Full Transcript

·YouTLDR

FREE Claude Prompt That Creates VOX Style Videos AUTOMATICALLY | Here's How

18:41EnglishBy Jacksons AITranscribed Aug 4, 2026
Analyze another video with Pro30-day money-back guarantee
0:00

Now, there's this crazy new trend going

0:01

viral, like crazy, that features Vox

0:04

style animations. Those paper cut style

0:07

visuals used to tell documentary stories

0:10

and keep viewers hooked and glued to the

0:11

screen. And they're pulling in some

0:14

crazy numbers. So, I checked YouTube to

0:16

see if anyone had covered how to make

0:17

these with free tools. But all I could

0:20

find were complicated tutorials using

0:22

Claude code or expensive paid software.

0:25

Nothing beginner-friendly at all. So, I

0:27

did my own research and I found a way to

0:29

create these animations completely free

0:31

without paying for any software. So, in

0:33

this course, I'm going to walk you

0:35

through exactly how to create

0:36

high-quality Vox style animations from

0:38

scratch using 100% free AI tools. And

0:42

simple enough that any beginner can

0:44

follow along. I mean, check out this

0:45

cool animation I made in like 5 minutes

0:47

with this free strategy.

0:49

>> May, 1925.

0:52

Paris.

0:53

A well-dressed man reads a newspaper

0:55

story about the Eiffel Tower rusting and

0:58

costing a fortune to maintain. His name

1:01

is Victor Lustig.

1:03

He has an idea.

1:05

Days later, forged government stationery

1:08

in hand, he invites five scrap metal

1:10

dealers to a private suite at the Hotel

1:13

de Crillon.

1:14

He calls himself a deputy director of

1:17

the Ministry of Posts. The city, he

1:20

explains, can no longer afford the

1:22

tower. It will be sold for scrap.

1:25

Quietly, to avoid public outcry.

1:28

He picks the most eager dealer, a man

1:30

named Andre Poisson.

1:32

He takes the payment and a bribe on top

1:34

of it.

1:35

Then he boards a train to Vienna.

1:38

Poisson,

1:38

>> [music]

1:39

>> too ashamed to admit he was fooled,

1:41

never goes to the police. So, 1 month

1:44

later, Lustig returns to Paris.

1:47

He gathers a new group of dealers. He

1:50

sells the tower again.

1:51

>> All right. So, for this walk-through,

1:52

the very first thing you're going to

1:54

need is access to the master prompts

1:56

we're going to use to automate this

1:57

entire process. You should find that

1:59

either in the description or in my

2:01

completely free school community. Once

2:03

you've got the master prompts, the next

2:05

thing you need is a chatbot. Now, there

2:07

are a couple of options here. For the

2:08

completely free users, Deep Seek should

2:11

be your go-to. Deep Seek is a

2:12

Chinese-launched chatbot that doesn't

2:14

have a paywall. It's just like GPT, but

2:16

you don't have to pay for it. Another

2:18

good option is my personal favorite,

2:20

Claude. And the last one is the OG,

2:23

ChatGPT. You can pick any of these to

2:25

work with. They all excel in different

2:27

places. For this walkthrough, I'm going

2:29

to go with Claude, but you can still use

2:30

Deep Seek or GPT. They both work the

2:33

exact same way. So, I'm just going to

2:34

choose Claude, head back to my master

2:36

prompts, scroll down to the bottom, hit

2:39

command plus A to highlight everything,

2:41

then control plus C to copy it all. Once

2:45

I've copied it, I'll head back to

2:46

Claude, paste it in, and send it. Now,

2:49

once you send it, Claude thinks for a

2:50

bit and then asks you for the source

2:52

PDF, which is basically the content

2:54

engine we're going to use for this whole

2:56

process. [music]

2:56

To get the source PDF, just head back to

2:58

the master prompts. I stored everything

3:00

in there, and scroll down to the engine

3:02

prompt. Click on that. Once you do, it

3:05

takes you to an eight-page PDF on a

3:07

Google Drive. Just click download to

3:09

save it to your computer. Then, head

3:10

back to Claude, drag it into the prompt

3:12

box, and type engine attached. Once you

3:14

send that, Claude proceeds to stage one.

