Pratica di Shadowing: Is AI About to Change Movies Forever? - Impara a parlare inglese con i video

Scegli una modalità di pratica

Ogni modalità apre una sala di pratica a schermo intero.

257 frasi
Se le frasi sono troppo corte o troppo lunghe, clicca su Edit per modificarle.
1
Thirty years ago, Toy Story came out, and it took seven hours to render just a single frame.
2
Three years ago, House of Dragon came out, and it still took seven hours to render a single frame, even though computers are literally a thousand times faster now.
3
How come?
4
And how does AI do all of this in only seven seconds?
5
Well, by creating my own House of Dragon shop, I found the answer.
6
So join me on the journey.
7
In order to make a believable Dragon VFX video, we needed the right location, and Norway doesn't quite cut it.
8
So I boarded a plane to Iceland to do this the right way.
9
The problem was that there were too many good options to pick from.
10
This entire country looks like it was ruled by dragons at some point.
11
But eventually, we found this really awesome looking waterfall that we could actually walk behind, and I knew I'd found the perfect spot.
12
For the video, I'm gonna be jumping off this ledge, and a dragon will come from underneath, and I will land on it and fly through this waterfall.
13
Quite ambitious, I know.
14
Then, we will see me ride the dragon through this epic valley, which my friend Luke helped me film on his drone that makes mine look like a toy.
15
I just a baby!
16
With these shots done, and a bit more...
17
Sightseeing?
18
AAAAAAAAHHHHHHHHHHHHHHHHHH!
19
I headed home to start working on the VFX.
20
Okay, we need to pause the video a bit, because something has happened.
21
So, after I posted my last video, an AI company reached out wanting to sponsor me, And did they even watch my last video?
22
I think you people have watched an AI video
23
and felt like it was a piece of art that really touched or inspired them.
24
But they did say something that intrigued me.
25
We don't see AI as a replacement for VFX artists, no final cut on our side, no script, no talking points.
26
Which gave me an idea.
27
So I pitched them a real VFX vs AI competition, where the audience decides which one is best, and they have no say in the script whatsoever.
28
I never thought they would go for this, but I jumped on a call with them to discuss it, and secretly recorded it.
29
What would be almost fun would be a very like open
30
and honest collaboration where I'm like in the video I'm like this is sponsored by Hicksfield
31
but they gave me complete control to say whatever I want
32
because I'm sure you guys are used to some hits but
33
if you're I'm not sure
34
if you're used to the creator you're sponsoring being critical about AI as well
35
and if you're you're fine with that Definitely you'll be receiving some negative feedbacks
36
but I'm totally okay in this matter Well okay then!
37
I got this in a solid contract so there would be no take backs And the competition with Hicksfield was on.
38
The first thing we need for our VFX shot is a dragon.
39
And I'm not holding back for this competition, so I'm gonna use this dragon, which is probably the most detailed and high-quality Dragon 3D model available online.
40
And I happen to know the creators, and they happily sent it over for this competition.
41
We need some animation for it though, so I assembled the next part of my VFX Avengers team, Roan.
42
My trusted animator from previous projects will handle the complex flying animation needed for this shot.
43
And while he gets to work on that, I will tackle the waterfall.
44
The waterfall can be replicated by trading a sphere that emits particles outwards and adding gravity to it.
45
We then add some wind turbulence and random wind resistance per particle, however, it's still not looking quite right.
46
But if we calculate the density of the particles by seeing how close each particle is to surrounding particles
47
and make the dense clumps drop faster than the floating solar particles, we end up with a pretty believable waterfall without using a slow real water simulation.
48
You might be wondering why we even need to create a waterfall when we have a real waterfall in the shot.
49
He dumb?
50
But we need it for when the dragon flies through the waterfall so we can get some interaction between them.
51
Rowan is almost done working on the animation, but in the meantime, let's take a look at the opening shot.
