfect but it's a start it also can
00:21:12
take scans from The Real World and
00:21:14
animate them so for example we have this
00:21:16
shot of this hand here and I'm not
00:21:18
entirely sure if fingers are designed to
00:21:21
go back that far I think uh there might
00:21:24
be some hyperextension Happening Here we
00:21:26
also have the shot of this dog like back
00:21:28
up I don't know if he's like scared of
00:21:30
something and we have the shot of this
00:21:32
Swan here and he's like Moon walking
00:21:34
backwards and this tool is also able to
00:21:36
automatically animate 3D assets that
00:21:39
were generated using AI so for example
00:21:41
we have this 3D Iguana here and he's
00:21:43
kind of walking around which is cool we
00:21:46
also have this shot of this dog which is
00:21:48
pretty interesting you can see the dog
00:21:49
is kind of like looking up at its owner
00:21:51
and kind of getting ready to sit down
00:21:53
and we have this shot of this kind of 3D
00:21:56
animated rabbit a research team from
00:21:58
Prague and Zurich came out with a white
00:22:01
paper called wild gaussians basically
00:22:04
this tool can take a ton of different
00:22:06
images and separate them based on the
00:22:08
time of day that they were shot then it
00:22:11
takes the images and creates a gaussian
00:22:13
Splat of the scene this allows you to
00:22:15
change the time of day as you move
00:22:18
around in 3D space it's pretty cool and
00:22:20
you can really imagine how this type of
00:22:22
Technology will be integrated in tools
00:22:25
like Google Maps in the future basically
00:22:27
any photo that's been uploaded to the
00:22:29
internet and ingested by Google could
00:22:31
hypothetically be stitched together to
00:22:33
where you could explore the world and
00:22:35
change the time of day and you would
00:22:37
have a pretty good 3D representation of
00:22:40
what the world looks like there's also
00:22:42
some pretty interesting news for VFX
00:22:44
artist and 3D artist basically when you
00:22:46
have a 3D asset that you're trying to
00:22:48
composite into a real life film scene
00:22:52
you need to go in and basically shoot a
00:22:56
360° image or video of the scene so that
00:22:59
you can have proper Reflections well
00:23:02
that has now changed with a brand new
00:23:05
tool it's called 3D gaussian Ray tracing
00:23:08
and basically it introduces Ray tracing
00:23:11
into gaussian Splats to where you can
00:23:14
take a 3D scan of a film set or an
00:23:17
environment and you'll be able to create
00:23:20
accurate 3D representations of what the
00:23:22
scene looks like so there's some pretty
00:23:24
interesting examples over on the website
00:23:27
for example we have the shot here of a
00:23:28
chrome car and you can see the
00:23:30
reflections look realistic like it looks
00:23:32
like that Chrome car is inside of the
00:23:34
scene it also has the ability to clean
00:23:36
up artifacts that are inside of your
00:23:38
scene and you can control the depth of
00:23:40
field which is really cool In Crowd
00:23:42
generation news there was a new white
00:23:45
paper called crowd Moen that basically
00:23:48
is an AI tool that intelligently
00:23:51
understands crowds this tool is
00:23:53
specifically designed for the
00:23:55
entertainment industry and it
00:23:57
intelligently understands crowd Dynamics
00:23:59
so for example right now if you want to
00:24:02
create a crowd simulation scene for a
00:24:04
film you have a few different options
00:24:06
the first is you can shoot a video plate
00:24:09
of a few different people and basically
00:24:12
copy and paste them to fill the stands
00:24:14
of let's say it's a sports arena you
00:24:17
also have the ability to use motion
00:24:18
capture or 3D scanning to basically get
00:24:21
a 3D model and you can duplicate that
00:24:23
model and create a crowd that way and
00:24:25
there are also procedural options using
00:24:28
tool like Houdini that allow you to do
00:24:30
slight variations to create different
00:24:32
crowd Dynamics this tool is more
00:24:34
interesting because it intelligently
00:24:36
understands how the crowd should behave
00:24:38
let me show you what I mean for example
00:24:40
the artist prompted a person falls down
00:24:44
and clearly there's a person that fell
00:24:45
down and someone's helping them up but
00:24:47
