Would you like to inspect the original subtitles? These are the user uploaded subtitles that are being translated:
1
00:00:00,256 --> 00:00:06,400
So let's take a look at how you can enlarge images inside of stable diffusion we're
2
00:00:06,656 --> 00:00:08,704
Is it going to be looking at methods that can be used
3
00:00:08,960 --> 00:00:15,104
Here in Stapleton 1.5 then what is it going to be mixing and matching
4
00:00:15,360 --> 00:00:18,688
According to taste so we're going to be looking at the control net
5
00:00:20,992 --> 00:00:26,368
An enlargement using a built-in method which is called latent upscale
6
00:00:26,624 --> 00:00:31,488
And what's going to be looking at another way of enlarging using models so this is going to be
7
00:00:31,744 --> 00:00:35,328
Feature packed video so you want to just
8
00:00:35,584 --> 00:00:37,120
Grab a seat and hold on
9
00:00:37,888 --> 00:00:41,728
So let me explain what we're looking at here first of all we've got the checkpoint coming in
10
00:00:41,984 --> 00:00:44,288
That is the dream Point checkpoint
11
00:00:44,544 --> 00:00:47,616
And that is taking us through to the
12
00:00:48,128 --> 00:00:49,408
Load vae
13
00:00:49,920 --> 00:00:56,064
And that's obviously the standard vae but we're now using a TI adapter why not
14
00:00:56,320 --> 00:00:56,832
It's very quick
15
00:00:57,344 --> 00:01:03,488
And we then have some text in almost that you seen from above forming an alien housing structure ocean
16
00:01:04,256 --> 00:01:06,560
Okay so that's the description
17
00:01:06,816 --> 00:01:08,608
No sailing ships we've got
18
00:01:09,120 --> 00:01:12,448
An empty Layton just 512 by 512
19
00:01:12,704 --> 00:01:16,032
Which is fairly standard for stable diffusion 1.5
20
00:01:18,592 --> 00:01:21,664
Wake up the depth map of the depth control
21
00:01:21,920 --> 00:01:26,528
Which I think we should really change to TI
22
00:01:29,856 --> 00:01:31,392
And we'll call that
23
00:01:31,648 --> 00:01:32,416
Apply
24
00:01:36,768 --> 00:01:38,304
Number two
25
00:01:45,472 --> 00:01:46,240
Adapter
26
00:01:50,592 --> 00:01:53,920
So we've got the empty latent image we've got the adapter
27
00:01:54,432 --> 00:01:58,528
Coming in and changing this to
28
00:01:58,784 --> 00:01:59,808
Beautiful
29
00:02:00,832 --> 00:02:01,600
Death mask
30
00:02:02,112 --> 00:02:07,744
Is being used to create this image which is the 51255 1/2
31
00:02:08,000 --> 00:02:14,144
And this image is again we're looking at an enormous that you've seen from about forming an alien housing structure
32
00:02:14,912 --> 00:02:16,192
Ocean
33
00:02:16,448 --> 00:02:22,592
Photographic we don't want text disgusting ugly Watermark or dark cuz I like everything's nice
34
00:02:23,616 --> 00:02:26,432
We've got some ocean in the background it's looking good
35
00:02:27,200 --> 00:02:33,344
When we zoom in we can see quite a little detail which is something I really like about this model and how it
36
00:02:34,112 --> 00:02:36,416
And obviously stable diffusion itself
37
00:02:37,184 --> 00:02:42,304
Now it might be a little bit surprising but we're looking at here we're seeing someone
38
00:02:42,560 --> 00:02:44,096
Who
39
00:02:44,608 --> 00:02:46,400
Obviously pretty alien
40
00:02:46,912 --> 00:02:52,544
And I think it's the Alien Part in Alien housing structure
41
00:02:52,800 --> 00:02:54,336
I want a futuristic
42
00:02:56,896 --> 00:03:03,040
I wanted something futuristic I didn't woman who looks like an alien but I've got a woman who
43
00:03:03,296 --> 00:03:04,064
Looks like an alien
44
00:03:04,320 --> 00:03:09,440
