Would you like to inspect the original subtitles? These are the user uploaded subtitles that are being translated:
1
00:00:00,170 --> 00:00:03,890
Now is the time for our Faceswap videos.
2
00:00:03,890 --> 00:00:07,070
I hope you have found some sauce videos.
3
00:00:07,070 --> 00:00:10,760
I have simply downloaded two videos from Pexels.
4
00:00:10,760 --> 00:00:12,350
I think this is enough.
5
00:00:12,350 --> 00:00:14,780
We don't have to post videos every day.
6
00:00:14,780 --> 00:00:18,890
So I think like one video every every week is okay.
7
00:00:18,890 --> 00:00:20,150
So we start with two.
8
00:00:20,180 --> 00:00:22,220
We have pictures for two weeks.
9
00:00:22,220 --> 00:00:24,380
We have also videos for two weeks.
10
00:00:24,380 --> 00:00:25,910
So we have a good start.
11
00:00:25,910 --> 00:00:29,240
But how do we make this deepfake videos?
12
00:00:29,240 --> 00:00:31,310
This is relatively easy.
13
00:00:31,310 --> 00:00:33,260
We work with Google Colab.
14
00:00:33,260 --> 00:00:37,610
The Google Colab is named Roop one click deepfake.
15
00:00:37,610 --> 00:00:41,360
So first of all, what the hell is Google Colab?
16
00:00:41,780 --> 00:00:46,280
Google Colab is simply a cloud solution from Google.
17
00:00:46,280 --> 00:00:50,810
They borrow you some GPU power and they do that for free.
18
00:00:50,810 --> 00:00:53,330
Yes, this is completely for free.
19
00:00:53,330 --> 00:01:00,860
You can simply run code in this colab notebooks and Google will borrow you some GPU power, and they
20
00:01:00,860 --> 00:01:02,840
will do that completely for free.
21
00:01:02,840 --> 00:01:05,450
Yes, you can also make a subscription.
22
00:01:05,450 --> 00:01:08,030
I must admit I have a subscription.
23
00:01:08,030 --> 00:01:12,740
If you don't have a subscription, you need to use the smaller GPUs.
24
00:01:12,740 --> 00:01:19,370
You will automatically use a T4 graphics card and they are not that fast.
25
00:01:19,400 --> 00:01:21,650
But how this all works.
26
00:01:21,650 --> 00:01:24,590
So like I said, we can run code here.
27
00:01:24,590 --> 00:01:28,130
You can see we have a lot of code and we can run this.
28
00:01:28,130 --> 00:01:32,330
But prior to running this code we need to install something.
29
00:01:32,330 --> 00:01:34,610
So the first cell right here is this.
30
00:01:34,610 --> 00:01:37,430
We have a little cell and we have a play button.
31
00:01:37,430 --> 00:01:42,890
If we press this play button we will install everything that is listed right here.
32
00:01:42,980 --> 00:01:45,620
First of all we clone this GitHub link.
33
00:01:45,620 --> 00:01:47,690
So this all comes from GitHub.
34
00:01:47,690 --> 00:01:50,810
And then we install the requirements.
35
00:01:50,810 --> 00:01:55,610
So if you read pip install something then we install something.
36
00:01:55,610 --> 00:01:59,480
If we if you read clone something we clone something.
37
00:01:59,480 --> 00:02:04,160
So if we press play right here we must give permission right here.
38
00:02:04,160 --> 00:02:06,290
Because this stuff comes from GitHub.
39
00:02:06,290 --> 00:02:09,470
A programmer has made this colab notebook.
40
00:02:09,470 --> 00:02:13,520
So somebody other has made an entire colab notebook.
41
00:02:13,520 --> 00:02:19,400
And we can use that for free, because the GitHub community is nice and they like to give away their
42
00:02:19,400 --> 00:02:20,690
stuff for free.
43
00:02:20,690 --> 00:02:25,190
And you can see right here I'm now connected with a graphics card.
44
00:02:25,190 --> 00:02:29,030
And I also and I'm also connected with the T4.
45
00:02:29,030 --> 00:02:31,970
I have forgotten to connect a stronger one.
46
00:02:31,970 --> 00:02:39,440
So in this video I will also work with the free version of Google Colab, even if I pay for the subscription.