52
The opening shot is this beautiful drone footage of the waterfall from above, But the problem is that the drone is moving backwards instead of forwards.
53
Usually, this would be an easy fix by putting the shot in reverse.
54
But this leads to the waterfall going the wrong way.
55
And this gave me an idea.
56
Because for this competition, I'll be sending Hicksfield all of my footage for them to create their version.
57
And I'm really curious how they're gonna fix this reverse waterfall.
58
I can fix it myself by tracking the shot, use that track to stabilize the shot, rotoscope out the waterfall, reverse it,
59
and then invert the stabilization to reapply the camera motion. And done.
60
However, I will definitely send Higgsfield the version of it going the wrong way to see how they deal with this dilemma.
61
Another thing I think AI will struggle with is close-up footage of a person with emotion.
62
I didn't film this in Iceland, but I want to make this as challenging as possible for AI, so I pull up a green screen and shoot footage of me starting to run towards the cliff.
63
I'll composite this later though because the dragon animation is finally ready.
64
I import the dragon animation to see how it's looking and it's absolutely perfect.
65
I also asked
66
if he could animate me jumping off the ledge since I didn't really feel like going for a swim that day.
67
And that animation is also looking so nice.
68
To start, I bring in the waterfall we made earlier and position it where the real waterfall is.
69
I then set up the dragon as a collision object and run the simulation again.
70
But the interaction doesn't really look that nice.
71
To fix this, I do a smoke simulation with the dragon as the source, which leads to a flowy wind field.
72
We can then plug this wind motion into our particle waterfall simulation, along with the collision we set up earlier, and that looks a lot better.
73
Now, it actually feels like the waterfall is wrapping around the dragon due to the air displacement.
74
To light the scene, I use this overcast image as a light source, and create a sphere with a hole in it to act as the cave.
75
I set this fake cave as a phantom object in the render, which means we won't see it in the render, but the light will still be affected by it, giving us this nice and realistic lighting.
76
I then give the waterfall a smoke shader, as this looks the most like the mist we see in a real waterfall.
77
Lastly, I use the same lighting for myself jumping off the cliff and hit render.
78
Now let's finally answer the question of why 30 years later it still takes hours to render each frame of a movie.
79
It all starts with looking at how they turned this into this.
80
Well, Toy Story used Reyes, render everything you ever saw, which broke the geometry into a ton of tiny pieces,
81
and for each piece read if the material was shiny or flat and what color it had.
82
It then used this to fake the look of the object.
83
It did this for every tiny piece, and finally added a pre-computed shadow map before the image was ready.
84
It's no wonder it took on average 7 hours to render each frame.
85
Nowadays, the process is completely different and almost simpler.
86
When we render, we run a light simulation
87
that bounces millions of lights around the scene to figure out
88
how the light would actually interact with it just like real life.
89
This is super computationally heavy though, which means even though we have computers that are thousands of times faster now, it still takes hours to render each frame of a movie.
90
Monsters University actually took 29 hours to render per frame with this new rendering technique.
91
Luckily, my render only took 12 hours, as I don't render at the insane quality that movies do.
92
But it still left me with this physically accurate dragon and waterfall.
93
I add the renders to my footage, paint out myself, and add fake me before the jump.
94
I then mask out the waterfall render for a smooth transition between the real and fake waterfall, and finally, I color match everything to the shot.
95
I found the waterfall interaction to be a bit underwhelming, so I went back into Houdini, added even more splash along the wings, and a bigger interaction overall,
96
and after rendering it again, I am super happy with the results.
97
Now, how in the world does AI do all of this in only a few seconds.