And it's a beautiful structure she's got these beautiful eyebrows she's got
45
00:03:09,952 --> 00:03:12,256
The things you know that
46
00:03:15,840 --> 00:03:18,400
Her head definitely looks overall
47
00:03:18,656 --> 00:03:20,448
The the color and shape
48
00:03:20,704 --> 00:03:22,752
That we would expect from an alien
49
00:03:23,264 --> 00:03:27,104
And aliens have done something beautiful with this with this little
50
00:03:28,128 --> 00:03:34,272
It looks good it looks good I think we we humans will be proud to create something that looked as spectacular as
51
00:03:34,784 --> 00:03:40,928
I really like that it's also respected the the idea that we have an enormous statue seen from above
52
00:03:41,184 --> 00:03:45,536
It is enormous and we are actually seeing it from above it did not want to do that
53
00:03:46,048 --> 00:03:48,352
With another of the images that I worked with
54
00:03:48,864 --> 00:03:53,984
So we've got an image being seen above and it's fascinating to see how it's taking this image this
55
00:03:54,496 --> 00:03:57,568
Deathmatch depth map and just really
56
00:03:57,824 --> 00:03:59,872
Played with it to create something
57
00:04:00,640 --> 00:04:04,224
Resembles the original image in terms of his overall composition
58
00:04:04,480 --> 00:04:07,296
But which is really creative and very original
59
00:04:09,088 --> 00:04:13,952
With an enlarged that image 1.5 times and you can see as we enlarge it we're getting
60
00:04:14,208 --> 00:04:15,744
Detail more detail
61
00:04:16,256 --> 00:04:20,351
We're getting we lose this with tentacles that he decided to create for a woman
62
00:04:20,863 --> 00:04:25,983
She has something now that looks like a crown or something I don't know what's going on there
63
00:04:27,263 --> 00:04:31,871
Are there these things coming out from her side it could be some kind of
64
00:04:32,127 --> 00:04:35,711
I don't know I don't know you never know with these aliens they've got all sorts of things going on
65
00:04:35,967 --> 00:04:39,039
And what we see now you can see
66
00:04:39,295 --> 00:04:43,647
It's a huge structure you can see little bits and pieces of vegetation
67
00:04:43,903 --> 00:04:47,231
Something that looks like maybe it could be a walkway there
68
00:04:47,743 --> 00:04:53,887
And it looks to me like that they worked hard to build this this is this is a giant building then you'd see this
69
00:04:54,143 --> 00:04:54,655
I'm a mile away
70
00:04:55,167 --> 00:04:57,471
And then we have another one
71
00:04:57,727 --> 00:05:01,311
Which is even larger one and a half times larger than this one so
72
00:05:01,567 --> 00:05:06,687
I don't know you can figure out how long it is in Total 1 1/2 times 1 1/2 times
73
00:05:07,199 --> 00:05:13,343
So it's now got a lot more detail it's almost as though we're looking at a whole kind of some kind of
74
00:05:13,599 --> 00:05:15,135
Auditorium or some kind of
75
00:05:15,903 --> 00:05:19,999
The people here will the aliens will be tiny if they were
76
00:05:20,255 --> 00:05:21,279
In that area there
77
00:05:21,535 --> 00:05:23,071
Please added more detail
78
00:05:23,327 --> 00:05:27,679
We've got something that looks like some kind of vegetation like moss
79
00:05:28,191 --> 00:05:29,471
Growing
80
00:05:31,519 --> 00:05:36,127
Whole surface I don't know what that look like original it looks like it was planned here
81
00:05:36,639 --> 00:05:38,175
It look like they they actually
82
00:05:38,431 --> 00:05:39,967
Created some kind of
83
00:05:40,735 --> 00:05:44,575
Effect where you have two colors and it looks really nice
84
00:05:45,855 --> 00:05:49,183
Begins to break down here we begin to see all kinds of decay