47
00:02:39,440 --> 00:02:41,900
So I do it just like everyone.
48
00:02:41,900 --> 00:02:45,620
We take this lower render time into account for this one.
49
00:02:45,950 --> 00:02:50,840
So simply press play at the first button and everything will get cloned.
50
00:02:50,840 --> 00:02:58,340
And then after the cloning, everything will get installed that we need to make our deepfake videos.
51
00:02:58,340 --> 00:03:02,450
This takes about two minutes and here we go.
52
00:03:02,450 --> 00:03:05,900
After roughly two minutes, all of this is installed.
53
00:03:05,900 --> 00:03:08,480
So we work with TensorFlow and so on.
54
00:03:08,480 --> 00:03:12,320
To be honest, this is not that important to understand.
55
00:03:12,320 --> 00:03:16,160
We simply scroll down and don't worry if there is some red text.
56
00:03:16,160 --> 00:03:18,260
This is all fine.
57
00:03:18,380 --> 00:03:24,740
Now we have this right here download model and we need to press play again.
58
00:03:24,740 --> 00:03:28,610
Now we go to huggingface and download some stuff again.
59
00:03:28,880 --> 00:03:30,860
So this is like really easy.
60
00:03:30,860 --> 00:03:37,010
We simply press play on this button and then we download most of the stuff from hugging face.
61
00:03:37,010 --> 00:03:41,450
Hugging face is also a community and most of them are programmers.
62
00:03:41,450 --> 00:03:46,010
Here you find models, checkpoints for stable diffusion and different stuff.
63
00:03:46,010 --> 00:03:49,760
All we have to do is to simply press play on this button.
64
00:03:49,760 --> 00:03:52,310
And this is really, really fast.
65
00:03:52,640 --> 00:03:55,700
So six seconds and everything is done.
66
00:03:55,700 --> 00:03:57,920
And now comes the fun part.
67
00:03:57,920 --> 00:04:00,410
We go to the deepfake cell.
68
00:04:00,410 --> 00:04:02,960
We have to press play on this button.
69
00:04:02,960 --> 00:04:06,380
But first we need to upload our stuff.
70
00:04:06,380 --> 00:04:08,570
And this is also really easy.
71
00:04:08,570 --> 00:04:14,660
We need to connect our videos, we need to upload our videos, and then we need to change some paths
72
00:04:14,660 --> 00:04:16,280
and we are ready to go.
73
00:04:16,460 --> 00:04:22,670
And how we do that, we can either connect our Google Drive or the simpler solution.
74
00:04:22,670 --> 00:04:25,460
We press on this folder right here.
75
00:04:25,580 --> 00:04:32,450
Then we go to our videos and pictures and then we start of course with our source image.
76
00:04:32,450 --> 00:04:34,640
This right here is the source image.
77
00:04:34,640 --> 00:04:39,260
And we simply drag and drop it right here and press okay.
78
00:04:39,260 --> 00:04:44,450
All of this will get deleted as soon as we close this colab notebook.
79
00:04:44,450 --> 00:04:50,090
And after this is uploaded, we also need to upload our videos.
80
00:04:50,090 --> 00:04:58,910
I start with video A and then we do also video B, so we simply upload video A also in this folder.
81
00:04:58,910 --> 00:04:59,810
And then we can.
82
00:04:59,980 --> 00:05:03,640
And simply change the paths and we are ready to go.
83
00:05:04,030 --> 00:05:05,290
And there we go.
84
00:05:05,290 --> 00:05:06,640
And now you see this.
85
00:05:06,640 --> 00:05:10,480
So afterwards we will run a bit of of Python code.
86
00:05:10,480 --> 00:05:14,530
And we need to give our path to the target video.
87
00:05:14,530 --> 00:05:19,150
So we simply need to delete this path right here.
88
00:05:19,150 --> 00:05:21,010
So we delete this.
89
00:05:21,010 --> 00:05:25,360
Then we go to our video that is our target video.
90
00:05:25,360 --> 00:05:27,280
We press on the three dots.
91
00:05:27,280 --> 00:05:29,320
We click Copy path.
92
00:05:29,320 --> 00:05:35,530
We go back and here into the target stuff we simply paste it in right here.
93
00:05:35,530 --> 00:05:39,550
And make sure to don't delete this space right here.
