Everybody is making tutorials on "how to use AI to make 300 videos in 30 minutes". But all of those AI videos are just terrible. I mean they don’t even get views! So in this video, I challenged myself to find a way to make a video with AI that’s actually good, and gets views. But throughout the challenge, things take an interesting turn...
If you want to learn how to use AI in your videos, or how to make videos with AI, I'm sure you're gonna get something from this one.
Hope you enjoy
If you want to learn how to use AI in your videos, or how to make videos with AI, I'm sure you're gonna get something from this one.
Hope you enjoy
Category
🛠️
LifestyleTranscript
00:00In this video, I'm going to teach you step-by-step how to create a YouTube automation video using
00:06our amazing AI tool that can generate 10 short videos every 30 seconds.
00:10Today, I'll be showing you how to recreate 180,000 videos in the fastest way to create
00:11a YouTube video.
00:12We're going to use two AI tools.
00:13I hate AI videos.
00:14Open any social media and start scrolling.
00:15Half of the videos you'll see will be short.
00:16Half of the videos you'll see will be long.
00:17Half of the videos you'll see will be long.
00:18Half of the videos you'll see will be short.
00:19Half of the videos you'll see will be long.
00:20Half of the videos you'll see will be short.
00:21Half of the videos you'll see will be long.
00:22Half of the videos you'll see will be short.
00:23Half of the videos you'll see will be long.
00:24I hate AI videos.
00:26Open any social media and start scrolling.
00:28Half of the videos you'll see are made with AI.
00:31An AI voice commenting on a funny video, an AI text that's showing a random motivational
00:35quote, AI image slideshow, a video with AI script, AI voiceover, AI captions, AI footage,
00:40AI music, AI...
00:41They're breaking the internet.
00:42And I'm not okay with it.
00:43Because they're soulless, boring, inconsistent.
00:44They make us dumb.
00:49So that they could take over the world?
00:51I don't know.
00:52Maybe.
00:53It's AI.
00:54You know, the biggest revolution in human history since the invention of Bitcoin.
00:59So what?
01:00Think about it.
01:01Isn't there a better way to use it?
01:02There has to be a better way to use it, right?
01:05We just need to find it.
01:06He was right.
01:08So I spent the next few months searching all over the internet to find a way to make better
01:12AI videos.
01:13And after dodging around 1,000 videos, teaching how to make 300 shorts in 30 minutes with
01:18ChatGPT and Canva, I think we got something.
01:28It turns out it's not a brand new thing.
01:31People have been making cinematic AI videos for the last few years, but they all have
01:35some big problems.
01:36The art style in each shot looks completely different.
01:39The characters don't look the same.
01:40The story is too AI generated.
01:42And most importantly, they don't get views.
01:45We need to fix them all.
01:49Here's the plan.
01:50I will try to create a video with AI completely from scratch, upload it to a new channel,
01:54and see how many views I can get.
02:04Before starting this video, I ask you guys for ideas.
02:06Here are all the cool video concepts that you came up with.
02:09Animated scenes with full emotions and peace.
02:12Add some emotions.
02:13Emotions.
02:14Emotions.
02:15Emotions.
02:16Emotions.
02:17Emotions.
02:18Emotions.
02:19Emotions.
02:20Emotions.
02:21Emotions.
02:22Emotions.
02:23Emotions.
02:24Emotions.
02:25Emotions.
02:26Emotions.
02:27Emotions.
02:28Emotions.
02:29Emotions.
02:30Emotions.
02:31Emotions.
02:32Emotions.
02:33Emotions.
02:34Emotions.
02:35Emotions.
02:36Emotions.
02:37Emotions.
02:38Emotions.
02:39Emotions.
02:40Emotions.
02:41Emotions.
02:42Emotions.
02:43Emotions.
02:45I wrote the story with chatGPT's ideas and tweaked some parts a little bit.
02:49It turned out pretty well.
02:50You can pause to read it, but we're gonna see it at the end anyways.
02:53So...
02:54Now, I already made a video about thumbnails.
02:57But since we're making a video with AI, we gotta make an AI thumbnail too.
03:01So, ladies and gentlemen, introducing the channel's very first sponsor.
03:07Pixels.
03:08It's an AI image generator that's trained specifically on YouTube thumbnails.
03:12So that with a single prompt, it can generate stuff like this.