3:17

So, now Claude asks you what niche you'd

3:19

like to make these animations in.

3:21

There's history, there's money and

3:23

power, there's disaster and survival,

3:26

there's documentary, tech, and sports.

3:29

But, from what we've seen from Vox,

3:30

they're mostly make stuff around crime

3:32

and documentaries. So, I'm going to

3:34

choose number one, which is the default.

3:37

But, you could pick whatever style you

3:38

want. And here's the thing, even if the

3:40

style you want isn't listed here, you

3:42

can just type your own style and the

3:43

model will still proceed with the next

3:45

steps, which is creating video ideas for

3:48

whatever niche you chose. So, I chose

3:49

the the and documentary niche, and it

3:51

gave me 10 ideas to pick [music] from

3:53

for a specific crime-based video. I'm

3:56

just going to go through them real

3:57

quick. All right, for this one, I think

3:59

I like idea number four. The man who

4:01

sold the iPhone 12 twice.

4:04

Sounds interesting. So, I'm going to

4:05

copy that, paste it into my chatbot, and

4:08

send it.

4:09

Once I do that, the next thing Claude

4:11

does is ask me how long I'd like my

4:13

script to be. Now, you can choose

4:14

anywhere from 30 seconds to 20 minutes.

4:17

It just depends on the kind of video you

4:19

want to make.

4:20

For documentary videos, you should aim

4:22

for around 10 to 20 minutes. That's the

4:24

sweet spot for ad revenue. But, since

4:26

this is just a brief walk-through, I'm

4:27

going to go with the 1-minute video.

4:29

Once you choose that, the chatbot takes

4:31

in your request and gives you a precise

4:33

1-minute long script, along with the

4:35

exact word count it created. Now, you've

4:37

got your script. For the voiceover,

4:39

there are platforms like Eleven Labs and

4:41

a couple of other free options, but

4:43

we're going to get into that later on.

4:45

For now, let's just type next to move to

4:47

the next step. One thing to keep in

4:49

mind, depending on the chatbot you're

4:50

using, this can be a little different.

4:53

Claude is pretty strict. I used Deep

4:55

Seek before and typing something like

4:57

proceed would work fine for next. But,

5:00

with Claude, you'd have to type voice or

5:03

proceed for it to actually move on. Once

5:06

it does, it starts converting the script

5:08

it just gave you into visual beats that

5:10

we'll use to create the images for your

5:12

specific video. For our 1-minute script,

5:14

we ended up with about 27 beats based on

5:16

what Claude detected, which is actually

5:18

pretty accurate since this involves a

5:19

lot of movement. This looks pretty good.

5:22

You can ask it to give you a specific

5:24

number of beats, but I'm happy with

5:25

this. So, what I'm going to do now is

5:27

type next.

5:29

When I do that, the model starts

5:31

creating a very detailed TXT file that

5:34

we're going to use to create our images

5:35

automatically. All right, once you get

5:37

your TXT file from Claude, the next

5:39

thing you need to do is download it.

5:41

This part is really important because

5:42

we're going to use it to automate the

5:43

image creation. Depending on how long

5:45

your script is, you might need to

5:47

generate up to 50, 100, maybe even 200

5:50

images. With this TXT file, all you've

5:52

got to do is head over to a platform

5:54

called Google Flow. Click on create a

5:56

new project. Once you're in, click on

5:58

the video model at the bottom and make

6:00

sure you switch from the video models

6:02

over to the image [music] model. Google

6:03

Flow gives you access to both.