94
00:05:39,550 --> 00:05:41,860
And also some space right here.
95
00:05:41,860 --> 00:05:49,390
So dash dash target space then your path space and then dash dash output video quality and so on.
96
00:05:49,390 --> 00:05:51,520
We can leave this as it is.
97
00:05:51,520 --> 00:05:54,100
You can also delete the video quality.
98
00:05:54,100 --> 00:05:56,830
But I would just leave it as it is.
99
00:05:57,070 --> 00:06:00,730
So this right here is the path to our target video.
100
00:06:00,730 --> 00:06:05,080
And of course now we need to give the path to our source image.
101
00:06:05,080 --> 00:06:08,620
So we do exactly the same stuff.
102
00:06:08,620 --> 00:06:12,430
You can see it right here dash dash source.
103
00:06:12,430 --> 00:06:16,870
And after the source we need to delete this path.
104
00:06:16,870 --> 00:06:19,150
This is just an example path.
105
00:06:19,150 --> 00:06:22,930
After it's deleted we go back to our source image.
106
00:06:22,930 --> 00:06:24,370
We press three dots.
107
00:06:24,370 --> 00:06:26,080
We press copy path.
108
00:06:26,080 --> 00:06:30,730
We go back in it and we simply paste it right here.
109
00:06:30,730 --> 00:06:38,920
And of course always dash dash source a little bit of space and then a little bit of space after your
110
00:06:38,920 --> 00:06:39,250
path.
111
00:06:39,250 --> 00:06:40,870
And then you are ready to go.
112
00:06:40,870 --> 00:06:45,160
Everything that we need to do right now is to hit this play button.
113
00:06:45,160 --> 00:06:50,920
And then I think this will most likely take about 2 to 6 minutes.
114
00:06:50,920 --> 00:06:52,060
It depends a bit.
115
00:06:52,060 --> 00:06:56,350
Roughly one second of video takes one minute of generation.
116
00:06:56,860 --> 00:07:00,220
You see we download stuff from GitHub and so on.
117
00:07:00,220 --> 00:07:02,350
I think this is done in a few minutes.
118
00:07:02,350 --> 00:07:04,450
Just think about it this way.
119
00:07:04,450 --> 00:07:11,470
If you have a video that lasts 10s, this will most likely last ten minutes.
120
00:07:11,470 --> 00:07:14,800
So I see you as soon as this is rendered.
121
00:07:15,280 --> 00:07:17,110
And here we have it.
122
00:07:17,110 --> 00:07:22,150
This right here took 11 minutes to run, but now it's completely done.
123
00:07:22,150 --> 00:07:25,390
And normally we have our data right here.
124
00:07:25,390 --> 00:07:29,350
If the data is not here, we need to refresh this.
125
00:07:29,350 --> 00:07:34,120
And now it's here you can see Swaped dot mp4.
126
00:07:34,120 --> 00:07:40,930
And it's important to download this because if you disconnect your drive this will get automatically
127
00:07:40,930 --> 00:07:41,470
deleted.
128
00:07:41,470 --> 00:07:43,030
And that is not nice.
129
00:07:43,030 --> 00:07:47,350
So press the three dots and then download this.
130
00:07:47,710 --> 00:07:53,620
As soon as this is downloaded you can and should maybe delete it and make other videos.
131
00:07:53,620 --> 00:07:56,620
That's simply the more efficient way.
132
00:07:56,620 --> 00:08:00,970
You simply need to delete the stuff that you need to make new.
133
00:08:00,970 --> 00:08:08,650
So the source image, of course we need that the swapped video we can delete because now we have downloaded
134
00:08:08,650 --> 00:08:08,920
it.
135
00:08:08,920 --> 00:08:11,320
So delete the swapped video.
136
00:08:11,320 --> 00:08:15,910
It's also important to delete the first target video.
137
00:08:15,910 --> 00:08:19,000
So we delete also the target video.
138
00:08:19,000 --> 00:08:22,390
And now we upload the target video too.
139
00:08:22,390 --> 00:08:25,720
We simply drag and drop it right here in.
140
00:08:25,720 --> 00:08:28,570
And after that we change the path.
141
00:08:28,570 --> 00:08:36,520
So we go back to the path, the target path also this path we will delete and then put our new path.