03:15Or this.
03:16Or this.
03:17Alright, let's create our thumbnail.
03:19Our video is gonna be about time travel.
03:21So, create a dramatic thumbnail with turbulent background.
03:24Close up on a running bearded ginger man centered in the middle of the image.
03:27He has a slight smile.
03:28There are dinosaurs looks like.
03:29The background forest.
03:30And there's a lot of motion going on.
03:32Pretty impressive for 10 seconds of work.
03:34Even if you know how to create it in Photoshop, it would take hours to create something like
03:38this.
03:39But it doesn't stop here.
03:40Click this button and edit the thumbnail with your own words.
03:43Let's add some muscles to our guy.
03:45Ah, that's a bit too much.
03:47Perfect.
03:48If you have a face in the thumbnail, like this one, you can replace it with your own
03:52face by just clicking this button.
03:56You can even search for a thumbnail concept that already worked on YouTube and just paste
04:00the video link here.
04:01The AI will analyze it and come up with a slightly different concept.
04:05Pixels recently released a new feature for helping you generate titles as well.
04:09Just write what you have in mind and it will come up with really impressive title variations.
04:13They're constantly improving the algorithm with the new pixels 2.0 update.
04:17The results look 10 times more impressive.
04:20You can even add a custom text to the thumbnail, a feature that almost no other AI tool is
04:25able to do at this point.
04:26It has more features coming soon, like adding your own photos, even creating a thumbnail
04:29from your sketch.
04:31You can check out pixels with the link in description and with the code Isaac, you'll
04:34get a 50% discount for the first month or use Isaac year for 33% on the yearly plan.
04:39It's only valid for the first 200 people.
04:41So go use it while you still can.
04:44So thank you pixels for sponsoring this video.
04:47Now that we got our title and thumbnail, let's move on to the next step.
04:53Before doing anything else, we need to plan what we want to create.
04:56So I used my impressive drawing skills to make this thing called storyboard.
05:01Every professional filmmaker uses this technique to plan their shots before actually taking
05:06it.
05:07So why not make it for AI films?
05:08The key is that it has to be easy to understand.
05:10After a few hours of work, here's what they look like.
05:16There are dozens of AI video generators out there, but the most popular ones are called
05:20stable diffusion, picolabs, runway and Sora.
05:23The results from this one look great, but it's not released to the public yet.
05:26So we can't use it.
05:27Stable diffusion looks like a bit.
05:29It makes the videos using your computer's and the UI looks a bit okay.
05:33It's too complicated.
05:34These two are almost the same thing with only slight differences in style.
05:38Basically, runway is more cinematic while picolabs has more cartoonish vibe.
05:42I gave them the exact same prompt and these are the videos they made.
05:46I like the runway style, though you can try both and see which one works for you, but
05:50it's not important for now because after this video, you can use any software with the same
05:55principles.
05:56Instead of breaking down all the technical stuff about the software, we will just learn
05:59what we need and focus on turning the ideas into videos.
06:03As you might guess, you can generate videos in three different ways.
06:06From text, from an image, and from a video.
06:09We don't have a video yet, so...
06:11Text to video is nice for generating quick b-rolls for your videos, but it's hard to
06:15predict the results if you're looking for a consistent style and characters.
06:18So the plan is we will generate images using AI and animate it with another AI.
06:23It's gonna be a piece of cake.
06:24No, god, that doesn't make any sense.
06:28Shit!
06:33I'm gonna use Mid Journey here because everyone says it's the best image generator out there,
06:37but you can use any software you want.
06:40First, get a subscription.
06:41Go to Discord, create a new server, and add the Mid Journey bot to the server.
06:45That way you can see all the images you generate in a single place.
06:48Now it's time to see what this thing can do.
06:50Write slash imagine, and...
06:52Uh, I haven't used it for a while.
06:54Let me catch up with the updates.
06:57I read all the update notes they shared on their Discord.
07:00I know boring stuff.
07:01Then I watched a few hundred videos on how to use it.
07:03Here are some cool channels I found the most valuable information from.
07:06I tried to apply what I learned to the actual software, and to be honest, most of the images
07:11turned out bad, creepy, even kinda suspicious.
07:14But with practice, they started to get pretty good.
07:16And a few hundred attempts later, here's the strategy I came up with to generate the scenes.