it also understands that when someone
00:24:49
falls down in a public space usually
00:24:51
other people run to help the person and
00:24:54
so it actually generated the movements
00:24:56
of other people going to help this
00:24:57
person out we also have a prompt here of
00:25:00
people running to avoid a car and you
00:25:02
can see that as the car approaches they
00:25:04
all kind of go their separate ways now
00:25:06
is this tool ready for prime time I
00:25:08
don't think so but it is interesting
00:25:10
that there are notable enhancements In
00:25:13
Crowd Dynamics and simulations crowds
00:25:16
are notoriously one of the more
00:25:18
challenging things to render because
00:25:20
there's just so many different Dynamics
00:25:22
and physical considerations that go into
00:25:25
creating a realistic crowd scene our
00:25:27
final white paper for this week comes
00:25:29
from Korea where a team put together a
00:25:31
project called click gaussian long story
00:25:34
short this tool allows you to select
00:25:36
objects from a gaussian Splat and scale
00:25:39
them up and move them around your scene
00:25:42
having the ability to change an
00:25:44
environment that you scanned gives you
00:25:46
greater creative control whenever you
00:25:48
hop into a 3D application and that
00:25:51
brings us to our AI films of the week
00:25:54
the first film that I want to highlight
00:25:56
is a film concept by Cavin the kid this
00:26:00
project is called Untold and it's one of
00:26:03
the best visual AI projects that I've
00:26:05
seen today Kavin is very skilled at
00:26:08
curating and selecting shots and that's
00:26:11
why he's one of the best AI filmmakers
00:26:13
out there the next project that I want
00:26:15
to highlight is called aari by Julie
00:26:19
Wheeland Julie is a fantastic AI artist
00:26:22
and she explored this beautiful vintage
00:26:24
aesthetic that takes place in Paris and
00:26:27
what I'm guessing is like the' 60s it's
00:26:30
clearly using Luma or some sort of AI
00:26:33
video tool that is more advanced and I
00:26:35
just love how the entire project seems
00:26:38
like it's a vintage travel film so
00:26:40
fantastic work Julie and actually we
00:26:42
interviewed Julie on our podcast a few
00:26:44
months ago if you want to learn more
00:26:45
from Julie and hear how she gets
00:26:48
consistency in her shots you can check
00:26:50
out that episode and our final project
00:26:52
of the week comes from our friend Tim at
00:26:55
theoretically media he put together a
00:26:57
project called called Dead Sea and it
00:27:00
really is one of the better AI films
00:27:02
that we've seen up to this point he
00:27:05
really cleverly used a flashback to tell
00:27:08
a story and I think that's a great way
00:27:10
to get around some of the limitations
00:27:12
with lip sync inside of AI video it has
00:27:16
fantastic visuals there's good
00:27:18
consistency and on a really exciting
00:27:21
note I want to celebrate that Tim hit
00:27:23
100,000 subscribers on YouTube
00:27:26
congratulations Tim you are one of the
00:27:29
best resources for AI video creators out
00:27:32
there and we wish you the best of luck
00:27:34
with your channel and if you watched the
00:27:36
video up to this point then you've
00:27:38
earned some secret insights into how to
00:27:40
win the game of the week this week my
00:27:43
tip to you is that cing likes to create
00:27:46
slow motion video in most of the
00:27:49
generations thank you so much for
00:27:51
watching this week's episode of AI film
00:27:54
news of course you can get AI film news
00:27:56
sent directly to your email email inbox
00:27:58
by subscribing over at curious Refuge we
00:28:02
also have a free AI film making course
00:28:04
that you can get completely for free and
00:28:07
of course you can join our Discord
00:28:08
Channel where we share the latest AI
00:28:11
news updates and meetups from around the
00:28:13
world and like always you can like And
00:28:16
subscribe here on YouTube to get the
00:28:19
latest tutorials and news directly here
00:28:22
on the platform thank you so much for
00:28:24
being an awesome member of the community
00:28:27
remember we have have enrollment for our
00:28:29
August session opening up on July 31st
00:28:32
at 11:00 a.m. Pacific time we would love
00:28:35
to see you inside of the courses thanks
00:28:37
for watching we'll see you next time
