1
00:00:01,100 --> 00:00:04,519

covering the week's top textbooks like

2
00:00:04,519 --> 00:00:04,529
covering the week's top textbooks like
 

3
00:00:04,529 --> 00:00:07,639
covering the week's top textbooks like
Linux - a new paper published by Disney

4
00:00:07,639 --> 00:00:07,649
Linux - a new paper published by Disney
 

5
00:00:07,649 --> 00:00:09,980
Linux - a new paper published by Disney
research describes a fully automated

6
00:00:09,980 --> 00:00:09,990
research describes a fully automated
 

7
00:00:09,990 --> 00:00:12,049
research describes a fully automated
neural network based method for swapping

8
00:00:12,049 --> 00:00:12,059
neural network based method for swapping
 

9
00:00:12,059 --> 00:00:15,320
neural network based method for swapping
faces in photos and videos the first

10
00:00:15,320 --> 00:00:15,330
faces in photos and videos the first
 

11
00:00:15,330 --> 00:00:17,029
faces in photos and videos the first
method that results in high resolution

12
00:00:17,029 --> 00:00:17,039
method that results in high resolution
 

13
00:00:17,039 --> 00:00:19,760
method that results in high resolution
results of sufficient quality to be used

14
00:00:19,760 --> 00:00:19,770
results of sufficient quality to be used
 

15
00:00:19,770 --> 00:00:21,320
results of sufficient quality to be used
in film and TV

16
00:00:21,320 --> 00:00:21,330
in film and TV
 

17
00:00:21,330 --> 00:00:24,290
in film and TV
the researchers specifically intend this

18
00:00:24,290 --> 00:00:24,300
the researchers specifically intend this
 

19
00:00:24,300 --> 00:00:26,089
the researchers specifically intend this
tech for use in replacing it and

20
00:00:26,089 --> 00:00:26,099
tech for use in replacing it and
 

21
00:00:26,099 --> 00:00:28,310
tech for use in replacing it and
existing actors performance with a

22
00:00:28,310 --> 00:00:28,320
existing actors performance with a
 

23
00:00:28,320 --> 00:00:30,919
existing actors performance with a
substitute actors face for instance

24
00:00:30,919 --> 00:00:30,929
substitute actors face for instance
 

25
00:00:30,929 --> 00:00:33,440
substitute actors face for instance
wendy aging or increasing the age of

26
00:00:33,440 --> 00:00:33,450
wendy aging or increasing the age of
 

27
00:00:33,450 --> 00:00:35,900
wendy aging or increasing the age of
someone or potentially when portraying

28
00:00:35,900 --> 00:00:35,910
someone or potentially when portraying
 

29
00:00:35,910 --> 00:00:38,600
someone or potentially when portraying
an actor who has passed away they also

30
00:00:38,600 --> 00:00:38,610
an actor who has passed away they also
 

31
00:00:38,610 --> 00:00:40,459
an actor who has passed away they also
suggest it could be used for replacing

32
00:00:40,459 --> 00:00:40,469
suggest it could be used for replacing
 

33
00:00:40,469 --> 00:00:43,310
suggest it could be used for replacing
the faces of stunt doubles in cases

34
00:00:43,310 --> 00:00:43,320
the faces of stunt doubles in cases
 

35
00:00:43,320 --> 00:00:45,080
the faces of stunt doubles in cases
where the conditions of a scene call for

36
00:00:45,080 --> 00:00:45,090
where the conditions of a scene call for
 

37
00:00:45,090 --> 00:00:47,900
where the conditions of a scene call for
them to be used this method is unique

38
00:00:47,900 --> 00:00:47,910
them to be used this method is unique
 

39
00:00:47,910 --> 00:00:49,700
them to be used this method is unique
from other approaches in a number of

40
00:00:49,700 --> 00:00:49,710
from other approaches in a number of
 

41
00:00:49,710 --> 00:00:52,369
from other approaches in a number of
ways including that any face used in the

42
00:00:52,369 --> 00:00:52,379
ways including that any face used in the
 

43
00:00:52,379 --> 00:00:54,380
ways including that any face used in the
set can be swapped with any recorded

44
00:00:54,380 --> 00:00:54,390
set can be swapped with any recorded
 

45
00:00:54,390 --> 00:00:56,240
set can be swapped with any recorded
performance making it possible to

