0:00
gamescom
0:01
hasn't been done for a couple of years
0:03
since covet and this was the first time
0:06
we kind of went
0:07
and met all our friends and tried to
0:09
make new connections at the recently
0:12
opened event
0:13
it was great meeting everybody and it
0:15
was even more exciting to kind of
0:17
explore the show floor
0:19
uh for the first time in two years and
0:22
we've seen a lot of uh returning
0:24
pavilions like there were a lot of
0:26
spaces where you could find you know
0:28
companies from a particular area in
0:31
europe like you know spain or area and
0:34
asia like malaysia
0:35
and some other spaces there were a lot
0:37
of publishers as well uh there were not
0:40
that many big publishers out there but
0:42
it was still interesting to see a lot of
0:45
local european houses kind of providing
0:47
and showing the games they're building
0:49
we haven't seen that many tools and tool
0:52
developers at gamescom seems like this
0:56
is not the kind of event that developers
0:58
of new technology for game production
1:00
usually like to attend and a lot of them
1:04
actually they don't have they do attend
1:06
but they don't have boost there but
1:07
there was one company that we were very
1:10
fortunate to have a little
1:12
time with
1:14
and
1:15
talk about
1:16
the future of video game production and
1:19
the future of animation
1:21
in particular
1:22
this company is plask
1:24
and i'm very happy to present our
1:26
interview with the ceo of this amazing
1:29
new startup
1:30
greetings and welcome to the 80 level
1:32
roundtable podcast in each episode host
1:36
corel tokorev invites video game
1:38
industry leaders to talk about the world
1:40
of game development
1:42
no topic is off limits as long as it
1:45
relates to video game development
1:47
new episodes are in the works so
1:49
remember to follow us or subscribe and
1:51
share with someone you know will also
1:53
enjoy the podcast so
1:57
before we kind of start
2:00
we ask this question a lot to our uh the
2:02
people that we're just talking with
2:04
do you actually play games
2:07
oh that's very um important question so
2:10
yes
2:11
yes but like i'm quite maniac i would
2:14
say because you know um
2:17
i'm not quite sure you know that but in
2:20
korea most of people play either
2:22
starcraft or
2:24
league of legends
2:25
so i'm not kind of the mainstream gamer
2:28
in queen
2:29
society but i like to play the
2:32
civilization or like oh really yeah
2:34
those kind of jump some boring games i
2:36
like those kind of things like this is
2:38
weird but a lot of people
2:41
who are doing business
2:43
in tech
2:44
they're really into civilization oh i
2:47
think yeah i think that's not a
2:49
coincidence yeah yes that's not a
2:51
coincidence like i think mark zuckerberg
2:53
he's into civilization
2:55
the boss of our company he's also a big
2:58
fan of civilization yeah and it seems
3:00
like everyone who's like
3:01
like to manage and figure those things
3:04
out they yeah even playing games
3:06
we'd like to manage but um
3:11
kind of what brought you to gamescom
3:13
this year what are you showcasing at
3:15
this show
3:16
yeah i am i can say that i'm a gamer but
3:20
the game uh the reason coming to
3:22
gamescom is not kind of poor like really
3:24
to enjoy the game industry but
3:27
it's more like the business
3:28
purpose because you know flask have been
3:31
on a great infrastructure for game
3:34
companies and productions and
3:36
we know that we can make the
3:38
productivity of the game industry really
3:41
really improved by
3:42
um
3:44
our service so i think we think it's a
3:47
great opportunity to introduce our tool
3:49
to the industry and show what we're
3:52
doing right now yeah that's i think the
3:54
main purpose
3:56
so tell us a little bit about your
3:58
product what does uh flask do what kind
4:01
of products do you provide
4:03
a little bit about that direction um so
4:05
yeah flask is an ai motion capture
4:08
[Music]
4:09
software we provide uh so you know
4:12
motion capture needs a lot of sensors
4:14
suits studios especially money and time
4:17
and class don't need any of those so it
4:20
is uh it only uses a video in order to
4:22
get a 3d motion data so
4:25
what you have to do is just
4:27
a prepare a video it can be video taken
4:29
from your phone or download it from
4:31
youtube it all works
4:32
so you just can drag and drop the video
4:35