07:23First, type slash settings, and choose Mid Journey version 6, or whatever is the latest
07:28version right now.
07:29Then experiment with the most basic prompts you can think of, and see what they do.
07:33At the end of each prompt, add dash dash AR 16 by 9 if you want to generate them in landscape
07:39mode, or dash dash AR 9 by 16 for shorts.
07:42Now, after each image, we get some buttons below.
07:45The ones at the top increase the resolution, and the bottom line creates slightly different
07:49variations.
07:50When you upscale the one you like, more buttons appear.
07:52Try them once, and you'll figure out what they do.
07:55Alright, these are a good starting point, but it isn't really possible to control much
07:59of the results this way.
08:01We need to get more specific.
08:03We can tell the AI where we want our character.
08:07Instead of asking it to generate an image of a bearded ginger man, we can add a background
08:11based on our story.
08:12Now let's add one more person to the shot.
08:15Tell where you want them to stand in the shot, and what they do.
08:21After getting the basis, the next step is adding more details.
08:24You can do it by adding some adjectives, and writing more detailed character descriptions.
08:29That was the last prompt we wrote.
08:31Let's see what else we can add.
08:32A charismatic 30 year old adventurer man is watching the Giza pyramid in ancient Egypt
08:37while his dog wanders around.
08:38The man has a lean and athletic body, wearing a light grey shirt with black cargo trousers.
08:42The Giza pyramid is made of glass.
08:45Alright, let that sink in.
08:46We can add much more details to our prompts.
08:51Like time of the day, season, color scheme, and so on.
08:56Most of the characters made by Mid Journey lack emotion in their faces.
09:00I mean, is this guy happy?
09:02Confused?
09:03Confident?
09:04None of them.
09:05He just exists.
09:06So it's always nice to specify the character's emotion in your prompt.
09:10Happy?
09:11Sad?
09:12Scared?
09:13Angry?
09:14Confused?
09:15Disgusted?
09:16Jealous?
09:17Guilty?
09:18Curious?
09:19You know what I mean?
09:20It doesn't even matter.
09:21We can get a similar effect by writing some keywords to our prompt.
09:24I like this one.
09:25It's all up to what style you're looking for.
09:27Speaking of styles, you've probably noticed almost every single image has a different style.
09:31Some are realistic.
09:32Others are more animation-like or illustration.
09:34It's cool.
09:35But when we animate these, we can't really make them look like a part of a single video.
09:39So we need to give them a single style.
09:43AI image generators usually make photos in between realistic and animated looks.
09:48I don't know what it's called, but it kind of became the signature of AI.
09:51If you want to avoid this style, you may go towards either side.
09:54Or try totally different art styles to get something unique.
09:57First, let's go with the realistic.
09:59You can add a movie genre to give it a certain vibe.
10:01Fantasy?
10:02Sci-fi?
10:03Action?
10:04Comedy?
10:05I don't know.
10:06Or a director name that has his own style.
10:07Like Wes Anderson, who's known for his symmetric shots and weird colors.
10:09Tim Burton?
10:10Edgar Wright?
10:11James Cameron?
10:12Spielberg?
10:13Or any director you like.
10:14I don't know much about cameras, but I saw most people like to add a camera type in their prompts like that.
10:17Or we can go the opposite way and get more animated style with words like Pixar, animation, or 3D.
10:22Other styles can be oil paintings, watercolors, anime, comics, 2D cartoon, surreal, Van Gogh, cubism, cyberpunk, renaissance, pixel art, sketch, Japanese art.
10:31I don't know.
10:32There are countless art styles you can choose from.
10:34Pick the one you like and add it to your prompt.
10:36But wait.
10:37It's still AI.
10:38And AI is not consistent.
10:40Even if you write the exact same prompt, it will give slightly different styles each time.
10:45But don't worry.
10:46There's a solution for that.
10:50It turns out you can give Mid Journey a reference image to look for.
10:53So it would generate all other images with a similar style.
10:56Now I want you to give your full attention to this point.
10:59Because we're about to get a bit technical.
11:01Hey!
11:02Turn off the music.
11:03Alright, you got this.
11:04You got this.
11:05Let's take this shot.
11:06Upscale.
11:07Upscale again.
11:08Open in browser.
11:09Download.
11:10Upload to Discord.
11:11Now write your prompt as usual.