85
00:05:49,439 --> 00:05:50,207
And then
86
00:05:50,719 --> 00:05:53,279
We begin to see like the you know
87
00:05:53,791 --> 00:05:55,839
Nature is beginning to reclaim
88
00:05:56,095 --> 00:05:56,863
This area
89
00:05:57,631 --> 00:06:01,471
Her eyebrows have begun to really sort of Fall Apart
90
00:06:01,983 --> 00:06:06,847
And what were these little things that were coming out from her sides these
91
00:06:07,359 --> 00:06:08,639
I don't know what those are
92
00:06:10,175 --> 00:06:15,807
One of them has turned into a kind of like decayed area where you can see sunlight coming through
93
00:06:16,063 --> 00:06:17,343
Some kind of
94
00:06:18,111 --> 00:06:19,647
I don't know deck
95
00:06:20,415 --> 00:06:26,559
Dereliction I don't know what's going on there that they haven't maintained that part of the statue or the part of
96
00:06:30,399 --> 00:06:34,495
Somewhat coherent but the original idea the original image
97
00:06:34,751 --> 00:06:37,823
Is beginning to the original
98
00:06:38,079 --> 00:06:43,199
Idea here this world built structure is we've lost it to some extent and we have something which is
99
00:06:43,711 --> 00:06:44,735
Much more decay
100
00:06:44,991 --> 00:06:46,783
Much less impressive
101
00:06:47,807 --> 00:06:53,439
And in many ways I don't know if it's almost like it's grown in size and also
102
00:06:53,695 --> 00:06:55,999
It's aged as well
103
00:06:56,511 --> 00:06:57,791
Whilst it was growing
104
00:06:58,303 --> 00:07:03,423
What we're saying is the creation of detail was seeing
105
00:07:03,679 --> 00:07:07,007
The creation of something that's meaningful within the context of
106
00:07:07,263 --> 00:07:12,127
The thing that we're looking at it it is this weird alien structure
107
00:07:12,639 --> 00:07:15,455
Scene from above with the ocean around it
108
00:07:15,711 --> 00:07:21,855
Is coherent but it's not coherent across the scaling so we lose something
109
00:07:22,111 --> 00:07:23,135
Play the original design
110
00:07:24,415 --> 00:07:29,535
It creates detail but the detail takes us away from where we were originally
111
00:07:30,559 --> 00:07:33,631
Now let's take a look at how we did this this affect here
112
00:07:34,399 --> 00:07:38,751
The effect is achieved by what's known as upscale latent so
113
00:07:39,263 --> 00:07:40,543
Upscale method
114
00:07:41,055 --> 00:07:44,639
Nearest exact you can change this to different options
115
00:07:44,895 --> 00:07:50,527
What else can you buy 1.5 as I mentioned before this method allows us to use a case sampler
116
00:07:52,063 --> 00:07:53,087
Upscale
117
00:07:53,343 --> 00:07:57,951
Moving from 20 to 40 in terms of the steps
118
00:07:58,207 --> 00:08:01,023
So the original one is the first 20 steps
119
00:08:01,535 --> 00:08:04,095
And then we moved from 20 to
120
00:08:05,119 --> 00:08:11,263
240 in the second one and the third one I've moved from 40 to 60 so we're just building on what
121
00:08:11,519 --> 00:08:15,871
What was the original which might explain this kind of progression that we see
122
00:08:17,151 --> 00:08:19,455
The this image and that image
123
00:08:19,711 --> 00:08:22,527
And also from this image to that image there's a progression
124
00:08:22,783 --> 00:08:27,391
But there's something of the essence of the of the idea that's retained
125
00:08:27,647 --> 00:08:31,231
It is in this in this building on steps
126
00:08:31,743 --> 00:08:35,583
20 to 40 and 40 to 60 that's why we see that progression
127
00:08:36,095 --> 00:08:37,887
A lot of what we're seeing here
128
00:08:38,655 --> 00:08:43,007
