Rupal Patel: Synthetic voices, as unique as fingerprints

112,811 views ใƒป 2014-02-13

TED


ืื ื ืœื—ืฅ ืคืขืžื™ื™ื ืขืœ ื”ื›ืชื•ื‘ื™ื•ืช ื‘ืื ื’ืœื™ืช ืœืžื˜ื” ื›ื“ื™ ืœื”ืคืขื™ืœ ืืช ื”ืกืจื˜ื•ืŸ.

ืžืชืจื’ื: Elazar Raab ืžื‘ืงืจ: Ido Dekkers
00:12
I'd like to talk today
0
12719
1490
ื‘ืจืฆื•ื ื™ ืœื“ื‘ืจ ื”ื™ื•ื
00:14
about a powerful and fundamental aspect
1
14209
2927
ืขืœ ื”ื™ื‘ื˜ ื™ืกื•ื“ื™ ื•ืขื•ืฆืžืชื™
00:17
of who we are: our voice.
2
17136
3598
ืฉืœ ืžื™ ืื ื—ื ื•:
ื”ืงื•ืœ ืฉืœื ื•.
00:20
Each one of us has a unique voiceprint
3
20734
2746
ืœื›ืœ ืื—ื“ ืžืื™ืชื ื• ื™ืฉ ื—ืชื™ืžืช ืงื•ืœ ื™ื™ื—ื•ื“ื™ืช
00:23
that reflects our age, our size,
4
23480
2289
ื”ืžืฉืงืคืช ืืช ื’ื™ืœื ื•, ื”ื’ื•ื“ืœ ืฉืœื ื•,
00:25
even our lifestyle and personality.
5
25769
3237
ืืคื™ืœื• ืืช ืกื’ื ื•ืŸ ื—ื™ื™ื ื• ื•ื”ืื™ืฉื™ื•ืช ืฉืœื ื•.
00:29
In the words of the poet Longfellow,
6
29006
2142
ื‘ืžื™ืœื•ืชื™ื• ืฉืœ ื”ืžืฉื•ืจืจ ืœื•ื ื’ืคืœื•,
00:31
"the human voice is the organ of the soul."
7
31148
3870
"ื”ืงื•ืœ ื”ืื ื•ืฉื™ ื”ื•ื ืื™ื‘ืจ ืฉืœ ื”ื ืฉืžื”."
00:35
As a speech scientist, I'm fascinated
8
35018
2747
ื›ืžื“ืขื ื™ืช ื“ื™ื‘ื•ืจ, ืื ื™ ืžื•ืงืกืžืช
00:37
by how the voice is produced,
9
37765
1829
ืžืื™ืš ืฉืžื•ืคืง ื”ืงื•ืœ,
00:39
and I have an idea for how it can be engineered.
10
39594
3658
ื•ื™ืฉ ืœื™ ืจืขื™ื•ืŸ ื›ื™ืฆื“ ื ื™ืชืŸ ืœื”ื ื“ืก ืื•ืชื•.
00:43
That's what I'd like to share with you.
11
43252
2210
ื–ื” ืžื” ืฉื”ื™ื™ืชื™ ืจื•ืฆื” ืœืฉืชืฃ ืื™ืชื›ื.
00:45
I'm going to start by playing you a sample
12
45462
1814
ืื ื™ ืืชื—ื™ืœ ื‘ืœื ื’ืŸ ืœื›ื ื“ื’ื™ืžื”
00:47
of a voice that you may recognize.
13
47276
1871
ืฉืœ ืงื•ืœ ืฉื™ื™ืชื›ืŸ ื•ืชื–ื”ื•.
00:49
(Recording) Stephen Hawking: "I would have thought
14
49147
1304
(ื”ืงืœื˜ื”): "ื”ื™ื™ืชื™ ื—ื•ืฉื‘
00:50
it was fairly obvious what I meant."
15
50451
2749
ืฉื–ื” ื”ื™ื” ื“ื™ ื‘ืจื•ืจ ืœืžื” ื”ืชื›ื•ื•ื ืชื™."
00:53
Rupal Patel: That was the voice
16
53200
1280
ื–ื” ื”ื™ื” ืงื•ืœื•
00:54
of Professor Stephen Hawking.
17
54480
2086
ืฉืœ ืคืจื•ืคืกื•ืจ ืกื˜ื™ื‘ืŸ ื”ื•ืงื™ื ื’.
00:56
What you may not know is that same voice
18
56566
3849
ืžื” ืฉื™ื™ืชื›ืŸ ืฉืื™ื ื›ื ื™ื•ื“ืขื™ื ื–ื” ืฉืื•ืชื• ื”ืงื•ืœ
01:00
may also be used by this little girl
19
60415
2478
ืขืฉื•ื™ ืœื”ื™ื•ืช ื‘ืฉื™ืžื•ืฉ ืฉืœ ื”ื™ืœื“ื” ื”ืงื˜ื ื” ื”ื–ื•
01:02
who is unable to speak
20
62893
1697
ืฉืื™ื ื” ืžืกื•ื’ืœืช ืœื“ื‘ืจ
01:04
because of a neurological condition.
21
64590
2597
ื‘ื’ืœืœ ืžืฆื‘ ื ื•ื™ืจื•ืœื•ื’ื™.
01:07
In fact, all of these individuals
22
67187
2068
ืœืžืขืฉื”, ื›ืœ ื”ืื ืฉื™ื ื”ืœืœื•
01:09
may be using the same voice,
23
69255
2012
ืขืฉื•ื™ื™ื ืœื”ืฉืชืžืฉ ื‘ืื•ืชื• ื”ืงื•ืœ,
01:11
and that's because there's only a few options available.
24
71267
3557
ื•ื–ืืช ืžื›ื™ื•ื•ืŸ ืฉื™ืฉ ืจืง ืžืกืคืจ ืืคืฉืจื•ื™ื•ืช ืžืฆื•ืžืฆื.
01:14
In the U.S. alone, there are 2.5 million Americans
25
74824
4317
ื‘ืืจื”"ื‘ ืœื‘ื“ื” ื™ืฉ 2.5 ืžืœื™ื•ืŸ ืืžืจื™ืงืื™ื
01:19
who are unable to speak,
26
79141
1610
ืฉืื™ื ื ืžืกื•ื’ืœื™ื ืœื“ื‘ืจ,
01:20
and many of whom use computerized devices
27
80751
2622
ื•ืจื‘ื™ื ืžื”ื ืžืฉืชืžืฉื™ื ื‘ืžื›ืฉื™ืจื™ื ืžืžื•ื—ืฉื‘ื™ื
01:23
to communicate.
28
83373
1522
ืขืœ ืžื ืช ืœืชืงืฉืจ.
01:24
Now that's millions of people worldwide
29
84895
3479
ื›ืœื•ืžืจ, ืžื“ื•ื‘ืจ ืขืœ ืžืœื™ื•ื ื™ ืื ืฉื™ื ื‘ืขื•ืœื
01:28
who are using generic voices,
30
88374
1652
ื”ืžืฉืชืžืฉื™ื ื‘ืงื•ืœื•ืช ื’ื ืจื™ื™ื,
01:30
including Professor Hawking,
31
90026
1446
ื›ื•ืœืœ ืคืจื•ืคืกื•ืจ ื”ื•ืงื™ื ื’,
01:31
who uses an American-accented voice.