6:06

Once you've switched to the image model,

6:07

select the aspect ratio you want. For

6:10

this, I'm going with 16 by 9 because

6:12

we're creating our images in the

6:13

long-form aspect ratio. Then for the

6:15

model, I'm going to choose Nano Banana.

6:18

Now I need a way to batch produce images

6:20

here automatically. For that, you're

6:22

going to need an extension called Zappy

6:24

Flow made by a creator on here. Really

6:26

nice extension. Head over to the Chrome

6:28

Web Store and click add to Chrome. Once

6:30

it's downloaded, you should see a pop-up

6:32

appear on the right side of your screen.

6:35

Once you've got that pop-up, head over

6:36

to Google Flow and make sure it shows

6:38

connected to your Google Flow project.

6:39

Down here, you can customize how you

6:41

want the images to be created. The

6:42

default settings work just fine. The

6:44

only thing I'd say you should turn off

6:46

is include serial number in file name.

6:48

If that one's on, just turn it off. Now

6:50

what you've got to do is upload the TXT

6:52

file you got from Claude. Upload it

6:54

right in here. According to Claude,

6:55

we've got 27 beats, which means Claude

6:57

made 27 image prompts for us. And as you

7:00

can see, we've got 27 prompts here. What

7:02

I'm going to do now is scroll down and

7:03

click on start generating. And just like

7:06

that, the Zappy Flow extension starts

7:08

creating my images completely from

7:10

scratch.

7:11

I could literally just leave this here

7:12

and go do something else while the

7:14

extension automates the whole thing for

7:16

me. And the cool thing about this

7:18

extension is that once it creates your

7:20

image, it saves it straight to your

7:22

computer. So you've already got all your

7:24

images generated and saved

7:26

automatically, and you can just leave it

7:28

running and go do something completely

7:30

different. Pretty cool. All right, so

7:32

after a couple of minutes, we've now got

7:34

every single image we need for our

7:35

video. As you can see, the extension

7:37

generated all 27 images.

7:39

>> [music]

7:40

>> Not going to lie, this is really, really

7:42

good. Everything's generated in the

7:43

style we want and I'm genuinely blown

7:45

away. One thing I like is that if we

7:47

click on the downloads folder, we can

7:49

see every single image that's been

7:51

downloaded and saved to our computer,

7:53

which is really cool. Now that we've got

7:54

our images, we can close the extension.

7:57

Next, we need to animate these images by

7:59

turning them into videos. For this,

8:01

we're going to head back to Claude and

8:02

type next. Once you do, Claude starts

8:05

creating a very detailed universal video

8:08

generation prompt. Now, the default is

8:10

10 seconds, but here's what I recommend.

8:13

Ask Claude to make the duration less

8:15

than 10 seconds because we're planning

8:17

to use free software. For video

8:19

generation, the more time an image has

8:21

to animate, the more it costs. A

8:23

10-second animation

8:25

>> [music]

8:25

>> is going to cost more than a 5-second

8:27

one. So, what I like to do is head back

8:29

to Claude or whatever chatbot you're

8:31

using and ask, "Could you change the

8:34

duration to 5 seconds instead of 10?"

8:37

Then send it. When you do, Claude gives

8:39

you a new prompt. Of course, you can

8:41

still use the 10-second prompt if you're

8:43

planning to use a paid software, [music]

8:44

but for the free users, I think 5

8:46

seconds works just fine. All you got to

8:48

do is copy the entire video prompt. Once

8:50

you've copied it, head over to Flow,

8:52

click on the Nano Banana option, and

8:54

switch from the image model to the video

8:56

model. Then select one output. For the

8:59

duration, select about 6 seconds. Once

9:01

you do that, start dragging the images

9:03

into the prompt interface. Paste in the

9:05

prompt you got and send it. Flow then

9:07

starts creating your video from scratch.