In terms of the difference is is the desire for the software
129
00:08:43,263 --> 00:08:46,591
To create detail and it's doing that creation of detail
130
00:08:46,847 --> 00:08:50,687
Within latent space so we're upscaling violent
131
00:08:50,943 --> 00:08:53,759
And that allows it to invent detail
132
00:08:54,015 --> 00:08:55,807
Using its own knowledge
133
00:08:56,063 --> 00:08:59,135
Of what's happening and what's happening
134
00:09:00,671 --> 00:09:02,719
Explain to it in two different ways
135
00:09:03,999 --> 00:09:08,607
We have one prompt which explains what we want to see which is the control net
136
00:09:09,119 --> 00:09:13,727
And then we have enough 100% is 0.86
137
00:09:14,751 --> 00:09:16,543
And then we have another prompt
138
00:09:16,799 --> 00:09:17,567
The word
139
00:09:17,823 --> 00:09:20,383
And it's using those two to create
140
00:09:20,639 --> 00:09:22,175
This structure
141
00:09:22,431 --> 00:09:26,783
At every stage and the reason why it's using both of them is because
142
00:09:27,295 --> 00:09:29,343
On all of these K samples
143
00:09:32,159 --> 00:09:33,695
I've connected the positive
144
00:09:33,951 --> 00:09:35,487
Note to the prompt
145
00:09:37,535 --> 00:09:38,303
To the control net
146
00:09:41,375 --> 00:09:43,423
To the negative to the positive
147
00:09:43,935 --> 00:09:45,471
And also to the positive here
148
00:09:45,983 --> 00:09:47,775
We have
149
00:09:50,335 --> 00:09:51,359
The negative
150
00:09:51,615 --> 00:09:56,223
So we could have tried something like text disgusting ugly Watermark dark
151
00:09:56,479 --> 00:09:57,759
And we could have tried
152
00:09:58,783 --> 00:10:00,319
I could have put in decay
153
00:10:00,575 --> 00:10:01,855
Or we could have put in
154
00:10:02,367 --> 00:10:06,463
Plants to try to remove some of the vegetation here
155
00:10:06,719 --> 00:10:12,863
In the negative that might have worked because the negative are all being taken into this case
156
00:10:13,119 --> 00:10:16,447
I'll give you this particular workflow
157
00:10:16,703 --> 00:10:22,591
And it will allow you to play around and that's really the same here there's so many ways of upscaling
158
00:10:22,847 --> 00:10:26,175
That you really need to play around with them to see what they can do
159
00:10:26,431 --> 00:10:29,247
And there's so many different outcomes that we can get
160
00:10:29,759 --> 00:10:35,903
It's a good idea to play around play around with all of the the settings here not all of them but
161
00:10:36,159 --> 00:10:37,183
Is there any play around
162
00:10:37,439 --> 00:10:40,255
With the sampler name so change some of these
163
00:10:41,535 --> 00:10:43,583
Very much play around with the
164
00:10:43,839 --> 00:10:47,167
Scheduler so exponential Cara's normal
165
00:10:47,423 --> 00:10:49,471
See which one produces the better
166
00:10:49,983 --> 00:10:51,263
The better outcome
167
00:10:51,519 --> 00:10:52,543
So for instance
168
00:10:55,871 --> 00:11:00,991
What I could do is to change this to let's say SG
169
00:11:01,247 --> 00:11:02,015
Uniform
170
00:11:02,783 --> 00:11:06,111
And we could take this from simple to exponential
171
00:11:07,135 --> 00:11:09,183
Then we accuse that and see what happens
172
00:11:13,535 --> 00:11:14,559
Oh that's beautiful
173
00:11:19,167 --> 00:11:20,191
Okay okay
174
00:11:24,031 --> 00:11:27,359
And this one takes quite a long time this is the one where we're going from the
175
00:11:27,871 --> 00:11:30,431
Enlarged one to the ultra enlarged one
176
00:11:38,623 --> 00:11:40,671
You can see that
177
00:11:40,927 --> 00:11:47,071