32
91472
4833
ื”ืžืฉืชืžืฉ ื‘ืงื•ืœ ื‘ืžื‘ื˜ื ืืžืจื™ืงืื™.
01:36
This lack of individuation of the synthetic voice
33
96305
3328
ื—ื•ืกืจ ื”ืื™ื ื“ื™ื‘ื™ื“ื•ืืœื™ื–ื ื‘ืงื•ืœื•ืช ืกื™ื ื˜ืชื™ื™ื
01:39
really hit home
34
99633
1416
ื”ื™ื” ืžืžืฉ ื‘ืจื•ืจ
01:41
when I was at an assistive technology conference
35
101049
2472
ื›ืืฉืจ ื”ื™ื™ืชื™ ื‘ื•ืขื™ื“ื” ืœื˜ื›ื ื•ืœื•ื’ื™ื•ืช ื ื’ื™ืฉื•ืช
01:43
a few years ago,
36
103521
1850
ืœืคื ื™ ืžืกืคืจ ืฉื ื™ื,
01:45
and I recall walking into an exhibit hall
37
105371
3604
ื•ืื ื™ ื–ื•ื›ืจืช ืฉื ื›ื ืกืชื™ ืœืื•ืœื ื”ืชืฆื•ื’ื”
01:48
and seeing a little girl and a grown man
38
108975
3044
ื•ืจืื™ืชื™ ื™ืœื“ื” ืงื˜ื ื” ื•ืื“ื ืžื‘ื•ื’ืจ
01:52
having a conversation using their devices,
39
112019
2916
ืžื ื”ืœื™ื ืฉื™ื—ื” ืชื•ืš ืฉื™ืžื•ืฉ ื‘ืžื›ืฉื™ืจื™ ื”ืขื–ืจ ืฉืœื”ื,
01:54
different devices, but the same voice.
40
114935
4284
ืžื›ืฉื™ืจื™ื ืฉื•ื ื™ื, ืื‘ืœ ืื•ืชื• ื”ืงื•ืœ.
01:59
And I looked around and I saw this happening
41
119219
1909
ื•ื”ื™ื‘ื˜ืชื™ ืžืกื‘ื™ื‘ื™ ื•ืจืื™ืชื™ ืืช ื–ื” ืงื•ืจื”
02:01
all around me, literally hundreds of individuals
42
121128
4190
ืžื›ืœ ื”ื›ื™ื•ื•ื ื™ื, ืœืžืขืฉื” ืžืื•ืช ืื ืฉื™ื
02:05
using a handful of voices,
43
125318
2738
ืžืฉืชืžืฉื™ื ื‘ืงื•ืžืฅ ืฉืœ ืงื•ืœื•ืช,
02:08
voices that didn't fit their bodies
44
128056
3091
ืงื•ืœื•ืช ืฉืœื ื”ืชืื™ืžื• ืœื’ื•ืฃ ืฉืœื”ื
02:11
or their personalities.
45
131147
2082
ืื• ืœืื™ืฉื™ื•ืช ืฉืœื”ื.
02:13
We wouldn't dream of fitting a little girl
46
133229
2727
ืœื ื ืขืœื” ืขืœ ื“ืขืชื ื• ืœื”ืชืื™ื ืœื™ืœื“ื” ืงื˜ื ื”
02:15
with the prosthetic limb of a grown man.
47
135956
3396
ืคืจื•ื˜ื–ื” (ืจื’ืœ ืชื•ืชื‘ืช) ืฉืœ ืื“ื ืžื‘ื•ื’ืจ.
02:19
So why then the same prosthetic voice?
48
139352
3304
ืื– ืžื“ื•ืข ืื ื›ืš ืื•ืชื• "ืชื•ืชื‘ ืงื•ืœ"?
02:22
It really struck me,
49
142656
1291
ื–ื” ื”ืžื ืื•ืชื™,
02:23
and I wanted to do something about this.
50
143947
3151
ื•ืจืฆื™ืชื™ ืœืขืฉื•ืช ืžืฉื”ื• ื‘ื ื•ื’ืข ืœื–ื”.
02:27
I'm going to play you now a sample
51
147098
1953
ืื ื’ืŸ ืœื›ื ืขื›ืฉื™ื• ื“ื’ื™ืžื”
02:29
of someone who has, two people actually,
52
149051
3288
ืฉืœ ืžื™ืฉื”ื•... ืฉืœ ืฉื ื™ ืื ืฉื™ื ืœืžืขืฉื”,
02:32
who have severe speech disorders.
53
152339
1768
ืฉื™ืฉ ืœื”ื ื”ืคืจืขืช ื“ื™ื‘ื•ืจ ื—ืžื•ืจื”.
02:34
I want you to take a listen to how they sound.
54
154107
3230
ืื ื™ ืจื•ืฆื” ืฉืชืงืฉื™ื‘ื• ื›ื™ืฆื“ ื”ื ื ืฉืžืขื™ื.
02:37
They're saying the same utterance.
55
157337
2357
ื”ื ืื•ืžืจื™ื ืืช ืื•ืชื• ื”ืžืฉืคื˜.
02:39
(First voice)
56
159694
2432
(ืงื•ืœ ืจืืฉื•ืŸ)
02:42
(Second voice)
57
162126
3617
(ืงื•ืœ ืฉื ื™)
02:45
You probably didn't understand what they said,
58
165743
2412
ื‘ื˜ื— ืœื ื”ื‘ื ืชื ืžื” ื”ื ืืžืจื•,
02:48
but I hope that you heard
59
168155
1854
ืื‘ืœ ืื ื™ ืžืงื•ื•ื” ืฉืฉืžืขืชื
02:50
their unique vocal identities.
60
170009
4283
ืืช ื”ื–ื”ื•ืช ื”ืงื•ืœื™ืช ื”ื™ื™ื—ื•ื“ื™ืช ืฉืœื”ื.
02:54
So what I wanted to do next is,
61
174292
2813
ืื– ื”ื“ื‘ืจ ื”ื‘ื ืฉืจืฆื™ืชื™ ืœืขืฉื•ืช ื”ื•ื
02:57
I wanted to find out how we could harness
62
177105
2384
ืจืฆื™ืชื™ ืœื‘ืจืจ ื›ื™ืฆื“ ื ื•ื›ืœ ืœืจืชื•ื
02:59
these residual vocal abilities
63
179489
1821
ืืช ืฉืืจื™ืช ื”ื™ื›ื•ืœื•ืช ื”ืงื•ืœื™ื•ืช ื”ืœืœื•
03:01
and build a technology
64
181310
2016
ื•ืœื‘ื ื•ืช ื˜ื›ื ื•ืœื•ื’ื™ื”
03:03
that could be customized for them,
65
183326
2143
ืฉืชื•ื›ืœ ืœื”ื™ื•ืช ืžื•ืชืืžืช ืœื”ื,
03:05
voices that could be customized for them.