98
Well, it completely bypasses all of the old techniques
99
and uses a neural network trained on millions of videos to guess what the output should look like based on your prompt.
100
Is that enough to create a convincing dragon scene?
101
Well, it's time to find out.
102
Last video, you guys complained that I misrepresented AI by not utilizing it to its full potential.
103
Well, this time there will be no excuse because Higgsfield, one of the biggest AI companies, is going to be creating it.
104
They start by trying to run this shot through a video-to-video model using a prompt written by Claude
105
and a reference of Adele and the dragon.
106
But it fails spectacularly.
107
So instead they extract the first frame and use an image to video model to create the full motion from scratch.
108
Which gives them a jump they're happy with, but the dragon emerging doesn't look so good.
109
So they keep running the prompt until they get a version of the dragon that they like, and then finally stitch the shots together in post.
110
For my reverse waterfall challenge, they just use the first frame and prompt the motion with an image to video model, which fixes it instantly.
111
But I'm not too worried yet, as we still have the hardest shots coming up, and I will tackle them first.
112
The next shot is this one from the side, and finally we'll tackle this ending drone shot of the dragon flying towards us.
113
After a few days of animating, Rowan sent over this awesome animation, which has the dragon shake off the water from the waterfall.
114
So I set up a water simulation that activates based on the speed of the dragon, so when he shakes, even more water falls off.
115
I then convert the water particles to a mesh
116
and create some geometry to block the light from the mountain so we can render with accurate lighting.
117
For the final shot, I told him to go all out for the animation and he nailed it, with the dragon flying super close to the camera for an epic ending.
118
Here, I use the same lighting setup as the previous shot and render overnight.
119
To composite the side shot, I add the dragon and water, do some color matching, paint out some people, and finally add this mist asset to blend everything in even better.
120
And you can get this asset plus a bunch more on Maria and I's website, vfx.store.
121
And you can use the code AI30 for 30% off.
122
To be featured in my next video, create a VFX shot with an asset from the store and send it to create at vfx.store.
123
And so everyone has a chance to participate, we are actually making the HUD and UI pack free for a limited time.
124
Anyways, with the mist added, the shot is complete.
125
The ending shot was pretty similar to composite.
126
I just make sure to add some camera shake as it flies over and it's done.
127
Finally, let's composite this green screenshot.
128
I use Corridor Key to remove the green screen, shoutout Nico, and I think this is the best use of AI.
129
As a tool that doesn't take away creative control, but just speeds up the work.
130
I then add a cave background and do my best to make it look realistic, but it doesn't really look that good.
131
This is due to bad lighting when we film, and a lack of compositing skills from my end.
132
So I headed out to a real cave to film it there instead.
133
Sometimes you just can't fake reality, and I'll use real footage and practical assets whenever I can, But good luck doing the same, Mr. AI.
134
Hicksfield gives video to video another go for this shot, and at first it struggles, but eventually they end up with a version they're happy with.
135
They go through the same process for this one before getting a good version, which leaves only this one left.
136
But to make it fair, I also sent them what I shot in the real cave, and they have the audacity to use a video to video model to replace me with a deal.
137
Identity theft is not a joke, Jim.
138
And with that, both the VFX and AI versions are complete, so let's take a look.
139
Okay, are you ready to compare VFX versus AI finally?
140
Yes, I'm so so excited.
141
I've been waiting for weeks for this, so...
142
Okay, well...
143
Let's do it.
144
Let's start with my VFX shot.
145
Okay.
146
Oh, it's so epic!
147
Nooo!
148
Oh my god!
149
It looks so real!
150
That could have been Game of Thrones!
151
Maybe, no. Sick!
152
What?
153
I appreciate it.
154
No, it was really fun to work on this one.