9:09

Now, all you've got to do is repeat this

9:11

exact same process. Drag in the image,

9:13

paste in the prompt, send it, go back to

9:15

your chatbot, grab the next image, drag

9:17

it onto the prompt box, paste it in,

9:19

send it, and [music] just keep doing

9:20

this until you've dragged every single

9:22

one of your images through so we can

9:24

generate all our videos. Now, while Flow

9:26

is handling this, I like to compare my

9:28

results. The model we're using right now

9:30

on Flow is the Omni Flash model. So,

9:32

while it's generating the videos for us,

9:35

I want to try out SeaArt Dance and see

9:36

what it can do. I'm going to access

9:38

SeaDance through Higgsfield.

9:40

Full disclosure, [music] Higgsfield is a

9:42

paid platform, so that's something to

9:43

keep in the back of your mind. I'm going

9:44

to click on the video panel option once

9:46

I sign up. Then I'll click the upload

9:48

button on Higgsfield and upload the

9:50

images I want to animate. Now, ZappyFlow

9:52

saves all the assets it generated in a

9:55

folder. So, just head over to that

9:57

folder and upload the image you plan to

9:59

animate. Once you do that, select the

10:00

image so [music] it's added to the

10:01

prompt box. Then, paste in the prompt

10:03

Claude gave you. In the duration

10:05

section, reduce it to just 5 seconds and

10:07

make sure the aspect ratio is the exact

10:09

one you want. Set the resolution to 720p

10:13

so you don't waste credits. For this,

10:15

it's charging us just 23 credits, which

10:17

is fine. So, I'm going to click

10:18

generate. And while that one's

10:20

generating, I'm going to upload a couple

10:21

more just to compare the results with

10:23

what OmniFlash can do. All right, after

10:25

a couple of seconds, we've got our

10:27

animations from both Higgsfield and

10:29

OmniFlash. [music]

10:30

So, let's preview what we got. Let me

10:32

click on the first video to see what the

10:33

output looks like.

10:39

All right, this looks pretty decent.

10:41

Next one.

10:45

This looks pretty good. Matches the

10:47

animation style we're going for.

10:50

Let's preview what Flow was able to

10:52

create.

10:54

All right, honestly, the output from

10:55

OmniFlash looks a bit better to me. I

10:58

don't know what you think, but I like

11:00

the OmniFlash output. It looks much more

11:02

animated. Yeah, I think OmniFlash did a

11:04

wonderful job compared to the rest of

11:05

the models because it was even able to

11:07

create this scene that starts out with a

11:09

photo.

11:10

OmniFlash added in that photo of the

11:12

next soldier who wasn't even in the end

11:13

frame. It added it in on its own, which

11:16

is really, really cool. So, if I click

11:18

on it, everything comes up real quick

11:20

and looks real good. Yeah, it's really

11:22

cool. So, all I've got to do now is

11:24

repeat this process over and over. Drag

11:27

the next image onto the prompt box,

11:29

>> [music]

11:29

>> paste in my prompt, send it, and rinse

11:31

and repeat for every image until I've

11:33

turned every single frame into a video.

11:36

Shouldn't take very long. I think within

11:38

5 minutes you should be done with this.

11:40

All right, once you're done with the

11:41

video creation, the next thing you need

11:42

to do is generate your voice-overs. Now,

11:45

there are a couple of options. The first

11:47

is Microsoft Clipchamp. Clipchamp gives

11:49

you unlimited voice-over generations.

11:51

It's actually a video editor, but it's

11:53

got a really good AI voice generator

11:55

built in. So, if you want unlimited,

11:57

just install the software and you can

11:59

generate as many voice-overs [music] as

12:00

you want. Another good option is CapCut

12:02

TTS, which gives you access to over 200

12:05

AI voices inside CapCut. So, if you've

12:08

got the mobile or desktop version, you

12:10

can create really high-quality unlimited

12:12

AI audio. And for a more web-based

12:15

platform, you could use Noise AI. Noise

12:18

AI is pretty good. However, the quality

12:21

isn't the best.