66
185469
2429
ืงื•ืœื•ืช ืฉื™ื•ื›ืœื• ืœื”ื™ื•ืช ืžื•ืชืืžื™ื ืœื”ื.
03:07
So I reached out to my collaborator, Tim Bunnell.
67
187898
2685
ืื– ืคื ื™ืชื™ ืœืฉื•ืชืฃ ืฉืœื™, ื˜ื™ื ื‘ื ืœ.
03:10
Dr. Bunnell is an expert in speech synthesis,
68
190583
3063
ื“"ืจ ื‘ื ืœ ื”ื•ื ืžื•ืžื—ื” ื‘ืกื™ื ื˜ื–ื” ืฉืœ ื“ื™ื‘ื•ืจ,
03:13
and what he'd been doing is building
69
193646
2033
ื•ืžื” ืฉื”ื•ื ืขืฉื” ื–ื” ืœื‘ื ื•ืช
03:15
personalized voices for people
70
195679
1881
ืงื•ืœื•ืช ืื™ืฉื™ื™ื ืœืื ืฉื™ื
03:17
by putting together
71
197560
2097
ืขืœ ื™ื“ื™ ืฆื™ืจื•ืฃ
03:19
pre-recorded samples of their voice
72
199657
2150
ื“ื’ื™ืžื•ืช ืžื•ืงืœื˜ื•ืช ืฉืœ ืงื•ืœื
03:21
and reconstructing a voice for them.
73
201807
2879
ื•ืฉื—ื–ื•ืจ ืฉืœ ืงื•ืœ ืขื‘ื•ืจื.
03:24
These are people who had lost their voice
74
204686
1712
ืืœื• ืื ืฉื™ื ืฉืื™ื‘ื“ื• ืืช ืงื•ืœื
03:26
later in life.
75
206398
1911
ืžืื•ื—ืจ ื™ื•ืชืจ ื‘ื—ื™ื™ื”ื.
03:28
We didn't have the luxury
76
208309
1394
ืœื ื• ืœื ื”ื™ืชื” ืืช ื”ืคืจื™ื•ื•ื™ืœื’ื™ื”
03:29
of pre-recorded samples of speech
77
209703
1774
ืฉื‘ื–ืžื™ื ื•ืช ืฉืœ ื“ื’ื™ืžื•ืช ื“ื™ื‘ื•ืจ ืžื•ืงืœื˜ื•ืช
03:31
for those born with speech disorder.
78
211477
2292
ืขื‘ื•ืจ ืืœื• ืฉื ื•ืœื“ื• ืขื ื”ืคืจืขืช ื“ื™ื‘ื•ืจ.
03:33
But I thought, there had to be a way
79
213769
2537
ืื‘ืœ ื—ืฉื‘ืชื™ - ื—ื™ื™ื‘ืช ืœื”ื™ื•ืช ื“ืจืš
03:36
to reverse engineer a voice
80
216306
1944
ืœื”ื ื“ืก ืœืื—ื•ืจ ืงื•ืœ
03:38
from whatever little is left over.
81
218250
2291
ืžื”ืžืขื˜ ืฉื ื•ืชืจ.
03:40
So we decided to do exactly that.
82
220541
2714
ืื– ื”ื—ืœื˜ืชื ื• ืœืขืฉื•ืช ื‘ื“ื™ื•ืง ืืช ื–ื”.
03:43
We set out with a little bit of funding from the National Science Foundation,
83
223255
3403
ื™ืฆืื ื• ืœื“ืจืš ืขื ืžืขื˜ ืžื™ืžื•ืŸ ืžื”ืงืจืŸ ื”ืœืื•ืžื™ืช ืœืžื“ืขื™ื,
03:46
to create custom-crafted voices that captured
84
226658
3565
ืœื™ืฆื•ืจ ืงื•ืœื•ืช ื‘ื”ืชืืžื” ืื™ืฉื™ืช ืฉื™ื›ื™ืœื•
03:50
their unique vocal identities.
85
230223
1536
ืืช ื”ื–ื”ื•ืช ื”ืงื•ืœื™ืช ื”ื™ื™ื—ื•ื“ื™ืช ืฉืœื”ื.
03:51
We call this project VocaliD, or vocal I.D.,
86
231759
3203
ืื ื• ืงื•ืจืื™ื ืœืคืจื•ื™ื™ืงื˜ ื”ื–ื” "ื•ื•ืงืืœื™-ื“ื™" ืื• "ื•ื•ืงืืœ ืื™ื™-ื“ื™"
03:54
for vocal identity.
87
234962
2033
ืขื‘ื•ืจ "ื–ื”ื•ืช ืงื•ืœื™ืช".
03:56
Now before I get into the details of how
88
236995
2674
ื•ื‘ื›ืŸ, ืœืคื ื™ ืฉืืจื“ ืœืคืจื˜ื™ื ื›ื™ืฆื“
03:59
the voice is made and let you listen to it,
89
239669
2048
ื”ืงื•ืœ ื ื•ืฆืจ ื•ืืชืŸ ืœื›ื ืœื”ืงืฉื™ื‘ ืœื•,
04:01
I need to give you a real quick speech science lesson. Okay?
90
241717
3350
ืื ื™ ืฆืจื™ื›ื” ืœื”ืขื‘ื™ืจ ืœื›ื ืฉืขื•ืจ ื–ืจื™ื– ื‘ืžื“ืขื™ ื”ื“ื™ื‘ื•ืจ. ื‘ืกื“ืจ?
04:05
So first, we know that the voice is changing
91
245067
3159
ืจืืฉื™ืช, ืื ื• ื™ื•ื“ืขื™ื ืฉื”ืงื•ืœ ืžืฉืชื ื”
04:08
dramatically over the course of development.
92
248226
2854
ื‘ืื•ืคืŸ ื“ืจืžื˜ื™ ืœืื•ืจืš ืชื”ืœื™ืš ื”ื”ืชืคืชื—ื•ืช.
04:11
Children sound different from teens
93
251080
2090
ื™ืœื“ื™ื ื ืฉืžืขื™ื ืฉื•ื ื” ืžื‘ื ื™ ืขืฉืจื”
04:13
who sound different from adults.
94
253170
1463
ืฉื ืฉืžืขื™ื ืฉื•ื ื” ืžืžื‘ื•ื’ืจื™ื.
04:14
We've all experienced this.
95
254633
2642
ื›ื•ืœื ื• ื”ืชื ืกื ื• ื‘ื–ื”.
04:17
Fact number two is that speech
96
257275
3363
ืขื•ื‘ื“ื” ืžืกืคืจ ืฉืชื™ื™ื ื”ื™ื ืฉื“ื™ื‘ื•ืจ
04:20
is a combination of the source,
97
260638
2553
ื”ื™ื ืฉื™ืœื•ื‘ ืฉืœ ื”ืžืงื•ืจ,
04:23
which is the vibrations generated by your voice box,
98
263191
3479
ืฉื”ื•ื ื”ืชื ื•ื“ื•ืช ืฉืžื•ืคืงื•ืช ืขืœ ื™ื“ื™ ืชื™ื‘ืช ื”ืงื•ืœ ืฉืœื›ื,
04:26
which are then pushed through
99
266670
1939
ืฉืœืื—ืจ ืžื›ืŸ ื ื“ื—ืคื•ืช ื“ืจืš
04:28
the rest of the vocal tract.