155
The waterfall integration looks super realistic.
156
That took a long time.
157
You nailed that.
158
Oh, and the animation.
159
I had good help.
160
So like obviously I put it all together and did the lighting
161
and all that but like the animation and the model
162
and all the help I had with that from other like Really really good artist was yeah, it made this come to life.
163
Well, are you ready to see the AI then?
164
Oh, I don't know if I want to.
165
I have no idea what to expect for the Higgs field one Well, I've seen it once just to get it ready,
166
but I'm very very excited to see your reaction All right,
167
let's go Oh, it's a different guy!
168
We have been replaced.
169
Whoa, that's a big dragon.
170
Yeah.
171
And now it's super small.
172
Yeah.
173
I like the water touch.
174
Me too.
175
I want to go add that now.
176
You should actually, that's actually so good.
177
There are definitely some inconsistencies between the shots.
178
There's so many inconsistencies.
179
And between the dragon.
180
It doesn't really float together as well.
181
But the way he falls...
182
Whee!
183
That was kind of cool.
184
I like the way he falls.
185
But now it looks goofy, like he looks so big on the dragon.
186
Yeah, and here it looks more like a falcon, almost.
187
Yeah, yeah, that's true.
188
It doesn't get that epic really close to the camera.
189
But I like the replacement here.
190
That's pretty impressive.
191
Yeah, that's actually better than yours, to be honest.
192
Yeah, I know.
193
Who wins in your eyes?
194
Oh, definitely yours.
195
At least now.
196
Like, so far, yours.
197
Like, the consistency, the quality, the inconsistency definitely ruined it.
198
Well, there we go.
199
I guess the effects wins.
200
Okay, then.
201
That was ZZ.
202
We can finally move on with our lives and never think about AI again.
203
Is what I thought.
204
But Higgsfield wasn't ready to give up quite yet.
205
But I have a small surprise for you, because they actually sent over one more version.
206
No way!
207
What?
208
I didn't know that!
209
Yeah, yeah, no, where they just made their own original version of this, just completely without the footage.
210
Oh, okay.
211
Just to see how far they can push AI.
212
It's fully prompt?
213
Full prompt, and they're using Seedance 2.5 for this as well, so this is the best of the best.
214
So we'll see how it holds up.
215
Okay, you ready?
216
Mhm.
217
Okay, you can do this.
218
You just have to jump.
219
Just run and jump.
220
That's it.
221
A lot better acting than I did.
222
The sound design is really nice.
223
Yeah, yeah.
224
Oh, that's cool.
225
It's pretty epic.
226
Wow, they really...
227
It sounds like amazing.
228
So how do you feel now after watching that?
229
Well, it's pretty good.
230
You seem like reluctantly wanting to say that.
231
Yes, I agree.
232
I think it's a lot better.
233
And I think that's kind of, I guess, where AI can shine a little bit, where it can make everything from scratch.
234
But I still feel like you're missing that creative control.
235
Like when they're trying to make it match my vision of what I wanted to make, they couldn't do it.
236
And now they get a pretty epic video, but I'm sure they struggled controlling it exactly what they wanted what they wanted.
237
I'm sure they were just like, oh, this looks nice.
238
There's still inconsistency there.
239
Right, right.
240
And I also think they directed it quite well with him kind of getting hyped up before
241
and with the jump and the location and all that.
242
So maybe they're just better directors than I am.
243
But in terms of like matching my vision where I wanted what I shot in Iceland to have a dragon, the VFX one.
244
That didn't hold up, no. But it still has this AI look to it.
245
True.
246
There's still something that says, hmm, this is not human made.
247
Yeah, but I think it'll be really interesting to see what the audience says in terms of who they think wins.
248
So be unbiased and be honest
249
and like vote in the comments of who you think is better and who wins and why.
250
Like, why is one better than the other?
251
Oh, I'm excited to see and a bit scared.
252
Oh, this was fun and a bit sad.
253
But I want to bias you a little bit by ending the video with my VFX version and some epic human-made music.
254
Enjoy.
255
You've been keeping secrets?
256
I've kept one secret from you.
257
Yeah, I didn't want to tell you this.