12:22

>> [music]

12:22

>> It's not my go-to, but it does have some

12:24

good voices, and they give you about

12:26

2,000 free credits every single day. So,

12:28

those are my three free ish voice-over

12:30

options. Noise AI is limited since it

12:33

only gives you 2,000 credits, but

12:35

Clipchamp and CapCut are unlimited, so

12:37

check them out. But, my go-to, the one I

12:40

love using the most, is 11 Labs. 11 Labs

12:44

gives you high-quality AI voices that

12:46

even let you add emotional tags, which

12:49

means you can monetize the audio you get

12:51

from it. YouTube doesn't like glitchy or

12:53

weird robotic-sounding AI voices. So, if

12:56

you want to monetize your content, I'd

12:57

always advise you use 11 [music] Labs

13:00

because it'll stand the test of time.

13:02

So, just head over to 11 Labs, sign in

13:04

with your Google account, and they give

13:06

you 10,000 free credits a month. Once

13:08

you're in, click on the voices option to

13:09

pick whatever voice you want to use.

13:11

Now, what I like to do is add square

13:13

brackets and put in an emotional tag.

13:15

Let's say, whisper. Then, close the

13:18

bracket. It turns purple, which means

13:20

whatever I generate in there is going to

13:21

have that emotional tag influencing how

13:23

the voice-over sounds. So, I'm going to

13:25

head back to Claude,

13:27

scroll all the way up to where it gave

13:30

me my 1-minute script,

13:33

and copy just the first chunk. I'll

13:35

paste it into 11 Labs and generate it to

13:37

hear how it sounds. If it sounds good,

13:40

I'll download it, then head back to

13:41

Claude to copy the next chunk. And I'll

13:43

just keep copying chunk by chunk until

13:45

I've generated every single part of my

13:47

script and turned it into a clean,

13:49

high-quality sounding voice-over. Once

13:51

I'm done with that, I need a good

13:53

platform for music generation. For this

13:56

tutorial, I'm going to use Suno AI. Just

13:58

head over to your browser, search for

14:00

Suno AI, and click the first link you

14:02

see.

14:03

That'll take you to their website, which

14:05

looks something [music] like this. Suno

14:07

AI is an AI music generation platform.

14:09

You give it text, and it creates

14:11

high-quality, monetizable AI tracks that

14:14

you don't have to worry about copyright

14:15

issues with.

14:16

Once you sign up, click on create. Over

14:19

here, you can create any kind of song. I

14:22

always advise you toggle on the

14:23

instrumental button when you're

14:24

describing a background track. For me,

14:26

I'm going to type something like silent,

14:29

growing, intense,

14:32

driven,

14:33

blockbuster audio, and then just click

14:35

generate. [music]

14:36

Suno takes that simple prompt and

14:38

creates a specific background track for

14:40

me. The reason I'm going with this

14:42

specific track is because I want the

14:44

audio to be subtle in the background and

14:46

fit the style of video I'm making.

14:48

Whatever you're making, you can write a

14:49

prompt that suits it and generate your

14:52

track. And of course, you could use

14:53

Claude or ChatGPT to help you with the

14:55

prompt. It's all up to you. Once Suno

14:57

generates your track, all you've got to

14:59

do is click on the three dots at the

15:01

bottom left, then click download, and

15:04

download the MP3 so it saves to your

15:06

computer. Once you've done that, we now

15:08

need an editing software to compile

15:10

everything together into [music] our

15:12

final published video. For this, we're

15:14

going to use CapCut. It's very

15:16

beginner-friendly, so if you've never

15:17

used it before, don't worry. It's really

15:20

easy to understand. All you've got to do

15:21

is create a new project and import the

15:23

folder where you stored all your assets.