Play probably I don't know about you but I probably wouldn't use exponential again so that kind of experimentation can
178
00:11:47,327 --> 00:11:49,887
Tell us a lot about what's working and what's not working
179
00:11:52,703 --> 00:11:55,007
So what I'm going to do I'm going to reload
180
00:11:55,263 --> 00:12:00,639
The one we just did will change this one maybe to normal again
181
00:12:01,151 --> 00:12:03,711
Because that seemed to work a little bit more
182
00:12:04,223 --> 00:12:07,551
A little bit better and maybe we'll put in
183
00:12:15,999 --> 00:12:16,511
Deck
184
00:12:18,303 --> 00:12:21,631
And we'll run the previous one again so
185
00:12:22,399 --> 00:12:26,239
I'm going to run this one again to see what it does with the new
186
00:12:26,751 --> 00:12:28,799
Wording and also with normal
187
00:12:30,335 --> 00:12:31,871
Now you might be wondering
188
00:12:32,127 --> 00:12:33,407
What happens
189
00:12:33,663 --> 00:12:35,711
So I've been incremented the
190
00:12:36,991 --> 00:12:39,551
Okay so it's changed a little bit in terms of the
191
00:12:40,319 --> 00:12:41,087
The noise
192
00:12:41,599 --> 00:12:44,415
We're getting more or less the same outcome
193
00:12:46,463 --> 00:12:51,839
No I think it is actually just the changes in the in the word prompt that created this result
194
00:12:52,095 --> 00:12:56,191
So that's something that actually work better without the
195
00:12:56,447 --> 00:12:59,519
Without the changes to the to the words to the
196
00:13:00,031 --> 00:13:02,079
The word prompt let's go back to where we were
197
00:13:02,591 --> 00:13:05,151
And let's change this to
198
00:13:05,407 --> 00:13:06,431
Normal
199
00:13:07,455 --> 00:13:11,295
And what will do is take a look at another option
200
00:13:12,063 --> 00:13:18,207
For enlarging and so what I'll do first of all is I'll explain how we get this this kind of enlargement
201
00:13:18,463 --> 00:13:22,047
What we do is that we start off with the
202
00:13:22,559 --> 00:13:23,327
Output
203
00:13:23,583 --> 00:13:25,887
That we have here so we have the sample the
204
00:13:26,143 --> 00:13:27,423
V a e d code
205
00:13:27,935 --> 00:13:30,751
And then the the outcome the the image
206
00:13:31,007 --> 00:13:35,359
And what we need to do is just basically grab the latent
207
00:13:36,639 --> 00:13:39,455
Push it through to a latent upscale
208
00:13:41,503 --> 00:13:43,807
We getting late in upscale by
209
00:13:44,319 --> 00:13:47,903
And that allows us to choose the method I'm going to keep it at that
210
00:13:48,415 --> 00:13:52,255
We can take the latent and bring it to a case sampler advanced
211
00:13:53,023 --> 00:13:59,167
And that allows us to set these settings so here would go from 60 to 80 or whatever whatever we wanted
212
00:13:59,679 --> 00:14:02,239
And then obviously we would do the latent
213
00:14:03,775 --> 00:14:06,335
The to the vaed code
214
00:14:07,359 --> 00:14:09,919
And then we'll create our image preview
215
00:14:11,967 --> 00:14:12,735
Preview image
216
00:14:13,503 --> 00:14:14,015
Like that
217
00:14:14,783 --> 00:14:19,135
And that would allow us to preview the image which is what I've got here these are all preview images
218
00:14:19,391 --> 00:14:20,671
That's a preview image
219
00:14:20,927 --> 00:14:25,791
This particular situation up with the upscaling
220
00:14:26,047 --> 00:14:29,887
And it's a very effective way of using the built-in processes to upscale
221
00:14:30,399 --> 00:14:36,287
I'm going to now delete these cuz we don't really need another upscale do we and it was getting a little bit slow
222
00:14:36,799 --> 00:14:37,567
So we don't want
223
00:14:37,823 --> 00:14:38,591
Push it