Scarica l'app

Tutto ciò che serve per parlare fluentemente

Pronuncia con IAVoto per ogni frase
Pratica IPAPadroneggia ogni suono
VocabolarioCostruisci il tuo lessico
Gioco di vocabolarioImpara giocando
FOR YOU

Video consigliato

Vocabolario e note di pronuncia per questa lezione

Questo video contiene 257 frasi e 3106 parole da ripetere con lo shadowing. Il parlato dura 16:51. Chi parla va veloce, circa 184 parole al minuto: aspettati suoni legati e ridotti. Solo il 81% delle parole rientra nelle 3.000 più comuni dell’inglese, quindi il lessico è impegnativo.

Vocaboli chiave di questo video

15 parole del video che vale la pena imparare, con pronuncia e significato:

ParolaPronunciaSignificato
dragon sostantivo/ˈdɹæɡən/drago, dragone
waterfall sostantivo/ˈwɔːtəfɔːl/cascata, rapida
render verbo/ˈɹɛn.dɚ/rendere
animation sostantivo/ˌæn.ɪˈmeɪ.ʃən/ravvivamento, animazione
particle sostantivo/ˈpɑɹtək(ə)l/granello, pezzetto
epic sostantivo/ˈɛp.ɪk/epico
footage sostantivo/ˈfʊtɪd͡ʒ/materiale filmato
simulation sostantivo/ˌsɪm.jəˈleɪ.ʃn̩/simulazione
prompt verbo/pɹɑmpt/incitare, guidare
cave sostantivo/keɪv/caverna, grotta
reverse sostantivo/ɹɪˈvɝs/contrario, opposto
interaction sostantivo/ˌɪn.təˈɹæk.ʃən/interazione
drone sostantivo/dɹoʊn/fuco
shake verbo/ˈʃeɪk/scuotere, agitare
asset sostantivo/ˈæsɛt/bene, proprietà

I phrasal verb che sentirai

ParolaPronunciaSignificato
set up verbo/ˌsɛt ˈʌp/preparare, installare
hold up verboritardare, rallentare
come up verbovenire fuori, emergere
figure out verboscoprire, rendersi conto
find out verboscoprire
give up verboarrendersi
go through verboattraversare
reach out verboprotendersi

La grammatica di questo video

Le strutture che chi parla usa di più, con le parole esatte del video:

StrutturaNel video
Present perfect continuous have/has been + -ing — un’azione iniziata prima e ancora in corsoI've been waiting · You've been keeping
Forma passiva be + participio passato — conta ciò che accade, non chi lo fawas ruled · is sponsored · can be replicated
Present perfect have/has + participio passato — un’azione passata che conta ancora adessohas happened · have watched · I've seen
Frasi relative who / which + frase — un’informazione in più su una persona o una cosadragon, which is · render, which means · though, which means

Pronuncia a cui fare attenzione

Chi parla usa 61 contrazioni e forme ridotte, come I'm, doesn't, they're. Pronunciale nella forma breve, così come le senti.

  • I suoni “th”: smooth /smuːð/, thirty /ˈθɝti/, theft /θɛft/, underneath /ˌʌndɚˈniθ/
  • I suoni “sh” e “zh”: animation /ˌæn.ɪˈmeɪ.ʃən/, simulation /ˌsɪm.jəˈleɪ.ʃn̩/, interaction /ˌɪn.təˈɹæk.ʃən/, shake /ˈʃeɪk/, collision /kəˈlɪʒn̩/
  • Parole lunghe — attenzione all’accento: animation /ˌæn.ɪˈmeɪ.ʃən/, interaction /ˌɪn.təˈɹæk.ʃən/, geometry /d͡ʒiˈɑ.mə.tɹi/, participate /pɑːˈtɪs.ɪ.peɪt/, collaboration /kəˌlæbəˈɹeɪʃən/

Come esercitarsi con questo video

  1. Ascolta tutto il video una volta senza parlare e annota le parole che non conosci.
  2. Inizia a velocità 0,75×, fai shadowing frase per frase e torna alla velocità normale quando diventa facile.
  3. Registrati e confronta con l’originale, facendo attenzione a parole come dragon, waterfall, render.

Cos'è la tecnica dello Shadowing?

Shadowing è una tecnica di apprendimento delle lingue supportata da studi scientifici, originariamente sviluppata per la formazione dei traduttori professionisti e resa popolare dal poliglotta Dr. Alexander Arguelles. Il metodo è semplice ma potente: ascolti un audio in inglese di madrelingua e lo ripeti immediatamente ad alta voce — come un'ombra che segue il parlante con un ritardo di solo 1–2 secondi. A differenza dell'ascolto passivo o degli esercizi di grammatica, lo shadowing costringe il tuo cervello e i muscoli della bocca a elaborare e riprodurre simultaneamente i modelli di discorso reale. La ricerca dimostra che migliora significativamente la precisione della pronuncia, l'intonazione, il ritmo, il discorso connesso, la comprensione dell'ascolto e la fluidità del parlato — rendendolo uno dei metodi più efficaci per la preparazione alla prova di speaking dell'IELTS e per la comunicazione reale in inglese.

Tecnica dello shadowing: leggi la guida completa passo dopo passo →