100
268609
2437
ืฉืืจ ืžืขืจื›ืช ื”ืงื•ืœ.
04:31
These are the chambers of your head and neck
101
271046
2484
ืืœื• ื”ื—ืœืœื™ื ืฉืœ ื”ืจืืฉ ื•ื”ืฆื•ื•ืืจ
04:33
that vibrate,
102
273530
1239
ืฉืจื•ืขื“ื™ื,
04:34
and they actually filter that source sound
103
274769
2110
ื•ื”ื ืœืžืขืฉื” ืžืกื ื ื™ื ืืช ืžืงื•ืจ ื”ืฆืœื™ืœ
04:36
to produce consonants and vowels.
104
276879
2537
ืขืœ ืžื ืช ืœื”ืคื™ืง ืขื™ืฆื•ืจื™ื ื•ืชื ื•ืขื•ืช.
04:39
So the combination of source and filter
105
279416
3860
ืื– ื”ืฉื™ืœื•ื‘ ืฉืœ ืžืงื•ืจ ื•ืžืกื ืŸ
04:43
is how we produce speech.
106
283276
2630
ื–ื” ื”ื“ืจืš ื‘ื” ืื ื• ืžืคื™ืงื™ื ื“ื™ื‘ื•ืจ.
04:45
And that happens in one individual.
107
285906
3026
ื•ื–ื” ืงื•ืจื” ืืฆืœ ื›ืœ ืื—ื“.
04:48
Now I told you earlier that I'd spent
108
288932
2626
ื•ื‘ื›ืŸ ืืžืจืชื™ ืœื›ื ืงื•ื“ื ืœื›ืŸ ืฉื”ืฉืงืขืชื™
04:51
a good part of my career
109
291558
2025
ื—ืœืง ื ื™ื›ืจ ืžื”ืงืจื™ื™ืจื” ืฉืœื™
04:53
understanding and studying
110
293583
2453
ื‘ื”ื‘ื ื” ื•ืœื™ืžื•ื“
04:56
the source characteristics of people
111
296036
1958
ืฉืœ ืžืืคื™ื™ื ื™ ื”ืžืงื•ืจ ืฉืœ ืื ืฉื™ื
04:57
with severe speech disorder,
112
297994
2301
ืขื ื”ืคืจืขื•ืช ื“ื™ื‘ื•ืจ ื—ืžื•ืจื•ืช,
05:00
and what I've found
113
300295
1465
ื•ืžื” ืฉื’ื™ืœื™ืชื™
05:01
is that even though their filters were impaired,
114
301760
3366
ืฉืœืžืจื•ืช ืฉื”ืžืกื ื ื™ื ืฉืœื”ื ื”ื™ื• ืคื’ื•ืžื™ื,
05:05
they were able to modulate their source:
115
305126
2961
ื”ื ื”ืฆืœื™ื—ื• ืœืืคื ืŸ ืืช ื”ืžืงื•ืจ ืฉืœื”ื:
05:08
the pitch, the loudness, the tempo of their voice.
116
308087
3262
ืืช ื’ื•ื‘ื” ื”ืฆืœื™ืœ, ื”ืขื•ืฆืžื” ื•ื”ืงืฆื‘ ืฉืœ ื”ืงื•ืœ ืฉืœื”ื.
05:11
These are called prosody, and I've been documenting for years
117
311349
3368
ืžืจื›ื™ื‘ื™ื ืืœื• ื ืงืจืื™ื ืคืจื•ื–ื•ื“ื™ื”, ื•ืชื™ืขื“ืชื™ ื‘ืžืฉืš ืฉื ื™ื
05:14
that the prosodic abilities of these individuals
118
314717
2277
ืฉื”ื™ื›ื•ืœื•ืช ื”ืคืจื•ื–ื•ื“ื™ื•ืช ืฉืœ ืื ืฉื™ื ืืœื•
05:16
are preserved.
119
316994
1575
ืžืฉืชืžืจื•ืช.
05:18
So when I realized that those same cues
120
318569
4087
ืื– ื›ืืฉืจ ื”ื‘ื ืชื™ ืฉืื•ืชื ื”ืกื™ืžื ื™ื ื”ืœืœื•
05:22
are also important for speaker identity,
121
322656
2769
ื”ื ื’ื ื—ืฉื•ื‘ื™ื ืœื–ื”ื•ืช ื”ื“ื•ื‘ืจ,
05:25
I had this idea.
122
325425
2015
ื”ื™ื” ืœื™ ืจืขื™ื•ืŸ ื›ื–ื”.
05:27
Why don't we take the source
123
327440
2516
ืžื“ื•ืข ืฉืœื ื ื™ืงื— ืืช ื”ืžืงื•ืจ
05:29
from the person we want the voice to sound like,
124
329956
2213
ืžื”ืื“ื ืฉื ืจืฆื” ืฉื”ืงื•ืœ ื™ื™ืฉืžืข ื›ืžื•ืชื•,
05:32
because it's preserved,
125
332169
1463
ืžื›ื™ื•ื•ืŸ ืฉื”ื•ื ืžืฉืชืžืจ,
05:33
and borrow the filter
126
333632
2135
ื•ื ืฉืื™ืœ ืืช ื”ืžืกื ืŸ
05:35
from someone about the same age and size,
127
335767
3229
ืžืžื™ืฉื”ื• ื‘ืขืจืš ื‘ืื•ืชื• ื’ื™ืœ ื•ื’ื•ื“ืœ,
05:39
because they can articulate speech,
128
339011
2407
ืžื›ื™ื•ื•ืŸ ืฉื”ื ื™ื›ื•ืœื™ื ืœื”ื’ื•ืช ื“ื™ื‘ื•ืจ
05:41
and then mix them?
129
341418
1791
ื•ืื– ื ืขืจื‘ื‘ ืื•ืชื?
05:43
Because when we mix them,
130
343209
1787
ื›ื™ ื›ืืฉืจ ืื ื—ื ื• ืžืขืจื‘ื‘ื™ื ืื•ืชื,
05:44
we can get a voice that's as clear
131
344996
1698
ืื ื• ื™ื›ื•ืœื™ื ืœืงื‘ืœ ืงื•ืœ ืฉื”ื•ื ื‘ืจื•ืจ
05:46
as our surrogate talker --
132
346694
1754
ื›ืžื• ื”ื“ื•ื‘ืจ ื”ื—ืœื™ืคื™ ืฉืœื ื• --
05:48
that's the person we borrowed the filter fromโ€”
133
348448
2595
ื–ื” ื”ืื“ื ืฉื”ืฉืืœื ื• ืžืžื ื• ืืช ื”ืžืกื ืŸ --
05:51
and is similar in identity to our target talker.