224
00:14:39,103 --> 00:14:43,455
When we can do now is to look at another technique for upscaling another
225
00:14:43,711 --> 00:14:47,295
The thing that you should be aware of is if you just shift click
226
00:14:47,807 --> 00:14:49,599
A bunch of nodes
227
00:14:49,855 --> 00:14:50,879
Control C
228
00:14:53,183 --> 00:14:55,487
And then Ctrl shift
229
00:14:57,023 --> 00:15:00,351
It copies everything with all the links in place
230
00:15:00,607 --> 00:15:04,447
And then what you would do in that situation is the decide
231
00:15:04,703 --> 00:15:10,847
Basically where you going to get your positive from so you could say I'm going to produce a completely new
232
00:15:11,871 --> 00:15:13,151
I like that
233
00:15:13,407 --> 00:15:18,015
Grab this and say quit texting code we can put in a completely new
234
00:15:18,783 --> 00:15:21,343
Text we can put in a completely new prompt
235
00:15:21,855 --> 00:15:26,207
Or we could connect it to the original prompto we can connect it to the
236
00:15:26,463 --> 00:15:27,999
Control prompt
237
00:15:28,511 --> 00:15:30,047
Go to the control conditioning
238
00:15:30,303 --> 00:15:33,375
So control C control shift fee
239
00:15:35,167 --> 00:15:36,703
Allows you to copy everything
240
00:15:36,959 --> 00:15:41,823
With the links in place let's take a look at another technique which we could use
241
00:15:42,335 --> 00:15:44,127
For enlarging and this one
242
00:15:44,383 --> 00:15:46,431
Relies on the use of a model
243
00:15:47,711 --> 00:15:53,855
Change this to simple I think it was actually a simple
244
00:15:54,111 --> 00:15:54,623
Not normal
245
00:15:55,903 --> 00:16:02,047
Let's go ahead and create this new upscaling method so we'll go to the original one cuz this one is a huge upscale
246
00:16:02,559 --> 00:16:04,607
And we can actually grab
247
00:16:04,863 --> 00:16:06,143
Let's see
248
00:16:13,055 --> 00:16:14,591
We'll grab the image from
249
00:16:15,359 --> 00:16:17,151
From the vaed code
250
00:16:17,407 --> 00:16:18,431
And then add
251
00:16:19,711 --> 00:16:23,551
Lotus no let me let me do the different way so we'll just
252
00:16:24,063 --> 00:16:28,159
Go to a fresh ground here and double click
253
00:16:28,415 --> 00:16:29,695
And we'll choose
254
00:16:33,791 --> 00:16:36,095
Preview image so we put the image there
255
00:16:36,607 --> 00:16:40,447
We need to now connect this to something that will create an image
256
00:16:40,703 --> 00:16:42,495
And that's going to be using a model
257
00:16:43,775 --> 00:16:47,615
So we'll go ahead and right click add node loaders
258
00:16:47,871 --> 00:16:50,687
And we going to choose load upscale model
259
00:16:52,991 --> 00:16:57,343
That is on upscale model we're going to use Let's see we can use
260
00:16:57,855 --> 00:16:59,135
Will use Astro again
261
00:16:59,391 --> 00:17:02,463
Now there are a lot of models that you can use for upscaling this one
262
00:17:02,719 --> 00:17:05,279
You should have at least one model
263
00:17:05,535 --> 00:17:09,375
In your system but you can find a lot of models online
264
00:17:09,631 --> 00:17:11,679
This is an area where you really want to
265
00:17:11,935 --> 00:17:14,495
Differentiate and look at how
266
00:17:15,263 --> 00:17:20,895
What models work best with your own kind of workflow so for me I think we'll try
267
00:17:21,151 --> 00:17:22,175
Maybe this one here
268
00:17:23,711 --> 00:17:28,319
These models are quite often called Ganz or generative adversarial Networks
269
00:17:28,575 --> 00:17:31,391
They're very fast so what we can do is to
270
00:17:33,183 --> 00:17:39,327
