Transcription
You're going to learn how to create cinematic 3D AI B-roll to take your YouTube videos, documentaries, and storytelling content to the next level. We're going to go through the entire video creation process to turn that AI B-roll into a full scene with ChatGpt, Vio 3.1, and 11 apps. And this custom B-roll style is used in the content production by some of the top YouTube creators, which get millions of views. We're going to recreate the same quality B-roll using AI.
So, the very first step is to head to this GPT, which is going to help you turn either prompts or images into the 3D cinematic scenes that you're looking for. And the link to this GPT is going to be in the description down below. Now, if you didn't want to use this specific GPT or you wanted to use another LLM, I'm going to put the prompt on screen right now. And all you have to do is to replace the criteria within the square brackets. And you could just actually ask an LLM to do this for you based on an image. But essentially, this custom GPT right here does all of this for you automatically. And a couple of extra things, too.
So, the very first step is to go ahead and upload a file. And here, I'm simply going to choose an image from my computer. And now you can be pretty flexible with the actual image you upload because we're going to create it in the exact cinematic 3D style we want in the aspect ratio we like. So I'm just going to upload this image right here of me talking to a crowd. And then I simply click open. And now all I'm going to do is actually click send here.
The GPT then gives us the main scene prompt which describes what it sees in the image and then also the B-roll prompts. So we've got a few different choices as to what kind of angle we want. Now, at the end, we're then asked which image we would like to generate. So, we can either choose the main scene or we can have the B-roll 1, 2, or 3. And so, this actually allows us to then get different shots and different angles within the same location of the same character.
Now, here I want to generate the main scene prompt. So, let's just do main scene because I want it to look very similar to the original image. I'm just going to type out main scene. Now, we're asked what aspect ratio we want for our content. And so, here I want to do this for YouTube. So, I'm simply going to type one, or you could just type YouTube. And now we're going to get a 16x9 image generated based off of the original image we gave it and from the custom prompt that I showed you earlier that has been ran through this GPT.
And so now, as you can see, the image has been generated. And if we put both of these side by side, you can see that it's actually a very accurate representation of the original image. The character's wearing a cap, holding the microphone. We've even got the laptop, which is on the side of the image. The AI even picked that up. And then we've got the brick wall, the curtains, and the plant in the background. Below it, you'll also notice that we're given a video prompt. And this is the video prompt that we're going to use to then bring this AI B-roll image to life. And so, this video prompt has been tailored to this image. And so, we're going to take this prompt, we're going to copy it, and then we're also going to save the image.
Then, we're going to head to Google Labs, and we're going to use V3.1. And we're going to go ahead and upload the image that we've just saved from chat GPT. And then, we simply paste in the prompt. And then we click send. And now our image is going to be generated and turned into cinematic AI B-roll.
Now, while that's loading, I do want to mention that you can use different AI models to get different results. Not only just with the video, but also at the image level. So, for an example, first of all, for the image here, I could use Clling's latest video 2.5 Turbo model, upload the same image, and then also paste in the prompt. And I've actually been using the DeepSeek feature, which actually rewrites the prompt for you, even more tailored to that image. And sometimes the results are actually really good. So let's go ahead and use one of these prompts and then we'll click generate and compare it with VO3.1.
But if we're not happy with the style of the image generated within chat GPT, we can actually go and use plenty of other AI image models to generate different looks because they all give different results with the same prompt. And so here I would go back to the GPT and I would copy the prompt that I wanted to generate. So, for example, let's just take the main scene prompt. And then I can paste it into a tool like Freepick and select one of the image models. And I've been getting some really nice results with Google Imaging 4. And so here I can click generate. And as you can see, the style of the image is completely different even though we didn't use the original image as a reference frame.
If we did want to use the original image as a reference frame, we could go ahead and use a model such as Cadream Creree 44K and then upload the image and then click generate. And as you can see, we've got four different images that are generated with the Cream 4K model. However, I prefer the style of Imagining 4 here because it's got this really nice 3D render look. So, if I wanted, I could download this image and then use this as the base image to then turn into B-roll.
And when it comes to creating multiple scenes here, we would want to switch to the model flux flux context max. And then here, we drag the image that we like. And then we can go back to GPT and we can use one of the B-roll angle. So here we could do this side angle one. So I would copy this prompt right here. Go back to Flux Context, paste the prompt, and then here we simply click generate. And I'm going to generate two at a time.
