0
00:00:02.900 --> 00:00:05.425
Hello. We are PLUME-D Inc., a company creating 'Avakit', 

1
00:00:05.425 --> 00:00:07.120
which is a real-time avatar full-tracking solution

2
00:00:07.120 --> 00:00:09.625
that you can begin with just one webcam

3
00:00:10.375 --> 00:00:14.120
Have you ever imagined characters on the screen 
mirroring your actions and expressions?

4
00:00:14.120 --> 00:00:17.250
We at PLUME-D are creating such technology

5
00:00:17.825 --> 00:00:20.425
There is a market with a growth rate of over 30%

6
00:00:20.425 --> 00:00:22.800
It is the virtual live streaming market

7
00:00:22.800 --> 00:00:25.225
Although the current market size is not large,

8
00:00:25.225 --> 00:00:28.425
it is growing faster than any other market

9
00:00:28.920 --> 00:00:30.775
Leading this market

10
00:00:30.775 --> 00:00:34.025
are the virtual YouTubers, also known as 'VTubers'

11
00:00:34.320 --> 00:00:38.026
These are streamers who are real people wearing hardware devices

12
00:00:38.026 --> 00:00:41.840
with avatars that mimic their actions and expressions in real-time

13
00:00:41.840 --> 00:00:44.250
to represent themselves on screen

14
00:00:44.926 --> 00:00:48.226
When you look at the monthly views of 
VTuber-related content on YouTube

15
00:00:48.226 --> 00:00:50.800
they have been steadily increasing since 2016

16
00:00:50.800 --> 00:00:53.200
recently reaching record views of 2 billion

17
00:00:54.026 --> 00:00:55.676
However, existing corporate solutions

18
00:00:55.676 --> 00:00:58.600
are failing to share the visions of virtual YouTubers

19
00:00:58.600 --> 00:01:03.876
On my left is the type of rendering, appearance, and texture

20
00:01:03.876 --> 00:01:06.320
 the virtual YouTubers and viewers actually want

21
00:01:06.320 --> 00:01:08.276
However, domestic and foreign tech companies

22
00:01:08.276 --> 00:01:12.550
are adopting rendering methods 
that are close to 3D avatars and game characters

23
00:01:13.000 --> 00:01:16.276
Insufficient tracking technology is also causing problems

24
00:01:16.276 --> 00:01:19.760
There are motions and frames where the body looks unnatural

25
00:01:19.760 --> 00:01:22.690
which can lead to live-streaming mishaps

26
00:01:23.076 --> 00:01:24.826
Also, the price of tracking equipments

27
00:01:24.826 --> 00:01:28.840
ranges from a minimum of 3 million won 
to a maximum of 30 million won,

28
00:01:28.840 --> 00:01:31.326
so there is a financial burden in terms of price

29
00:01:31.326 --> 00:01:36.040
In addition, many softwares are not user-friendly

30
00:01:36.040 --> 00:01:40.176
However, thanks to our R&D personnel
specializing in AI vision and VR development,

31
00:01:40.176 --> 00:01:43.626
we are receiving feedback from more than 3,000 users

32
00:01:43.626 --> 00:01:46.080
through Discord and Twitter

33
00:01:46.080 --> 00:01:47.827
Based on this feedback

34
00:01:47.827 --> 00:01:51.427
we are using a deep learning model called AI Human Pose Estimation

35
00:01:51.427 --> 00:01:54.227
to create a full-body tracking model called Avakit

36
00:01:54.840 --> 00:01:57.477
When a user films their video with a webcam,

37
00:01:57.477 --> 00:02:01.200
Avakit extracts the user's motion sequence 
from that video in real-time

38
00:02:01.200 --> 00:02:03.227
Then, by combining and rendering the extracted motion sequence

39
00:02:03.227 --> 00:02:06.277
on the 3D avatar selected by the user,

40
00:02:06.277 --> 00:02:09.877
the user can create an avatar that mirrors their actions in real-time

41
00:02:10.240 --> 00:02:15.000
Our advantage is that you can start all this 
with just one camera anywhere, anytime