46
00:00:56,240 --> 00:00:56,250
performance making it possible to
 

47
00:00:56,250 --> 00:00:58,939
performance making it possible to
relatively easily re image the actors on

48
00:00:58,939 --> 00:00:58,949
relatively easily re image the actors on
 

49
00:00:58,949 --> 00:01:02,510
relatively easily re image the actors on
demand the other is that it recreates

50
00:01:02,510 --> 00:01:02,520
demand the other is that it recreates
 

51
00:01:02,520 --> 00:01:04,520
demand the other is that it recreates
contrast and lighting conditions to

52
00:01:04,520 --> 00:01:04,530
contrast and lighting conditions to
 

53
00:01:04,530 --> 00:01:06,170
contrast and lighting conditions to
ensure the actor looks like they are

54
00:01:06,170 --> 00:01:06,180
ensure the actor looks like they are
 

55
00:01:06,180 --> 00:01:08,690
ensure the actor looks like they are
actually present in the same conditions

56
00:01:08,690 --> 00:01:08,700
actually present in the same conditions
 

57
00:01:08,700 --> 00:01:11,300
actually present in the same conditions
as the scene you can check out the

58
00:01:11,300 --> 00:01:11,310
as the scene you can check out the
 

59
00:01:11,310 --> 00:01:13,570
as the scene you can check out the
results for yourself in this video as

60
00:01:13,570 --> 00:01:13,580
results for yourself in this video as
 

61
00:01:13,580 --> 00:01:15,920
results for yourself in this video as
you can see there's still a hint of

62
00:01:15,920 --> 00:01:15,930
you can see there's still a hint of
 

63
00:01:15,930 --> 00:01:18,020
you can see there's still a hint of
uncanny valley going on here but the

64
00:01:18,020 --> 00:01:18,030
uncanny valley going on here but the
 

65
00:01:18,030 --> 00:01:19,640
uncanny valley going on here but the
researchers acknowledge that in their

66
00:01:19,640 --> 00:01:19,650
researchers acknowledge that in their
 

67
00:01:19,650 --> 00:01:22,310
researchers acknowledge that in their
paper calling this a major step toward

68
00:01:22,310 --> 00:01:22,320
paper calling this a major step toward
 

69
00:01:22,320 --> 00:01:24,320
paper calling this a major step toward
photorealistic face swapping that can

70
00:01:24,320 --> 00:01:24,330
photorealistic face swapping that can
 

71
00:01:24,330 --> 00:01:26,590
photorealistic face swapping that can
successfully bridge the uncanny valley

72
00:01:26,590 --> 00:01:26,600
successfully bridge the uncanny valley
 

73
00:01:26,600 --> 00:01:29,210
successfully bridge the uncanny valley
it is still a lot more realistic than

74
00:01:29,210 --> 00:01:29,220
it is still a lot more realistic than
 

75
00:01:29,220 --> 00:01:30,770
it is still a lot more realistic than
other attempts which is especially

76
00:01:30,770 --> 00:01:30,780
other attempts which is especially
 

77
00:01:30,780 --> 00:01:32,240
other attempts which is especially
apparent when you've seen the

78
00:01:32,240 --> 00:01:32,250
apparent when you've seen the
 

79
00:01:32,250 --> 00:01:34,249
apparent when you've seen the
side-by-side comparisons with other

80
00:01:34,249 --> 00:01:34,259
side-by-side comparisons with other
 

81
00:01:34,259 --> 00:01:37,219
side-by-side comparisons with other
techniques most notably it works at much

82
00:01:37,219 --> 00:01:37,229
techniques most notably it works at much
 

83
00:01:37,229 --> 00:01:38,960
techniques most notably it works at much
higher resolution which is key for

84
00:01:38,960 --> 00:01:38,970
higher resolution which is key for
 

85
00:01:38,970 --> 00:01:41,620
higher resolution which is key for
actual entertainment industry use

86
00:01:41,620 --> 00:01:41,630
actual entertainment industry use
 

87
00:01:41,630 --> 00:01:43,639
actual entertainment industry use
considering the example of using this

88
00:01:43,639 --> 00:01:43,649
considering the example of using this
 

89
00:01:43,649 --> 00:01:45,800
considering the example of using this
technique for a stunt double the real is

90
00:01:45,800 --> 00:01:45,810
technique for a stunt double the real is
 

91
00:01:45,810 --> 00:01:47,810
technique for a stunt double the real is
realism could come across as being