142
00:08:36,520 --> 00:08:43,720
And the source path is of course the same because we use the same face to put on this video.
143
00:08:43,720 --> 00:08:46,510
So I have to wait until this is uploaded.
144
00:08:46,510 --> 00:08:49,660
Then we change this path and we press play again.
145
00:08:49,660 --> 00:08:52,000
And then we have both our videos.
146
00:08:52,420 --> 00:08:55,150
So now it's uploaded, we do the same stuff.
147
00:08:55,150 --> 00:09:01,390
We press the three dots, we copy the path, we go back, we delete the old path.
148
00:09:01,390 --> 00:09:03,130
So delete this right here.
149
00:09:03,130 --> 00:09:05,380
No need to install other stuff.
150
00:09:05,380 --> 00:09:06,670
This is all installed.
151
00:09:06,670 --> 00:09:09,970
We have our models, we have our checkpoints and so on.
152
00:09:09,970 --> 00:09:14,080
We simply press play and then we will get our second video.
153
00:09:14,080 --> 00:09:17,590
And I think this will also take something like ten minutes.
154
00:09:17,590 --> 00:09:19,930
But the first video we have downloaded.
155
00:09:19,930 --> 00:09:21,400
And now let's take a look.
156
00:09:21,400 --> 00:09:25,360
If the first video looks somewhat like our influencer.
157
00:09:25,660 --> 00:09:29,950
So we press three dots right here and we go into the downloads.
158
00:09:30,220 --> 00:09:33,250
In the downloads we will find our video.
159
00:09:33,640 --> 00:09:38,440
I called the video, swapped one and we simply look at it.
160
00:09:38,680 --> 00:09:40,990
It's here just a moment.
161
00:09:43,760 --> 00:09:46,190
So yeah, I think it's okay.
162
00:09:46,190 --> 00:09:54,440
The quality is maybe not 100% perfect, but it looks relatively similar than our influencer.
163
00:09:54,440 --> 00:09:55,490
We should.
164
00:09:55,490 --> 00:10:02,210
Or maybe we could upscale this, but the face, yeah, it looks relatively the same than our influencer.
165
00:10:02,210 --> 00:10:04,790
And she's dancing and that's what's all about.
166
00:10:04,790 --> 00:10:09,170
And remember on a phone this is really, really a lot smaller.
167
00:10:09,170 --> 00:10:15,890
Let's open up our face and see if that could actually be our eye influencer.
168
00:10:15,890 --> 00:10:19,370
And I think they are relatively similar.
169
00:10:19,370 --> 00:10:22,820
So I think the Colab notebook did a good job.
170
00:10:22,820 --> 00:10:27,530
We have our face on this eye influencer and she's dancing.
171
00:10:27,530 --> 00:10:27,740
Yeah.
172
00:10:27,740 --> 00:10:30,650
That looks that looks good in my mind.
173
00:10:30,650 --> 00:10:33,110
I think we did a good job.
174
00:10:33,110 --> 00:10:37,910
And I need to see how the how the second video behaves.
175
00:10:37,910 --> 00:10:40,820
I think also the second video will get.
176
00:10:41,440 --> 00:10:44,350
Nearly the same, or even better.
177
00:10:44,350 --> 00:10:48,880
So let's just take a look at the second video as soon as this is done.
178
00:10:49,270 --> 00:10:52,480
And this time it only take four minutes.
179
00:10:52,480 --> 00:10:55,300
And we have already the video right here.
180
00:10:55,300 --> 00:10:57,400
So swapped dot mp4.
181
00:10:57,430 --> 00:11:03,640
Remember please remember to download this immediately because as soon as you disconnect, this will
182
00:11:03,640 --> 00:11:05,110
get deleted.
183
00:11:05,110 --> 00:11:12,430
If you let it open for 15 to 20 minutes, it is possible that it disconnects automatically and then
184
00:11:12,430 --> 00:11:16,090
the video is gone and you need to start over again.
185
00:11:16,090 --> 00:11:23,140
So now we go again into the downloads and I want to see if this video worked as well.
186
00:11:23,560 --> 00:11:28,270
I name it what we put it right here in it.
187
00:11:28,330 --> 00:11:31,270
I open this swapped video.
188
00:11:31,270 --> 00:11:33,430
So this is the second one.