into our tool our tool is a
4:37
browser-based tool and no need for
4:39
installations so you just can drag and
4:40
drop video into the tool and you get the
4:42
3d motion data so
4:44
uh this makes the whole production much
4:47
much easier because you know animating
4:49
became the animation was uh
4:52
the most bottleneck of the production it
4:54
takes a lot of time a lot of effort but
4:57
just by dragging and drop and video and
5:00
getting up motion data is a really big
5:02
thing especially for like
5:04
the 3d games i mean the mmr pitch or
5:06
like big scale games they can save a lot
5:09
of time and money uh by using flask uh
5:12
for making their animation process much
5:14
more efficient
5:15
so
5:16
how does it work usually
5:18
when you think about motion capture like
5:20
you said you need to have a suit there
5:22
has to be cameras around right there has
5:24
to be trackers that are on you and this
5:26
allows for a very accurate
5:28
representation of motion yeah especially
5:31
when you think about
5:32
um you know facial animations or some
5:35
very complex motions when you're right
5:37
involving fingers or some other elements
5:40
those are super important how does
5:43
your technology work like what's the
5:45
secret there how does it know how to
5:48
track the
5:49
you know motion from the image which is
5:51
basically flat as far as in the exact
5:54
video so the previous method the motion
5:56
capture uh uses a optical or like
6:01
electromagnetic sensors in order to get
6:03
the exact position of its joints so you
6:07
would probably see that they put markers
6:09
on their suits so they are using the
6:12
cameras or the sensor setup in order to
6:15
get the position of that marker and
6:18
making it to recover into the motion
6:20
data that's how they used to do uh so
6:23
they are basically using um
6:26
the tracking sensor that's the main
6:28
difference we are using the ai
6:31
or like
6:32
uh computer vision technology to get uh
6:35
motion reconstructed from video so
6:37
it's like a human eye or human brain or
6:39
like human intelligence because
6:42
when we see a video we are not 100 sure
6:45
about the death data or like 100 sure
6:48
about edge of uh the position of each
6:50
joint but we somehow know
6:53
um
6:54
the
6:56
approximated uh position of each data
6:59
while you can see that someone come
7:02
closer to camera
7:05
or like going far from camera it all can
7:07
be reconstructed in human brain so
7:10
that's what ai do ai was trained by
7:13
3d motion data so they were trained to
7:16
estimate the
7:18
30 joints of
7:20
each joint of the person in the video by
7:22
looking at huge data set that's how our
7:25
ai reconstruct 3d motion data from video
7:28
and it allows it to
7:30
by not uh
7:32
getting a 3d depth data or like
7:35
sensor data it somehow
7:37
can get
7:39
reconstructed 3d positions so
7:43
when you think about this technology how
7:45
accurate
7:46
is the
7:47
traction of movement how accurate can
7:49
you go
7:50
so we should talk about uh the prison
7:53
state and how it can go beyond
7:57
so for now we can say that our
8:00
okay quality is like low budget motion
8:03
suits so compared to like high and
8:06
motion call motion capture studios yes
8:09
our motion data is a little bit
8:12
[Music]
8:14
noisy and like
8:15
uh many users are giving feedback that
8:18
the feeds are like
8:20
quite slightly
8:21
those are still we have still those have
8:24
problems but the fun part is that last
8:27
year
8:28
the ai
8:30
quality was like too awful that it isn't
8:33
like serviceable and two years ago
8:36
no one think about
8:38
to make this into service because you
8:40
know it was that terrible
8:42
and you know uh the growth of the ai
8:45
technology is really fast and we
8:47
estimate that most of current uh users
8:50
feedback i mean the fit sliding and
8:52
southern noise problems will be so
8:54
within four and five months and after
8:57
one year later i think uh we can
9:00
approach to the quality of
9:03
middle and uh
9:04
middle class motion capture suit and
9:07
like two or three years later you can go
9:11
like similar level with the high quality
9:13
motion capture studios
9:15
so that's the thing i guess um the
9:17
future of the motion capture should be
9:20
and
9:21
must be the ai motion should capture
9:23
because uh the performance will
9:26
uh finally converge into the high
9:28
quality market and um
9:31