92
00:01:47,810 --> 00:01:47,820
realism could come across as being
 

93
00:01:47,820 --> 00:01:51,050
realism could come across as being
incredibly realistic the examples

94
00:01:51,050 --> 00:01:51,060
incredibly realistic the examples
 

95
00:01:51,060 --> 00:01:54,469
incredibly realistic the examples
presented are a super small sample so it

96
00:01:54,469 --> 00:01:54,479
presented are a super small sample so it
 

97
00:01:54,479 --> 00:01:56,330
presented are a super small sample so it
remains to be seen how broadly this can

98
00:01:56,330 --> 00:01:56,340
remains to be seen how broadly this can
 

99
00:01:56,340 --> 00:01:59,270
remains to be seen how broadly this can
be applied subjects used appear to be

100
00:01:59,270 --> 00:01:59,280
be applied subjects used appear to be
 

101
00:01:59,280 --> 00:02:01,520
be applied subjects used appear to be
primarily white for instance also

102
00:02:01,520 --> 00:02:01,530
primarily white for instance also
 

103
00:02:01,530 --> 00:02:03,410
primarily white for instance also
there's always the question of the

104
00:02:03,410 --> 00:02:03,420
there's always the question of the
 

105
00:02:03,420 --> 00:02:05,600
there's always the question of the
ethical implication of any use of face

106
00:02:05,600 --> 00:02:05,610
ethical implication of any use of face
 

107
00:02:05,610 --> 00:02:08,029
ethical implication of any use of face
swapping technology especially in video

108
00:02:08,029 --> 00:02:08,039
swapping technology especially in video
 

109
00:02:08,039 --> 00:02:10,100
swapping technology especially in video
as it could be used to fabricate

110
00:02:10,100 --> 00:02:10,110
as it could be used to fabricate
 

111
00:02:10,110 --> 00:02:12,930
as it could be used to fabricate
credible video or photographic evidence

112
00:02:12,930 --> 00:02:12,940
credible video or photographic evidence
 

113
00:02:12,940 --> 00:02:15,840
credible video or photographic evidence
thing that didn't actually happen given

114
00:02:15,840 --> 00:02:15,850
thing that didn't actually happen given
 

115
00:02:15,850 --> 00:02:17,850
thing that didn't actually happen given
however that the technology is now in

116
00:02:17,850 --> 00:02:17,860
however that the technology is now in
 

117
00:02:17,860 --> 00:02:20,250
however that the technology is now in
development from multiple quarters it's

118
00:02:20,250 --> 00:02:20,260
development from multiple quarters it's
 

119
00:02:20,260 --> 00:02:22,140
development from multiple quarters it's
essentially long past the time for

120
00:02:22,140 --> 00:02:22,150
essentially long past the time for
 

121
00:02:22,150 --> 00:02:23,550
essentially long past the time for
debate about the ethics of its

122
00:02:23,550 --> 00:02:23,560
debate about the ethics of its
 

123
00:02:23,560 --> 00:02:26,820
debate about the ethics of its
development and exploration instead it's

124
00:02:26,820 --> 00:02:26,830
development and exploration instead it's
 

125
00:02:26,830 --> 00:02:28,890
development and exploration instead it's
welcomed at organizations like Disney

126
00:02:28,890 --> 00:02:28,900
welcomed at organizations like Disney
 

127
00:02:28,900 --> 00:02:30,720
welcomed at organizations like Disney
Research are following the academic path

128
00:02:30,720 --> 00:02:30,730
Research are following the academic path
 

129
00:02:30,730 --> 00:02:33,210
Research are following the academic path
and sharing the results of their work so

130
00:02:33,210 --> 00:02:33,220
and sharing the results of their work so
 

131
00:02:33,220 --> 00:02:34,680
and sharing the results of their work so
that others concerned about its

132
00:02:34,680 --> 00:02:34,690
that others concerned about its
 

133
00:02:34,690 --> 00:02:37,170
that others concerned about its
potential Melissa malicious use and

134
00:02:37,170 --> 00:02:37,180
potential Melissa malicious use and
 

135
00:02:37,180 --> 00:02:39,510
potential Melissa malicious use and
determine ways to flag identify and

136
00:02:39,510 --> 00:02:39,520
determine ways to flag identify and
 

137
00:02:39,520 --> 00:02:43,950
determine ways to flag identify and
protect against any bad actors