134
351043
4649
ื•ื”ื•ื ื“ื•ืžื” ื‘ื–ื”ื•ืช ืœื“ื•ื‘ืจ ื”ื™ืขื“ ืฉืœื ื•.
05:55
It's that simple.
135
355692
1427
ื–ื” ืžืžืฉ ืคืฉื•ื˜.
05:57
That's the science behind what we're doing.
136
357119
2934
ื–ื” ื”ืžื“ืข ื”ืขื•ืžื“ ืžืื—ื•ืจื™ ืžื” ืฉืื ื—ื ื• ืขื•ืฉื™ื.
06:00
So once you have that in mind,
137
360053
3533
ืื– ื›ืฉืืชื ืžื‘ื™ื ื™ื ืืช ื–ื”,
06:03
how do you go about building this voice?
138
363586
2258
ืื™ืš ืžืชืงื“ืžื™ื ื‘ื‘ื ื™ื™ืช ืงื•ืœื•ืช?
06:05
Well, you have to find someone
139
365844
1480
ื•ื‘ื›ืŸ, ืขืœื™ื›ื ืœืžืฆื•ื ืžื™ืฉื”ื•
06:07
who is willing to be a surrogate.
140
367324
2400
ืฉืžื•ื›ืŸ ืœื”ื™ื•ืช ื“ื•ื‘ืจ ื—ืœื™ืคื™.
06:09
It's not such an ominous thing.
141
369724
2264
ื–ื” ืœื ืกื™ืคื•ืจ.
06:11
Being a surrogate donor
142
371988
1523
ืœื”ื™ื•ืช ืชื•ืจื ื—ืœื™ืคื™
06:13
only requires you to say a few hundred
143
373511
2788
ื“ื•ืจืฉ ืžื›ื ืœื•ืžืจ ืจืง ืžืื•ืช ืกืคื•ืจื•ืช
06:16
to a few thousand utterances.
144
376299
2242
ืขื“ ืืœืคื™ื ืกืคื•ืจื™ื ืฉืœ ื‘ื™ื˜ื•ื™ื™ื.
06:18
The process goes something like this.
145
378541
2003
ื”ืชื”ืœื™ืš ืงื•ืจื” ื‘ืื•ืคืŸ ื›ืžื• ื–ื”:
06:20
(Video) Voice: Things happen in pairs.
146
380544
2190
"ื“ื‘ืจื™ื ืงื•ืจื™ื ื‘ืฆืžื“ื™ื"
06:22
I love to sleep.
147
382734
1925
"ืื ื™ ืื•ื”ื‘ ืœื™ืฉื•ืŸ"
06:24
The sky is blue without clouds.
148
384659
3882
"ื”ืฉืžื™ื™ื ื›ื—ื•ืœื™ื ืœืœื ืขื ื ื™ื"
06:28
RP: Now she's going to go on like this
149
388541
2002
ื•ื‘ื›ืŸ, ื”ื™ื ืขื•ืžื“ืช ืœื”ืžืฉื™ืš ื›ืš
06:30
for about three to four hours,
150
390543
1919
ื‘ืžืฉืš ืฉืœื•ืฉ ืขื“ ืืจื‘ืข ืฉืขื•ืช
06:32
and the idea is not for her to say everything
151
392462
3005
ื•ื”ืจืขื™ื•ืŸ ืื™ื ื• ืฉืชืืžืจ ื›ืœ ื“ื‘ืจ
06:35
that the target is going to want to say,
152
395467
2045
ืฉื”ื™ืขื“ ื™ืจืฆื” ืœื•ืžืจ,
06:37
but the idea is to cover all the different combinations
153
397512
3395
ืืœื ื”ืจืขื™ื•ืŸ ืœื›ืกื•ืช ืืช ื›ืœ ื”ืฆื™ืจื•ืคื™ื ื”ืฉื•ื ื™ื
06:40
of the sounds that occur in the language.
154
400907
3271
ืฉืœ ื”ืฆืœื™ืœื™ื ืฉืงื™ื™ืžื™ื ื‘ืฉืคื”.
06:44
The more speech you have,
155
404178
1638
ื›ื›ืœ ืฉื™ืฉ ืœืš ื™ื•ืชืจ ื“ื™ื‘ื•ืจ,
06:45
the better sounding voice you're going to have.
156
405816
2305
ื›ืš ืชืฉืชืคืจ ืื™ื›ื•ืช ื”ืงื•ืœ ืฉืชืงื‘ืœ.
06:48
Once you have those recordings,
157
408121
1673
ื›ืืฉืจ ื™ืฉ ืœืš ืืช ื”ื”ืงืœื˜ื•ืช ื”ืœืœื•,
06:49
what we need to do
158
409794
1413
ืžื” ืฉืื ื• ืฆืจื™ื›ื™ื ืœืขืฉื•ืช
06:51
is we have to parse these recordings
159
411207
2718
ื”ื•ื ืฉืขืœื™ื ื• ืœื ืชื— ื•ืœืคืจืง ืืช ื”ื”ืงืœื˜ื•ืช ื”ืœืœื•
06:53
into little snippets of speech,
160
413925
2449
ืœืงื˜ืขื™ื ืงื˜ื ื™ื ืฉืœ ื“ื™ื‘ื•ืจ,
06:56
one- or two-sound combinations,
161
416374
2337
ืฆื™ืจื•ืฃ ืฉืœ ืฆืœื™ืœ ืื—ื“ ืื• ืฉื ื™ื™ื,
06:58
sometimes even whole words
162
418711
1883
ืœืคืขืžื™ื ื’ื ืžื™ืœื™ื ืฉืœืžื•ืช
07:00
that start populating a dataset or a database.
163
420594
4516
ืฉืžืชื—ื™ืœื™ื ืœืžืœื ื‘ืกื™ืก ื ืชื•ื ื™ื.
07:05
We're going to call this database a voice bank.
164
425110
3717
ื ืงืจื ืœื‘ืกื™ืก ื”ื ืชื•ื ื™ื ื”ื–ื” ื‘ื ืง ืงื•ืœ.
07:08
Now the power of the voice bank
165
428827
2096
ื•ื‘ื›ืŸ, ื”ืขื•ืฆืžื” ืฉืœ ื‘ื ืง ื”ืงื•ืœ
07:10
is that from this voice bank,
166
430923
2014
ื”ื™ื ืฉืžื‘ื ืง ืงื•ืœ ื–ื”,
07:12
we can now say any new utterance,
167
432937
2011
ืื ื• ื™ื›ื•ืœื™ื ื›ืขืช ืœื•ืžืจ ื›ืœ ื‘ื™ื˜ื•ื™ ื—ื“ืฉ,
07:14
like, "I love chocolate" --
168
434948
1424
ื›ืžื• "ืื ื™ ืื•ื”ื‘ ืฉื•ืงื•ืœื“" --
07:16
everyone needs to be able to say thatโ€”
169
436372
1739
ื›ื•ืœื ืฆืจื™ื›ื™ื ืœื”ื™ื•ืช ืžืกื•ื’ืœื™ื ืœื•ืžืจ ืืช ื–ื” --
07:18
fish through that database
170
438111
1831
ื—ืคื•ืจ ื‘ื‘ืกื™ืก ื”ื ืชื•ื ื™ื ื”ื–ื”
07:19
and find all the segments necessary
171
439942
1940
ื•ืชืžืฆื ืืช ื›ืœ ื”ืžืงื˜ืขื™ื ื”ื“ืจื•ืฉื™ื
07:21
to say that utterance.