but the
9:32
accessibility or the
9:35
um
9:36
the efficiency the amount gives to the
9:39
production is uncomparable so
9:42
yeah that's what we are estimating about
9:45
the performance
9:47
so
9:48
let's talk a little bit about how it
9:50
works so as far understand you're a
9:51
web-based tool right you upload some
9:53
kind of video there then the ai does its
9:56
magic and it looks at the
9:58
human positions and
10:00
what happens next does your ai also
10:03
create a rig for the character can you
10:06
then kind of export this rig to any
10:08
character that you have in the game and
10:10
kind of end the animation as well right
10:12
so yeah um
10:15
so when you uh drag and drop the video
10:17
tool you get the rotation data of 203
10:21
joints which is exactly same with the
10:23
mixed summary and what we are doing is
10:25
we support an automatic retargeting
10:28
feature so user just can drag and drop
10:30
their custom 3d model to of course issue
10:33
week
10:34
they re-character to our tool and you
10:37
just can re-target the
10:39
ai extracted motion to they can apply
10:42
that to
10:44
their custom 3d characters so no we are
10:46
not um creating a rig for a 3d custom
10:50
character but
10:51
the ai extracting motion can be applied
10:54
to any of your humanoid characters so
10:57
we can say that we have some uh kind of
11:00
comparability so and uh after you apply
11:03
the motion to the
11:05
character you can export it to
11:07
fbx glbbh so you can bring it to other
11:10
tools like unity blender any other tools
11:12
you like right so
11:14
you need a rigged character to kind of
11:17
you utilize the motion that the
11:20
flask extracts from the video and after
11:23
that you can use it as uh
11:25
as much as possible whatever you can
11:27
right what about the
11:29
editing part so
11:31
if i want to create
11:33
like in
11:34
let's take for example like modern
11:36
action games or any kind of games right
11:38
now
11:40
if you've seen like last of us part two
11:42
yeah they did like
11:44
like a billion different animations for
11:47
every kind of movement that the
11:48
character can do and that's why the
11:50
character can open a door
11:52
six
11:53
different ways right or and it's like
11:55
every time you watch it it's like a
11:57
different motion
11:59
it requires a lot of motion capture but
12:01
it also requires a lot of editing
12:03
because you need to cut this motion and
12:05
that motion and so on so basically with
12:08
your data when it's exported can you
12:10
just
12:12
export it to maya i guess and play
12:14
around there and cut it into like
12:16
whatever pieces you want that's really
12:18
good point because
12:19
um actually um our internal
12:22
goal or like internal target is like
12:24
two-way the first one is uh to
12:27
automating the animation industry with
12:30
ai that's the one big part and the next
12:33
big part is collaboration so
12:36
um we are adding more and more uh
12:38
editing features to the tool
12:41
and we choose a browser to be the
12:44
platform uh of our tool uh there are a
12:47
lot of reasons so
12:49
uh as you just told you as you just told
12:52
so
12:53
animating is kind of like the
12:57
work of art you know it is not our just
13:00
a technical job so there should be a lot
13:03
of editing jobs after even if they
13:06
capture a motion from a fancy motion
13:08
capture studio they have to edit the
13:10
motion to make it into the form of art
13:12
so um editing is like the one of the
13:15
essential part when it comes to game
13:17
animations and we want but
13:20
what people are doing right now is quite
13:22
an old way it's like the way just like
13:25
30 years ago and um we think there
13:29
should be about two things two
13:31
innovation should be brought to the
13:33
industry the first one is about the
13:34
collaboration so you know
13:38
figma is a great example to the ui ux
13:41
area they bring a real-time
13:42
collaboration to the ui ux design and
13:45
they make the whole
13:47
stage to the
13:49
browser and it makes a lot of people to
13:51
access uh the project and giving a
13:54
feedback working simultaneously and
13:57
making something into a
13:59
like
14:00
i am actually a a developer before i
14:03
start a
14:04
company and we developers
14:07
uh have some like very good culture i
14:10
would say it's like the like you know
14:12
there's a git top there and there's an
14:13