11:13Then just like ____AR16x9, write ____SREF.
11:18Hit space and drop your image next to it.
11:21That was a style reference.
11:22It will copy its style to the new image.
11:24We need to learn a few more parameters.
11:26____S is called stylize value.
11:29It shows how much Mid Journey can add its imagination to your image.
11:32The default is 100.
11:34And the maximum value is a 1000.
11:36Let's try something like 500.
11:39____C, chaos value.
11:41It's like the unpredictability of the results.
11:43Zero makes them almost the same.
11:45And the shots get different as long as you increase this value up to 100.
11:49Style raw minimizes the improvisation and only follows your prompt.
11:54With all this knowledge, hopefully you can get any image in your mind with the best accuracy.
11:59But the biggest problem is still there since the very beginning.
12:03This is not the same guy.
12:04If we want to make a video out of these, we need a consistent character throughout all the results.
12:10Yeah, that's the hard part.
12:12What's your plan?
12:14Hey, you have a plan, right?
12:17Look, I'm gonna be honest.
12:18I watched all the tutorials about that.
12:20But none of them gets the perfect result.
12:23The images always turn out slightly different each time.
12:26And the process is too complex.
12:29I don't want that.
12:31I actually don't know what to do at this point.
12:34Don't worry, man.
12:35There's still time.
12:36You can figure it out.
12:37Can I?
12:38Yeah, I can.
12:42I couldn't figure it out.
12:50Maybe I should move on to another tutorial.
12:53A tutorial about how I edit.
12:56Nah, nobody would watch that.
12:59Having a bad day?
13:01Yeah, I think we had this conversation before.
13:03Don't even start.
13:04Yeah, I'm not here to give you a motivational speech this time.
13:07Just wanted to know how's your AI project going.
13:10Not good, man.
13:11It's clear it's not gonna work out.
13:13It's not the end of the world if I'm not gonna check off everything on the list, right?
13:17I'll just accept and move on.
13:19Hey, hey, hey, no.
13:20Look, you have to finish this video.
13:21You hear me?
13:22Otherwise...
13:23Why is it that important for you anyways?
13:25That's not the point.
13:26If you were gonna quit, why did you come this far?
13:28I don't know.
13:29You don't know?
13:30I know what you want to hear.
13:32What?
13:33So you're saying, I can't figure out this thing that even idiots are able to do.
13:36Is that right?
13:37What did you say?
13:38You're saying, I'm teaching people how to make videos while you're too afraid to finish a damn video yourself.
13:42Are you really that weak?
13:44So the tools people are playing like a toy is too complicated for you.
13:47Because you are a man with lower capacity than them, huh?
13:50I'm not.
13:51You know, Dave, this idiot made an AI video in one week.
13:54He probably got some help though.
13:56Doesn't matter. Dave is an idiot.
13:57Can you do what he did?
13:59I can.
14:00I didn't hear.
14:01I can.
14:02Louder.
14:03I can.
14:05We can finally create consistent characters inside of Midjourney.
14:08Consistent pictures of your character created in any scene you want.
14:13The best part is, we'll be able to keep the same face, hair, clothing, and even body type.
14:19That's it.
14:20That's what I've been looking for.
14:22I'm glad then.
14:23What are you waiting for?
14:24Get back to work.
14:25Yeah.
14:26Let me see how this thing works.
14:27Good luck.
14:28That was close.
14:29Hey.
14:30Thanks for coming, man.
14:31No problem, man.
14:32You do the exact same thing.
14:33Sure.
14:39It's time to finish what we started.
14:46So with this new feature, just like dash dash S ref, we can give a reference character to our prompt by typing dash dash C ref.
14:53So it can generate scenes with the same character every time.
14:56First, we need a reference character to use in all our images.
15:00To do that, generate your character based on what you learned until this point.
15:04A like 90s anime style portrait of a 30-year-old bearded ginger man with green eyes.
15:09The man is wearing a light gray shirt.
15:11Here's the crucial part.
15:12Split into four images shot from different angles.
15:14It will give you something like this.
15:16Upscale the one you like, take screenshots of each angle, and upload it to Discord.
15:21Now, it gets a bit technical.
15:22Stay with me.
15:23We can still use C ref command and paste each image link one by one.
15:27But doing that for every single image can be a bit annoying if you're going to do it a hundred times.