172
441882
1929
ืœื•ืžืจ ืืช ื”ื‘ื™ื˜ื•ื™ ื”ื–ื”.
07:23
(Video) Voice: I love chocolate.
173
443811
1789
"ืื ื™ ืื•ื”ื‘ ืฉื•ืงื•ืœื“"
07:25
RP: So that's speech synthesis.
174
445600
1391
ืื– ื–ื” ืกื™ื ื˜ื–ื” ืฉืœ ื“ื™ื‘ื•ืจ.
07:26
It's called concatenative synthesis, and that's what we're using.
175
446991
2573
ื”ื™ื ื ืงืจืืช ืกื™ื ื˜ื–ื” ืžืฉืจืฉืจืช, ื•ื‘ื–ื” ืื ื• ืžืฉืชืžืฉื™ื.
07:29
That's not the novel part.
176
449564
1533
ื”ื—ื“ืฉื ื•ืช ืื™ื ื” ื‘ื–ื”.
07:31
What's novel is how we make it sound
177
451097
2221
ืžื” ืฉื—ื“ืฉื ื™ ื”ื•ื ืื™ืš ืื ื• ื’ื•ืจืžื™ื ืœื–ื” ืœื”ืฉืžืข
07:33
like this young woman.
178
453318
1457
ื›ืžื• ื”ืื™ืฉื” ื”ืฆืขื™ืจื” ื”ื–ื•.
07:34
This is Samantha.
179
454775
1524
ื–ืืช ืกืžื ื˜ื”.
07:36
I met her when she was nine,
180
456299
2346
ืคื’ืฉืชื™ ืื•ืชื” ื›ืฉื”ื™ืชื” ื‘ืช ืชืฉืข,
07:38
and since then, my team and I
181
458645
1897
ื•ืžืื– ื”ืฆื•ื•ืช ืฉืœื™ ื•ืื ื™
07:40
have been trying to build her a personalized voice.
182
460542
2714
ืžื ืกื™ื ืœื‘ื ื•ืช ืœื” ืงื•ืœ ืื™ืฉื™.
07:43
We first had to find a surrogate donor,
183
463256
3099
ืจืืฉื™ืช ื”ื™ื” ืขืœื™ื ื• ืœืžืฆื•ื ืชื•ืจื ืงื•ืœ ื—ืœื™ืคื™,
07:46
and then we had to have Samantha
184
466355
1818
ื•ืื– ื”ื™ื” ืขืœื™ื ื• ืœื’ืจื•ื ืœืกืžื ื˜ื”
07:48
produce some utterances.
185
468173
1929
ืœื”ืคื™ืง ืžืกืคืจ ื‘ื™ื˜ื•ื™ื™ื.
07:50
What she can produce are mostly vowel-like sounds,
186
470102
2379
ืžื” ืฉื”ื™ื ืžืกื•ื’ืœืช ืœื”ืคื™ืง ื–ื” ื‘ืขื™ืงืจ ืฆืœื™ืœื™ื ื“ืžื•ื™ื™ ืชื ื•ืขื•ืช
07:52
but that's enough for us to extract
187
472481
2479
ืื‘ืœ ื–ื” ืžืกืคื™ืง ืœื ื• ืขืœ ืžื ืช ืœืžืฆื•ืช
07:54
her source characteristics.
188
474960
2285
ืืช ืžืืคื™ื™ื ื™ ืžืงื•ืจ ื”ืงื•ืœ ืฉืœื”.
07:57
What happens next is best described
189
477245
3271
ื”ื“ื‘ืจ ื”ื‘ื ืฉืงื•ืจื” ืžืชื•ืืจ ื”ื›ื™ ื˜ื•ื‘
08:00
by my daughter's analogy. She's six.
190
480516
2767
ืขืœ ื™ื“ื™ ืื ืœื•ื’ื™ื” ืฉืœ ื”ื‘ืช ืฉืœื™. ื”ื™ื ื‘ืช ืฉืฉ.
08:03
She calls it mixing colors to paint voices.
191
483283
5422
ื”ื™ื ืงื•ืจืืช ืœื–ื” ืœืขืจื‘ื‘ ืฆื‘ืขื™ื ื›ื“ื™ ืœืฆื‘ื•ืข ืงื•ืœื•ืช.
08:08
It's beautiful. It's exactly that.
192
488705
2555
ื–ื” ื™ืคื”ืคื”. ื–ื” ื‘ื“ื™ื•ืง ื–ื”.
08:11
Samantha's voice is like a concentrated sample
193
491260
2860
ื”ืงื•ืœ ืฉืœ ืกืžื ื˜ื” ื”ื•ื ื›ืžื• ื“ื’ื™ืžื” ืžืจื•ื›ื–ืช
08:14
of red food dye which we can infuse
194
494120
2609
ืฉืœ ืฆื‘ืข ืžืื›ืœ ืื“ื•ื ืฉืื ื• ื™ื›ื•ืœื™ื ืœื”ื–ืจื™ืง
08:16
into the recordings of her surrogate
195
496729
2540
ืœืชื•ืš ื”ื”ืงืœื˜ื•ืช ืฉืœ ื”ืงื•ืœ ื”ื—ืœื™ืคื™ ืฉืœื”
08:19
to get a pink voice just like this.
196
499269
4387
ืขืœ ืžื ืช ืœืงื‘ืœ ืงื•ืœ ื•ืจื•ื“, ื‘ื“ื™ื•ืง ื›ื›ื”:
08:23
(Video) Samantha: Aaaaaah.
197
503656
4491
ืกืžื ื˜ื”: "ืืืืืืื”."
08:28
RP: So now, Samantha can say this.
198
508147
2808
ืื– ืขื›ืฉื™ื• ืกืžื ื˜ื” ื™ื›ื•ืœื” ืœื•ืžืจ ืืช ื–ื”.
08:30
(Video) Samantha: This voice is only for me.
199
510955
3069
ืกืžื ื˜ื”: "ื”ืงื•ืœ ื”ื–ื” ื”ื•ื ืจืง ื‘ืฉื‘ื™ืœื™.
08:34
I can't wait to use my new voice with my friends.
200
514024
6305
ืื ื™ ืœื ื™ื›ื•ืœื” ืœื—ื›ื•ืช ืœื”ืฉืชืžืฉ ื‘ืงื•ืœ ื”ื—ื“ืฉ ืฉืœื™ ืขื ื”ื—ื‘ืจื™ื ืฉืœื™."