open source thing and there's
14:16
a collaboration between um total
14:18
strangers those kind of things should be
14:20
brought to the game
14:22
game design animation
14:24
as well i guess so that's why we are
14:27
adding more and more editing feature to
14:29
our browser-based tool and making it
14:30
into a real-time collaboration
14:32
uh or in uh
14:34
yeah i think those what
14:36
those are one of our great targets and
14:38
those kind of editing features that's
14:40
why we are adding more and more editing
14:42
feature to our tool yeah
14:45
so
14:45
i like this um analogy with sigma
14:48
because i think i use figma a lot
14:50
although i'm not a
14:52
designer but i do use it to kind of
14:53
review different forms and
14:56
it seems right that
14:58
kind of like sigma grew from just like a
15:01
ui
15:03
building tool to like more like a
15:05
photoshop kind of tool or like photoshop
15:08
plus coral draw because it's like vector
15:10
stuff and raster stuff and all of it
15:12
combined exactly and it's uh it is very
15:15
approachable so you can download it and
15:18
start playing around in the browser
15:20
basically without having to use
15:22
any other stuff
15:24
that you do with other comparable tools
15:27
um
15:28
like especially the 3d animation thing
15:31
the assets are too heavy and
15:34
the tool itself the maya the 3d max when
15:38
you mis-click the tool and it takes
15:41
on years to open up and yeah i think
15:44
those make the review and feedback
15:46
process really slow and
15:48
i think when it comes to browser it's
15:50
really really
15:52
fast accessible thing so i think it
15:54
would be great if you can bring it to
15:56
the industry
15:58
when i think about
15:59
[Music]
16:00
maya in particular
16:04
the rigs themselves
16:06
they can get pretty complex right right
16:09
and when we think about motion in
16:11
general
16:12
it is like like you said a very
16:14
complicated thing and that's why we also
16:17
run like a little bit of a recruiting
16:19
uh service and we see a lot of demand
16:22
for guys like technical animators or
16:24
animators in general because
16:27
like you said it's like you're sparkling
16:29
life
16:30
into a character certainly given motion
16:32
and making it sure that it moves
16:35
i feel like what you're saying about ai
16:38
could be
16:40
could bring out like the next level to
16:43
animation in general and the fluidity of
16:46
motion the versatility of motion because
16:49
right now everything is basically done
16:51
by hand i mean of course mocap helped a
16:55
lot
16:56
definitely took a lot of
16:58
pressure from animators but it's still a
17:00
lot of things that you need to do
17:03
for yourself or in teams and a lot of
17:05
cleaning like you said a lot of
17:07
you know
17:09
adapting the material and so on
17:11
when
17:13
we have ai a lot of those tasks are
17:15
being done by the machine
17:17
and my question is
17:19
what else do you feel in animation can
17:22
we outsource to computers so we don't
17:25
have to
17:26
think about it like
17:28
i think there is another piece of
17:30
software called the cascader or
17:32
something yeah they're doing a lot of
17:35
that like in that
17:36
it's basically like a keyframe approach
17:39
but aai kind of adds up and helps smooth
17:43
it out or right suggest motion or
17:47
help figure out when the impact is
17:49
happening what's going on with the
17:50
motion and the physics and all that
17:52
stuff
17:53
where do you think it's going to go from
17:55
here
17:56
how is plask kind of
17:58
going to push this ai into the animation
18:02
industry of course animation itself have
18:05
a lot of things to can be um automated
18:08
by ai because
18:10
i think at the end of
18:12
this world
18:13
maybe we can have um some kind of
18:15
animation features like
18:17
just like by using the natural language
18:19
maybe you can uh make a whole animation
18:22
so you are just like being a director so
18:25
like just move
18:26
in this way more or like just do it more
18:29
with a passion this can be be done with
18:32
ai because you know
18:34
similar things are done by the dali too
18:37
recently you know the natural language
18:39
can produce a
18:40
whole
18:41
artworks and
18:43
this can be applied to animations as
18:45
well so
18:46
those things so animation itself can be
18:48
really really
18:50
easy by using those kind of things but
18:52
not only for animation