This one is uploading all models so it's like one of the other uploaders and you
271
00:17:39,583 --> 00:17:40,351
Have to have
272
00:17:40,607 --> 00:17:43,423
The model stored away in a specific part
273
00:17:44,191 --> 00:17:44,703
What's
274
00:17:45,215 --> 00:17:48,287
Configure I folder so let me show you where that is
275
00:17:49,311 --> 00:17:51,615
So what you do is you would
276
00:17:51,871 --> 00:17:52,639
Basically
277
00:17:52,895 --> 00:17:53,663
Google
278
00:17:54,431 --> 00:17:56,479
For upscale models for
279
00:17:56,735 --> 00:17:58,527
Stable diffusion
280
00:17:58,783 --> 00:18:04,927
And then you would go and grab your models and place them inside models upscale models so I've got the three
281
00:18:05,183 --> 00:18:11,327
I think one of them is being picked up from automatic 11:11 and
282
00:18:11,583 --> 00:18:16,703
Once you've got the models and you probably want to pick up quite a lot because they behave differently and all of them
283
00:18:17,215 --> 00:18:21,311
Most of them are quite valuable because they can produce different results in different situations
284
00:18:21,567 --> 00:18:22,847
What we want to do is to
285
00:18:23,359 --> 00:18:26,431
Wrap this guy here and then just bring out
286
00:18:29,503 --> 00:18:30,527
Let's say
287
00:18:30,783 --> 00:18:33,343
Upscale image upscale with model okay
288
00:18:33,599 --> 00:18:38,463
So this is going to upscale the image using this model and then we can output
289
00:18:38,975 --> 00:18:40,255
To the image here
290
00:18:41,791 --> 00:18:47,935
Is a simple as that and then we need to actually grab the image and where where do we grab the image
291
00:18:48,191 --> 00:18:53,055
Will we grab the image from the vaed code here so that completes the loop
292
00:18:53,311 --> 00:18:56,127
We get our image from the original VA
293
00:18:56,383 --> 00:18:57,151
Deck
294
00:18:57,407 --> 00:19:00,735
Nope that's the wrong one let's go and grab the right one
295
00:19:05,343 --> 00:19:07,903
Right at the beginning that the very first one here
296
00:19:08,415 --> 00:19:12,255
So we'll grab that one and let's move this here
297
00:19:28,383 --> 00:19:32,479
Then I'll preview image let's run this one and you can actually see how quick
298
00:19:32,735 --> 00:19:33,247
Next day
299
00:19:33,503 --> 00:19:34,271
So if you
300
00:19:34,783 --> 00:19:38,367
See how quickly is running now and you'll be able to see
301
00:19:40,159 --> 00:19:42,463
Bake the images I think
302
00:19:42,975 --> 00:19:44,511
This one is four times
303
00:20:04,991 --> 00:20:09,855
So that was super quick and you can see we have a an image which is significantly larger than that one
304
00:20:10,879 --> 00:20:15,231
I don't think this is the best example what I want to do is just to run it again
305
00:20:15,743 --> 00:20:16,767
Using
306
00:20:17,279 --> 00:20:19,071
A slightly different
307
00:20:20,607 --> 00:20:21,375
Front
308
00:20:28,799 --> 00:20:31,871
And let's see if we can do a comparison of the speed
309
00:20:32,127 --> 00:20:34,943
So we're now running the case sampler for the first image
310
00:20:36,223 --> 00:20:38,527
Are we on and then for the second image
311
00:20:43,135 --> 00:20:47,487
And then we should see how slow this one is compared
312
00:20:47,743 --> 00:20:49,279
To the game so I think
313
00:20:57,727 --> 00:21:00,287
Straight from here
314
00:21:00,543 --> 00:21:01,311
Through to that
315
00:21:03,359 --> 00:21:09,503
And this guy here is still running so you can see this one is actually quite a bit slower than the Gan than the
316
00:21:11,039 --> 00:21:12,063
Ezra again
317
00:21:12,319 --> 00:21:13,087
Times
318