08:40
RP: Thank you. (Applause)
201
520329
6417
ืชื•ื“ื” ืจื‘ื”. (ืžื—ื™ืื•ืช ื›ืคื™ื™ื)
08:46
I'll never forget the gentle smile
202
526746
2333
ืœืขื•ืœื ืœื ืืฉื›ื— ืืช ื”ื—ื™ื•ืš ื”ืขื“ื™ืŸ
08:49
that spread across her face
203
529079
1902
ืฉื ืคืจืฉ ืขืœ ืคื ื™ื”
08:50
when she heard that voice for the first time.
204
530981
3649
ื›ืืฉืจ ืฉืžืขื” ืืช ื”ืงื•ืœ ื”ื–ื” ื‘ืคืขื ื”ืจืืฉื•ื ื”.
08:54
Now there's millions of people
205
534630
1882
ื•ื‘ื›ืŸ, ื™ืฉ ืžืœื™ื•ื ื™ ืื ืฉื™ื
08:56
around the world like Samantha, millions,
206
536512
2833
ื‘ื›ืœ ื”ืขื•ืœื ื›ืžื• ืกืžื ื˜ื”. ืžืœื™ื•ื ื™ื.
08:59
and we've only begun to scratch the surface.
207
539345
3440
ื•ืจืง ื”ืชื—ืœื ื• ืœื’ืœื•ืช ืืช ืงืฆื” ื”ืงืจื—ื•ืŸ.
09:02
What we've done so far is we have
208
542785
1642
ืžื” ืฉืขืฉื™ื ื• ืขื“ ื›ื” ื–ื” ืฉื™ืฉ ืœื ื•
09:04
a few surrogate talkers from around the U.S.
209
544427
3859
ื›ืžื” ื“ื•ื‘ืจื™ื ื—ืœื™ืคื™ื™ื ืžืจื—ื‘ื™ ืืจื”"ื‘
09:08
who have donated their voices,
210
548286
1507
ืฉืชืจืžื• ืืช ืงื•ืœื,
09:09
and we have been using those
211
549793
1928
ื•ื”ืฉืชืžืฉื ื• ื‘ืืœื”
09:11
to build our first few personalized voices.
212
551721
4472
ืœื‘ื ื•ืช ืืช ืžืกืคืจ ื”ืงื•ืœื•ืช ื”ืžื•ืชืืžื™ื ืื™ืฉื™ืช ื”ืจืืฉื•ื ื™ื ืฉืœื ื•
09:16
But there's so much more work to be done.
213
556193
1756
ืื‘ืœ ื™ืฉ ืขื•ื“ ื›ืœ ื›ืš ื”ืจื‘ื” ืขื‘ื•ื“ื” ืœืขืฉื•ืช.
09:17
For Samantha, her surrogate
214
557949
2188
ืขื‘ื•ืจ ืกืžื ื˜ื”, ื”ื“ื•ื‘ืจ ื”ื—ืœื™ืคื™ ืฉืœื”
09:20
came from somewhere in the Midwest, a stranger
215
560137
3046
ื”ื’ื™ืข ืžืื™ืคืฉื”ื• ื‘ืžืขืจื‘ ื”ืžืจื›ื–ื™ (ืฉืœ ืืจื”"ื‘) - ื–ืจ
09:23
who gave her the gift of voice.
216
563183
3841
ืฉื ืชืŸ ืœื” ืืช ืžืชื ืช ื”ืงื•ืœ.
09:27
And as a scientist, I'm so excited
217
567024
2153
ื•ื›ืžื“ืขื ื™ืช ืื ื™ ื›ืœ ื›ืš ื ืจื’ืฉืช
09:29
to take this work out of the laboratory
218
569177
1935
ืœืงื—ืช ืืช ื”ืขื‘ื•ื“ื” ื”ื–ื• ืืœ ืžื—ื•ืฅ ืœืžืขื‘ื“ื”
09:31
and finally into the real world
219
571112
1800
ื•ืกื•ืฃ ืกื•ืฃ ืืœ ื”ืขื•ืœื ื”ืืžื™ืชื™
09:32
so it can have real-world impact.
220
572912
3165
ืขืœ ืžื ืช ืฉืชื”ื™ื” ืœื” ื”ืฉืคืขื” ืืžื™ืชื™ืช.
09:36
What I want to share with you next
221
576077
1582
ื”ื“ื‘ืจ ื”ื‘ื ืฉืื ื™ ืจื•ืฆื” ืœื—ืœื•ืง ืื™ืชื›ื
09:37
is how I envision taking this work
222
577659
2175
ื–ื” ืื™ืš ืื ื™ ื—ื•ื–ื” ืฉื”ืขื‘ื•ื“ื” ื”ื–ื•
09:39
to that next level.
223
579834
2711
ืชืขื‘ื•ืจ ืœืฉืœื‘ ื”ื‘ื.
09:42
I imagine a whole world of surrogate donors
224
582545
3887
ืื ื™ ืžื“ืžื™ื™ื ืช ืขื•ืœื ืฉืœื ืฉืœ ืชื•ืจืžื™ ืงื•ืœ ื—ืœื™ืคื™
09:46
from all walks of life, different sizes, different ages,
225
586432
3260
ืžื›ืœ ืชื—ื•ืžื™ ื”ื—ื™ื™ื, ื‘ื’ื“ืœื™ื ืฉื•ื ื™ื, ื’ื™ืœืื™ื ืฉื•ื ื™ื,
09:49
coming together in this voice drive
226
589692
3058
ื‘ืื™ื ื™ื—ื“ื™ื• ืœืžื”ืœืš ื”ืงื•ืœื™ ื”ื–ื”
09:52
to give people voices
227
592750
2270
ืœืชืช ืœืื ืฉื™ื ืงื•ืœื•ืช
09:55
that are as colorful as their personalities.
228
595020
3799
ืฉื”ื ืžื’ื•ื•ื ื™ื ื›ืžื• ื”ืื™ืฉื™ื•ืช ื”ืฉื•ื ื” ืฉืœ ื›ืœ ืื—ื“ ืžื”ื.
09:58
To do that as a first step,
229
598819
2300
ืขืœ ืžื ืช ืœืขืฉื•ืช ืืช ื–ื”, ื›ืฆืขื“ ืจืืฉื•ืŸ
10:01
we've put together this website, VocaliD.org,
230
601119
3275
ื”ืงืžื ื• ืืช ื”ืืชืจ ื”ื–ื”: VocaliD.org
10:04
as a way to bring together those
231
604394
1624
ื›ื“ืจืš ืœื›ื ืก ืืช ื›ืœ ืืœื”
10:06
who want to join us as voice donors,
232
606018
2675
ืฉืจื•ืฆื™ื ืœื”ืฆื˜ืจืฃ ืืœื™ื ื• ื›ืชื•ืจืžื™ ืงื•ืœ,
10:08
as expertise donors,
233
608693
1772
ื›ืชื•ืจืžื™ ืžื•ืžื—ื™ื•ืช,
10:10
in whatever way to make this vision a reality.
234
610465
5339
ื‘ื›ืœ ื“ืจืš ืฉืชืืคืฉืจ ืœื”ืคื•ืš ืืช ื”ื—ื–ื•ืŸ ื”ื–ื” ืœืžืฆื™ืื•ืช.