18:54
so so example did a great job with uh
18:57
automatic rigging uh don't buy the
18:59
uh ai but you know rigging itself is
19:02
actually have much more
19:04
way to go than just mixamo because all
19:07
the deformation things and all the you
19:10
know that's why we have technical
19:12
artists in the game industry so i think
19:15
our next thing after we conquered the
19:17
animation job would be the
19:20
uh rigging thing and after that maybe we
19:22
can work with the building thing there
19:24
are a lot of things can be improved by
19:27
ai so
19:29
i think those kind of automations will
19:31
help
19:32
like i think this can be like youtube
19:35
thing so back then only videos were
19:38
mostly produced by the broadcasting
19:40
uh giants and normal people can make the
19:44
videos and shows but now everyone do
19:46
right
19:47
um i think after those kind of
19:50
animation after reading after modeling
19:53
after all the 3d pipeline productions
19:55
are automated i think really um everyone
19:58
would produce like
20:00
team of 10 make a
20:02
large-scale game and get
20:05
a like
20:06
hundreds millions of revenue i think
20:08
that would be possible after all the
20:12
um
20:13
bothering jobs will be automatically
20:15
automated do you feel like
20:18
uh adoption of ai tools will negatively
20:23
impact you know the workforce do you
20:25
think more people are going to be left
20:26
out of jobs
20:29
yeah that's very important question um
20:31
um no
20:33
because um
20:35
i think this kind of question was like
20:38
repeatedly when there's a new technology
20:41
coming to the industry and everyone uh
20:43
worrying about it replacing their jobs
20:46
but
20:47
what i found out is that
20:49
when
20:51
there is an ai coming to the industry
20:53
and replace the most
20:54
hard or like most repetitive jobs and
20:58
people just
21:00
move to more created jobs
21:02
our
21:03
goal is to
21:04
make every uh
21:06
members in the game team to focus on
21:09
their own creativity that's what we are
21:12
aiming for so like when ai makes the
21:15
whole draft by ai mocap then every
21:19
animators can now um work with the
21:22
sensitive the artistic or like the
21:25
uh craftsmanship thing to their emotions
21:28
and it can make their emotions like more
21:30
creative more sensitive more high
21:33
quality i think those will
21:36
yeah come to the area because you know
21:39
when there's like
21:43
yeah
21:44
so uh
21:45
final question how do we
21:47
uh start playing around with plus do we
21:49
just go online and click the link and
21:51
start working you don't need to
21:53
do you need to pay is there like a trial
21:55
period how does it work
21:57
um yeah you just visit flask.ai after
22:00
you make an account you can start with
22:03
plus right away we have a freemium plan
22:05
so you can have most of the features
22:08
with most of the credit uh most of those
22:10
you can use most of the feature without
22:12
any kind of payment method registration
22:15
but actually we are releasing first paid
22:17
version this weekend and it will have
22:20
some amazing explosive features uh like
22:24
uh previously we only can extract the
22:26
motion from a video with a one person
22:28
inside but uh with a newly coming paid
22:32
version that you can now
22:34
finally extract the motion of multiple
22:36
people in the video each of the motions
22:38
yeah those will be
22:41
will introduce the feature to the paid
22:43
version and there will be much more um
22:46
minor like
22:48
um features which which can make your uh
22:52
production much more effective so
22:54
maybe but the paid version will have the
22:56
free trial as well so you can try it out
22:59
and if you love it you can buy it but
23:02
our
23:03
vision is to provide but
23:05
our short-term vision is to introduce
23:07
the am or capture the industry that's
23:09
the first thing
23:10
more than a monetization so we will
23:12
focus on
23:14
more to provide or provide more ai-based
23:17
mocap without any cost so
23:20
maybe don't have to worry too much
23:23
all right thank you so much for your
23:24
interview
23:26
thanks for enjoying another episode of
23:28
the 80 level roundtable podcast
23:31
check out upcoming episodes on the 80
23:33
level website at 80.lv
23:36
join our career site at 80.lv
23:39
rfp
23:41
and share our podcast with friends and
23:43
on your social networks
23:47
[Music]