15:25

For me, I like to keep all my assets in

15:27

a single folder, my voiceover, my

15:29

background music, and all my footage. I

15:31

put it all in one folder, then click

15:33

import to add it to CapCut. Once the

15:36

folder's added, I'll first drag my

15:37

background track onto the timeline, crop

15:40

it, and reduce its volume so it's not

15:42

too loud. Then I'll use the magnifying

15:43

glass to stretch the timeline so I can

15:45

see clearly.

15:46

And I'll start dragging my voiceover

15:48

onto the timeline in the exact order I

15:50

generated it. Remember, we generated in

15:53

multiple batches. So, depending on how

15:56

long your script is, your voiceover

15:58

might be in a couple batches or more.

16:00

So, I'll align the voiceovers, then crop

16:02

the background track to make sure it's

16:03

lined up and synced to the voiceover.

16:06

Then I'll select both voiceovers by

16:08

holding command or control and clicking

16:10

on both, and I'll bump the volume up to

16:13

around 6.7 decibels. After that, I'll

16:16

start dragging in all my clips in the

16:17

exact order I generated them and just

16:19

make sure everything's synced up. All

16:21

you've got to do is drag the clips and

16:23

line them up properly. For some of them,

16:25

you might need to crop a bit, but I'm

16:26

pretty sure the animation software has

16:28

done [music] most of the work for you.

16:29

So, just make sure you crop the footage

16:31

to match the narration being said on

16:33

screen. Once all your clips are in and

16:36

synced properly to the voiceover, the

16:37

next thing to do is start adding

16:39

transitions between each clip. Now,

16:41

based on the kind of video we're making,

16:43

transitions need to be subtle or fit the

16:45

specific style. So, just go through all

16:47

the transition styles and pick whichever

16:49

works best. The camera one might look

16:51

pretty good. I'd suggest adding

16:53

different transitions in different

16:54

places so it doesn't get too repetitive.

16:57

Just make sure they fit the style or

16:58

animation you're going with.

17:00

For me, the ones I used the most were

17:02

the slide style transitions. [music]

17:05

So, add all your transitions, and if you

17:07

want captions, you can add those, too.

17:10

But I don't really like adding them for

17:11

videos like this, since the visuals

17:13

already have a lot of text pointers. So,

17:15

I'm going to preview it to see how it

17:17

looks and once it's perfect, I'll click

17:18

[music] export and export it in 1080p,

17:21

which is what I exported off flow. But,

17:23

if you exported your quality in 720p off

17:26

flow, then export that same 720p here.

17:29

You could use an upscaler to bump up the

17:31

quality later or even do it right here

17:33

in CapCut. But, that's pretty much it.

17:35

Check out how the video looks.

17:37

>> May 1925,

17:39

Paris.

17:41

A well-dressed man reads a newspaper

17:43

story about the Eiffel Tower rusting and

17:45

costing a fortune to maintain. His name

17:49

is Victor Lustig.

17:51

He has an idea.

17:53

Days later, forged government stationery

17:55

in hand, he invites five scrap metal

17:58

dealers to a private suite at the Hotel

18:00

de Crillon.

18:02

He calls himself a deputy director of

18:05

the Ministry of Posts. The city, he

18:08

explains, can no longer afford the

18:10

tower. It will be sold for scrap,

18:12

quietly, to avoid public outcry.

18:15

He picks the most eager dealer, a man

18:18

named Andre Poisson.

18:20

He takes the payment and a bribe on top

18:22

of it.

18:23

Then he boards a train to Vienna.

18:25

Poisson, too ashamed to admit he was

18:28

fooled, never goes to the police. So, 1

18:31

month later, Lustig returns to Paris.

18:35

He gathers a new group of dealers. He

18:37

sells the tower again.

Continue with YouTLDR

Analyze another video with Pro

Process a new video, search every timestamp, compare sources, and keep the result in your library.

Get Pro — $12/month30-day money-back guarantee

More transcripts

Explore other videos transcribed with YouTLDR.