00:21:13,343 --> 00:21:16,672
This is created an image four times larger than the original
319
00:21:16,928 --> 00:21:22,304
Didn't take any time at all and if we zoom in we can see it's done interesting stuff for the detail
320
00:21:22,816 --> 00:21:26,912
And I'm actually going to start running another one just to see what happens this time around
321
00:21:27,168 --> 00:21:32,032
But you can see that that is created in an image which I think looks better than this one
322
00:21:34,336 --> 00:21:38,432
And with this new reminder you can see again we have a very different outcome here
323
00:21:42,528 --> 00:21:43,296
I think
324
00:21:43,552 --> 00:21:48,160
Should have more of a bit more of resembles resemblance to this one
325
00:21:48,416 --> 00:21:50,720
So I'm kind of wondering what would happen if we instead of we
326
00:21:50,976 --> 00:21:53,024
Took this one and
327
00:21:54,560 --> 00:22:00,704
Increase it by 1.5% by say 1.2 and then 1.2
328
00:22:01,216 --> 00:22:02,240
And then 1.2
329
00:22:02,496 --> 00:22:07,360
Would we retain some of the original features some of the original logic
330
00:22:07,616 --> 00:22:11,712
Switch off the original image overall I think the original image is look quite beautiful
331
00:22:11,968 --> 00:22:14,272
But as we move down the scale here
332
00:22:14,528 --> 00:22:15,296
Play move down
333
00:22:15,808 --> 00:22:20,160
We begin to lose some of the coherence of the original image some of the
334
00:22:20,416 --> 00:22:26,560
Beauty of the original image and it becomes a bit more like almost a software is trying to find anything
335
00:22:26,816 --> 00:22:28,608
What they can use to fill the details
336
00:22:28,864 --> 00:22:29,632
In the image
337
00:22:29,888 --> 00:22:36,032
And it doesn't remain as coherent as it was before where is with this one with Ezra again it's actually
338
00:22:36,800 --> 00:22:41,664
I think still it retains the original kind of sense of what we had there
339
00:22:43,456 --> 00:22:47,296
Something which still looks similar to what we had originally
340
00:22:47,552 --> 00:22:50,624
And it is so fast every time I run a new one it is just
341
00:22:50,880 --> 00:22:52,160
Done very quickly
342
00:22:53,696 --> 00:22:59,840
You've got a number of options there and maybe what we can do is to give this to you maybe
343
00:23:00,096 --> 00:23:01,120
Play what I can do to give this
344
00:23:01,376 --> 00:23:02,656
This work Flow To You
345
00:23:02,912 --> 00:23:05,984
You can play around with the settings and see what the results you get
346
00:23:07,008 --> 00:23:11,104
I certainly don't want to give the impression that the latent
347
00:23:11,360 --> 00:23:14,688
Upscale is bad or something you shouldn't use
348
00:23:14,944 --> 00:23:17,248
I definitely don't want to give that impression
349
00:23:17,760 --> 00:23:23,904
Are there any situations where you definitely want to use it and it can give good results but I think that there's a little bit of food
350
00:23:24,160 --> 00:23:26,208
I thought for you right there and maybe
351
00:23:26,464 --> 00:23:31,328
You can experiment with many of there are literally dozens of
352
00:23:31,840 --> 00:23:37,984
Models that we can use for upscaling and they some of them are fantastic and give really good results so
353
00:23:38,240 --> 00:23:40,544
Let me try to experiment with the upscaling
354
00:23:40,800 --> 00:23:46,944
And I think it's an important part of getting professional-looking results because obviously the 512 by 512
29210
Can't find what you're looking for?
Get subtitles in any language from opensubtitles.com, and translate them here.