10:15
They say that giving blood can save lives.
235
615804
4153
ืื•ืžืจื™ื ืฉืชืจื•ืžืช ื“ื ืžืฆื™ืœื” ื—ื™ื™ื.
10:19
Well, giving your voice can change lives.
236
619957
4982
ื•ื‘ื›ืŸ, ืœืชืช ืืช ื”ืงื•ืœ ืฉืœืš ื™ื›ื•ืœ ืœืฉื ื•ืช ื—ื™ื™ื.
10:24
All we need is a few hours of speech
237
624939
3050
ื›ืœ ืžื” ืฉืื ื• ืฆืจื™ื›ื™ื ื–ื” ืžืกืคืจ ืฉืขื•ืช ืฉืœ ื“ื™ื‘ื•ืจ
10:27
from our surrogate talker,
238
627989
1491
ืžืชื•ืจื ื”ืงื•ืœ ื”ื—ืœื™ืคื™ ืฉืœื ื•,
10:29
and as little as a vowel from our target talker,
239
629480
4733
ื•ืจืง ืฆืœื™ืœ ืชื ื•ืขื” ืžื“ื•ื‘ืจ ื”ื™ืขื“,
10:34
to create a unique vocal identity.
240
634213
3711
ืขืœ ืžื ืช ืœื™ืฆื•ืจ ื–ื”ื•ืช ืงื•ืœื™ืช ื™ื™ื—ื•ื“ื™ืช.
10:37
So that's the science behind what we're doing.
241
637924
2626
ืื– ื–ื” ื”ืžื“ืข ืฉืžืื—ื•ืจื™ ืžื” ืฉืื ื• ืขื•ืฉื™ื.
10:40
I want to end by circling back to the human side
242
640550
4455
ืื ื™ ืจื•ืฆื” ืœืกื™ื™ื ื‘ื—ื–ืจื” ืœืกื™ืคื•ืจ ื”ืื ื•ืฉื™
10:45
that is really the inspiration for this work.
243
645005
4102
ืฉื”ื•ื ื‘ืขืฆื ื”ื”ืฉืจืื” ืœืขื‘ื•ื“ื” ื”ื–ื•.
10:49
About five years ago, we built our very first voice
244
649107
3699
ืœืคื ื™ ื›ื—ืžืฉ ืฉื ื™ื ื‘ื ื™ื ื• ืืช ื”ืงื•ืœ ื”ืจืืฉื•ืŸ ื‘ื”ื—ืœื˜ ืฉืœื ื•
10:52
for a little boy named William.
245
652806
2501
ืœื ืขืจ ืฆืขื™ืจ ื‘ืฉื ื•ื•ื™ืœื™ืื.
10:55
When his mom first heard this voice,
246
655307
2357
ื›ืฉืื™ืžื• ืฉืžืขื” ืืช ื”ืงื•ืœ ืฉืœื• ื‘ืคืขื ื”ืจืืฉื•ื ื”,
10:57
she said, "This is what William
247
657664
2345
ื”ื™ื ืืžืจื”: "ื–ื” ืื™ืš ืฉื•ื•ื™ืœื™ืื
11:00
would have sounded like
248
660009
1546
ื”ื™ื” ื ืฉืžืข
11:01
had he been able to speak."
249
661555
2449
ืื ื”ื•ื ื”ื™ื” ื™ื›ื•ืœ ืœื“ื‘ืจ."
11:04
And then I saw William typing a message
250
664004
2418
ื•ืื– ืจืื™ืชื™ ืืช ื•ื•ื™ืœื™ืื ืžืงืœื™ื“ ืžืกืจ
11:06
on his device.
251
666422
1362
ื‘ืžื›ืฉื™ืจ ืฉืœื•.
11:07
I wondered, what was he thinking?
252
667784
3293
ืชื”ื™ืชื™, ืžื” ื”ื•ื ื—ื•ืฉื‘?
11:11
Imagine carrying around someone else's voice
253
671077
3590
ื“ืžื™ื™ื ื• ืฉืืชื ื ื•ืฉืื™ื ืงื•ืœ ืฉืœ ืžื™ืฉื”ื• ืื—ืจ
11:14
for nine years
254
674667
2193
ื‘ืžืฉืš ืชืฉืข ืฉื ื™ื
11:16
and finally finding your own voice.
255
676860
4844
ื•ืกื•ืฃ ืกื•ืฃ ืžื•ืฆืื™ื ืืช ื”ืงื•ืœ ืฉืœื›ื.
11:21
Imagine that.
256
681704
1377
ื“ืžื™ื™ื ื• ืืช ื–ื”.
11:23
This is what William said:
257
683081
2797
ื–ื” ืžื” ืฉื•ื•ื™ืœื™ืื ืืžืจ:
11:25
"Never heard me before."
258
685878
4463
"ืืฃ ืคืขื ืœื ืฉืžืขืชื™ ืืช ืขืฆืžื™ ืœืคื ื™ ื›ืŸ."
11:32
Thank you.
259
692417
1619
ืชื•ื“ื”.
11:34
(Applause)
260
694036
4724
(ืžื—ื™ืื•ืช ื›ืคื™ื™ื)
ืขืœ ืืชืจ ื–ื”

ืืชืจ ื–ื” ื™ืฆื™ื’ ื‘ืคื ื™ื›ื ืกืจื˜ื•ื ื™ YouTube ื”ืžื•ืขื™ืœื™ื ืœืœื™ืžื•ื“ ืื ื’ืœื™ืช. ืชื•ื›ืœื• ืœืจืื•ืช ืฉื™ืขื•ืจื™ ืื ื’ืœื™ืช ื”ืžื•ืขื‘ืจื™ื ืขืœ ื™ื“ื™ ืžื•ืจื™ื ืžื”ืฉื•ืจื” ื”ืจืืฉื•ื ื” ืžืจื—ื‘ื™ ื”ืขื•ืœื. ืœื—ืฅ ืคืขืžื™ื™ื ืขืœ ื”ื›ืชื•ื‘ื™ื•ืช ื‘ืื ื’ืœื™ืช ื”ืžื•ืฆื’ื•ืช ื‘ื›ืœ ื“ืฃ ื•ื™ื“ืื• ื›ื“ื™ ืœื”ืคืขื™ืœ ืืช ื”ืกืจื˜ื•ืŸ ืžืฉื. ื”ื›ืชื•ื‘ื™ื•ืช ื’ื•ืœืœื•ืช ื‘ืกื ื›ืจื•ืŸ ืขื ื”ืคืขืœืช ื”ื•ื•ื™ื“ืื•. ืื ื™ืฉ ืœืš ื”ืขืจื•ืช ืื• ื‘ืงืฉื•ืช, ืื ื ืฆื•ืจ ืื™ืชื ื• ืงืฉืจ ื‘ืืžืฆืขื•ืช ื˜ื•ืคืก ื™ืฆื™ืจืช ืงืฉืจ ื–ื”.

https://forms.gle/WvT1wiN1qDtmnspy7