15:32So we're going to create our own command.
15:34Sounds complicated, but all we're going to do is instead of C ref and links next to it, we will create a custom keyword.
15:40So that when we write it, it will understand we mean this command.
15:43Write slash prefer option set.
15:46Inside the option box, write your character's name.
15:48Click plus one and select value.
15:50Here, write the command dash dash C ref and paste your reference image links.
15:55Now, each time I write dash dash Isaac, it will create the image based on these images.
16:03Finally!
16:04Of course, you don't need to use all of these in all your prompts.
16:07But by using as much detail as possible, you'll get closer to the image in your head.
16:11Which I guess is the main purpose of AI tools.
16:14Isn't it?
16:15Anyways, I repeated the same process for the rest of the shot.
16:18And here are the images we're going to use.
16:20Yeah, you better be watching this, man.
16:22I got a pterodactyl for you.
16:24Now, we need to turn these into actual videos.
16:26Uh-oh, we're a bit behind the schedule.
16:30The video part turned out to be easier than I expected.
16:33As I said in the beginning, I used Runway for that part.
16:36It gives you some free credits, but you're probably going to need a subscription after some point.
16:41Anyways, here's what I did.
16:43This is the main page when you open up your account.
16:45Click on generate videos and choose Gen 2, image to video.
16:49After uploading the image you generated in Mid Journey, write your prompt.
16:54I saw it works best if you write very simple prompts like the man looking up.
16:58Because we already did most of the heavy lifting with images.
17:02Here we need to adjust the camera motion and motion brush tools based on our scene.
17:07Camera motion is, well, controls the motion of the whole shot.
17:10Tweak them slightly based on your image and move on to the motion brush tool.
17:14This is where the magic happens.
17:15Choose the specific parts you want to move and brush them with different colors.
17:19Like head, hands, the entire character, dog, sky.
17:22And building.
17:23Then use these sliders to decide which direction you want them to move.
17:27Again, don't go too extreme with these or you might get some extreme results.
17:31Before hitting generate, tick this checkbox to get the final video with higher FPS.
17:36And...
17:40I mean it's not bad at all for a raw video, but it's going to get better.
17:43I repeated the same process for the rest of the images.
17:46A quick note from the future.
17:47I actually tried animating the rest of the shots, but even though the first few videos were quite nice,
17:52as soon as I tried some illustration style images...
17:55Yeah.
17:56I mean...
17:57What's that?
17:58However, when I tried the same thing with PikaLabs, it handled it way better.
18:02The settings here are almost the same as Runway, so the learning curve is a few minutes.
18:05So, realistic or 3D images for Runway, cartoon or anime style for PikaLabs.
18:11Now we can continue.
18:13And lastly, I used 11 laps to get the voiceovers needed for the story.
18:16Just like the voice you're listening to right now.
18:18And yeah, if you didn't realize it's AI until this point, welcome to the club.
18:22There's a link in the description if you want to try it yourself.
18:25Now, we need to edit them together.
18:32I will make a full guide on editing soon, so it's just a quick overview of what I did to edit this video.
18:38Step 1. Put the voiceovers in order and cut the pauses.
18:41Step 2. Find a nice music that fits in the story and add it to the timeline.
18:45Step 3. Add the AI-generated video clips based on the voiceover and music.
18:50Try to keep the flow between music and the visuals.
18:53Step 4. Add zooms, camera movements and transitions if needed.
18:57Step 5. Add extra details like smoke, dust or lighting overlays.
19:02Step 6. Add some sound effects to important points.
19:05If you're new to this, just use a popular sound effect pack from YouTube.
19:09Or if you need something more specific, again, search for it on YouTube.
19:12Step 7. Cover everything with an adjustment layer and add the Lumetri Color effect on it.
19:16Play with the colors to get the vibe you want.
19:19Step 8. If you want more cinematic look, add a crop effect on the adjustment layer and increase the top and bottom values.
19:26Step 9. Export the final video.
19:29It's show time.
19:32It's been one week since I released the video, but before showing the results, we need to talk.
19:37So the AI video and all that is cool.
19:39But the real question is, where are you going to use these?
19:42Is making story videos like these a good idea?
19:45Probably not.
19:47Wait, so all that stuff was for... nothing?
19:51Hey, did you really pay attention to the entire video?
20:21Thanks for watching.