42
00:02:15.000 --> 00:02:18.090
Also, we support up to 120FPS

43
00:02:18.840 --> 00:02:20.827
We are currently proceeding with two versions,

44
00:02:20.827 --> 00:02:23.320
Avakit Free version and the Pro version

45
00:02:23.320 --> 00:02:26.577
As for the Pro version, there are two options 
which are the one-time payment version

46
00:02:26.577 --> 00:02:30.177
and the monthly subscription version 
which includes the full-tracking version

47
00:02:31.000 --> 00:02:33.920
We are currently running an open beta test on Stream

48
00:02:33.920 --> 00:02:37.840
We have attracted over 3,100 organic users 
in approximately three months

49
00:02:37.840 --> 00:02:40.600
Avakit is being used in more than 80 countries

50
00:02:40.600 --> 00:02:43.677
Currently, it is being used the most in Korea

51
00:02:43.677 --> 00:02:47.377
But the number of users in the United States and Japan is increasing at a rapid pace

52
00:02:48.080 --> 00:02:51.827
Currently, we only have a monthly subscription software sales business model

53
00:02:51.827 --> 00:02:54.377
But we have three scale-up strategies

54
00:02:54.800 --> 00:02:59.428
The first scale-up strategy is a customized image creation service based on character IP

55
00:02:59.428 --> 00:03:00.528
Which is AvaMaker

56
00:03:00.830 --> 00:03:06.320
If you upload 6 to 10 photos to the server, we fine-tune it.

57
00:03:06.320 --> 00:03:10.678
Based on this character, you can easily and quickly create images

58
00:03:10.678 --> 00:03:13.228
where the character you want is the main focus

59
00:03:13.720 --> 00:03:17.490
We had CBT with virtual YouTubers

60
00:03:18.440 --> 00:03:20.320
The second strategy is an R&D-based scale-up

61
00:03:20.320 --> 00:03:23.560
The problem with our first deep learning model we are currently using is that

62
00:03:23.560 --> 00:03:25.928
there are unnatural frames in the lower body part

63
00:03:25.928 --> 00:03:29.360
and there are times when it can"t keep up with fast movements

64
00:03:29.360 --> 00:03:32.720
We are currently developing a second model that will make up those problems perfectly

65
00:03:32.720 --> 00:03:37.000
Based on this, we aim to sell entertainment agencies and various B2B motion solutions

66
00:03:37.600 --> 00:03:41.320
The third strategy is the expansion into virtual entertainment agency, MCN

67
00:03:41.320 --> 00:03:44.778
Not only the avatar and technical support, but also by content production support

68
00:03:44.778 --> 00:03:46.628
We plan to continuously expand our IP

69
00:03:46.628 --> 00:03:48.978
We plan to share the revenue from them

70
00:03:49.630 --> 00:03:52.278
Currently, we are only focusing on the virtual live streaming market

71
00:03:52.278 --> 00:03:55.578
But ultimately, we aim to expand into the character entertainment market

72
00:03:56.520 --> 00:03:58.528
We established the corporation last year in June,

73
00:03:58.528 --> 00:04:00.800
and successfully attracted seed investment from Futureplay at the same time

74
00:04:00.800 --> 00:04:04.840
We have also been selected by TIPS and are undergoing technology advancements

75
00:04:04.840 --> 00:04:07.179
And showed excellent results

76
00:04:07.179 --> 00:04:09.129
In the 'Jeong Joo-young Startup Competition' and 'Challenge! K-Startup'

77
00:04:09.720 --> 00:04:12.960
We are a team created by KAIST graduates

78
00:04:12.960 --> 00:04:15.529
Not only our team members have technical skills,

79
00:04:15.529 --> 00:04:19.179
but also have a deep understanding of the virtual market and subculture

80
00:04:19.179 --> 00:04:22.329
and are working hard to create prompts that suits VTubers

81
00:04:23.200 --> 00:04:26.960
We will develop technologies which will enable everyone to showcase their charms at best

82
00:04:26.960 --> 00:04:28.960
We were PLUME. D, thank you