And now here you'll notice that the two images we've generated have the exact same backdrop. This one's a little bit different, but I want the backdrop to be different. And so I'm simply going to modify the prompt and I'm going to put windows in the background. And as you can see, now we've got more of a side angle. And you can keep on reprompting until you get exactly what you want. But I like this because there's continuity. Right here, we can see the side of the whiteboard or just the edge. And then at the top here, we can actually see more of the the whiteboard. And so it looks like we're in completely the same environment just from a different angle. So I'm also going to download this one.
And now let's head back and see what the original videos looked like. >> Understanding the data is only the first step. The true challenge lies in its application to the human element. >> And so this actually looks really good. It's very slow movement. It's kept the overall style and aesthetic of the 3D render throughout the entire video. Ignore the audio because we're going to use our own in 11 Labs. And so, as you can see, both of them have succeeded in keeping that realistic 3D render look throughout the entire video. And they both got slow and static movement. So, which one you use is really just a question of preference.
And now, both of these were generated with audio, but it's not audio we're going to use because we're going to be using 11 Labs. And now for the scene that I'm going to create, we're going to use the two images that were made using Imagin 4 and Flux Context. I run both of those through VO3.1 with the same video prompt that we got from chat GPT. And once I have all of my cinematic 3D AI B-roll generated with V3.1, I can then pull everything into 11 Labs so we can create the voice over and also add some music and potentially some sound effects to create the entire final video.
So in 11 Labs, we're simply going to click on studio. And here I'm going to click on new blank project and select video project. And here I simply want to upload the B-roll that we've just generated with video 3.1. So I'm going to select my first video, click open, and now it's being imported into 11 Labs. I'm then going to click on the video tab again and upload another video. Click on the second B-roll, click open, and now that is also imported into 11 Labs.
And so now if I press play, >> though my form is synthesized, my message is clear. >> We've got the AI B-roll from VO3.1, but there's audio with it. And we don't want to use the audio because the audio isn't that good or it's not the audio that we want to use within our project. And so here I'm going to click on the clip and we're simply going to mute it.
Now what I'm going to do here is I'm going to type out the narration or the voice over that I want said in my video. And so I prepared a little something earlier. And here I want to make sure that it's a good voice that's fitting to my story and my scene. And so here I'm going to go and click on the voice. And instead of using one of my voice clones, I want to find a different narrator voice. And so I might just type in narrator right here. And then we've got a few different narrator voices that appear. And so let's preview one. >> Calm, clear, articulate, every word delivered with purpose. >> I actually think this one could work quite nicely. So I'm just going to select it, click add to my voices. So it's added to my voice library. And now I'm simply going to click play. And then the voice over is going to be generated. >> As he began his speech, there was a shift in the room. You could feel it like the start of something important but fragile. He wasn't trying to impress anyone. He was just honest.
So here I think this is pretty good. However, the B-roll scenes last a little bit too long. So what I might do is potentially trim one of them down so it cuts to the next one earlier and maybe trim off the beginning so it lands on that video in motion. And obviously this is the tutorial, so I'm not going to go ahead and generate every single B-roll scene, but you saw that it was pretty easy to generate as many B-roll scenes as you need. And then you just simply import them back and forth.
And now I actually want to change this voice. So I'm just going to click on the voice again. I'm going to go back to narrator. And we're going to use one I already added to my voice library. Jim narrator. This one right here. And now the last thing that I want to do to bring the entire scene together is to add some background music. And so here within 11 Labs, I click on music. We can actually generate AI music simply by describing the song that we want to hear. And so here I'm going to type out ominous documentary music. I want to generate two variations of my prompt and I want it to be instrumental. And then this scene is quite short so I'm going to select 30 seconds for the duration. And then we click generate.
And now we have two songs that have been generated which we can preview and directly add to my project. And to add them, I'm simply going to place the playhead where I want the music to begin. And then I click on the plus button on the song that I want to add. So let's select this one right here. And as you can see, it's now been imported. And I can actually trim off the end here because it doesn't need to be as long. And I might also want to trim off the beginning just a little bit. So it starts a little bit quicker and there's no blank silence at the beginning.
And now if I click play, let's take a look. >> As he began his speech, there was a shift in the room >> here. The music's a little bit too loud. So, I'm just going to select it and then we turn it down a little bit. And so, we want this to be background music. So, let's just make it 32%. And then we click play again. >> As he began his speech, there was a shift in the room. You could feel it like the start of something important, but fragile. >>
And so, that is how to create cinematic 3D AI B-roll using VO3.1 and then compiling it all together into a scene using 11 Lab Studio. And this AI workflow allows you to turn anything into a really compelling story. And we're going to leave you on an example to finish up the video. Thanks for watching. >> As Alec began his speech, there was a shift in the room. You could feel it like the start of something important but fragile. It would have been his best one yet, except it wasn't. [Music]