189
00:11:35,060 --> 00:11:36,770
Yeah, I think this works.
190
00:11:36,770 --> 00:11:38,570
Maybe even better.
191
00:11:39,540 --> 00:11:42,690
Um, we open our source image.
192
00:11:42,690 --> 00:11:50,400
So this is not our source image, but that's also fine I think because we can also look at other stuff.
193
00:11:50,430 --> 00:11:55,140
Yeah, I think this is nearly perfect with our AI influencer.
194
00:11:55,140 --> 00:11:59,430
So we have the same face and now we have some dancing videos.
195
00:11:59,460 --> 00:12:02,430
The videos are somewhat nice.
196
00:12:02,430 --> 00:12:09,960
They are maybe not 100% perfect, but this is the best that this stuff can do right now.
197
00:12:09,960 --> 00:12:10,590
If we.
198
00:12:10,590 --> 00:12:17,190
If you want to make it even better, you can make a little upscale of these videos, but I think they
199
00:12:17,190 --> 00:12:19,290
are good as they are.
200
00:12:20,240 --> 00:12:27,050
Just remember, 99% of the people will see this on a phone and then they are even better.
201
00:12:27,050 --> 00:12:30,350
And of course, we don't post such videos every day.
202
00:12:30,350 --> 00:12:38,480
They are a little they they take a little bit of time to make, but I think, yeah, it looks pretty
203
00:12:38,480 --> 00:12:39,770
-- realistic.
204
00:12:39,770 --> 00:12:41,690
It can create some clicks.
205
00:12:41,690 --> 00:12:44,150
This is really, really nice stuff.
206
00:12:44,240 --> 00:12:49,820
So in this video we took our target video and we swapped the faces.
207
00:12:49,820 --> 00:12:52,640
And the process is really, really simple.
208
00:12:52,640 --> 00:12:59,660
Just open the Colab notebook, press play, and then you upload your video and the source image.
209
00:12:59,660 --> 00:13:02,720
You change the path and press play again.
210
00:13:02,720 --> 00:13:09,020
You wait a few minutes and then you have your video and after your video is done, of course you can
211
00:13:09,020 --> 00:13:16,040
post also your video, maybe take a little bit of music over it and then you are really ready to go.
212
00:13:16,040 --> 00:13:23,330
And after this process we will even create our own voice for our AI influencer.
213
00:13:23,330 --> 00:13:26,360
And that we will do in the next video.
214
00:13:26,360 --> 00:13:31,460
Because right now I don't see any AI influencers with own voices.
215
00:13:31,460 --> 00:13:35,600
And maybe you can stand out if you create your own voice.
19004
Can't find what you're looking for?
Get subtitles in any language from opensubtitles.com, and translate them here.
Afrikaans
Akan
Albanian
Amharic
Armenian
Azerbaijani
Basque
Belarusian
Bemba
Bengali
Bihari
Bosnian
Breton
Bulgarian
Cambodian
Catalan
Cebuano
Cherokee
Chichewa
Chinese (Simplified)
Chinese (Traditional)
Corsican
Croatian
Czech
Danish
Dutch
English
Esperanto
Estonian
Faroese
Finnish
Galician
Georgian
German
Greek
Guarani
Haitian Creole
Hausa
Hawaiian
Hebrew
Hmong
Hungarian
Icelandic
Indonesian
Irish
Italian
Japanese
Kazakh
Kinyarwanda
Kirundi
Kongo
Korean
Krio (Sierra Leone)
Kurdish
Kyrgyz
Laothian
Latin
Latvian
Lithuanian
Luganda
Luo
Luxembourgish
Macedonian
Malagasy
Malay
Maltese
Maori
Mauritian Creole
Moldavian
Mongolian
Myanmar (Burmese)
Montenegrin
Nepali
Norwegian
Pashto
Persian
Polish
Portuguese (Brazil)
Portuguese (Portugal)
Quechua
Romanian
Romansh
Russian
Samoan
Scots Gaelic
Serbian
Sesotho
Setswana
Seychellois Creole
Shona
Sindhi
Sinhalese
Slovak
Slovenian
Somali
Spanish
Swedish
Tajik
Thai
Tigrinya
Tonga
Turkish
Turkmen
Ukrainian
Uzbek
Vietnamese
Welsh
Wolof