FREE & UNLIMITED | How I Make VIRAL 3D Documentary Videos 100% With AI

0:00This YouTube channel called Fern has generated around 20 million views in just the last month alone. This other channel pulled 2.3 million views in only 7 days, and this one brought in over 1 million views in just the past week. And what they all post is literally the exact same thing. 3D mannequin style animation used to tell documentary stories. And judging by the type of audience this niche attracts, we're looking at a healthy RPM of at least $6 per thousand views. Now, if you multiply that RPM by 20 million views, that's an absolutely insane amount of revenue. And I'm pretty sure you've already come across videos like these and gotten

0:38completely glued to the screen because of how clean the animation looks. Now, in the past I've covered how to make animations like this, and the response was insane. [music] But a lot of you kept asking for a more in-depth course that really breaks down the entire process, from branding your channel to writing your scripts, generating your footage, adding the voiceover, creating your thumbnails, and editing everything together so it actually looks professional. So, in this video, that's exactly what I'm going to do. And to make things even better, we're using completely free tools. So, anyone, anywhere, can follow along and try this right now. Plus, somewhere in

1:18this tutorial, I'm going to share a really important tip that can help protect your channel from getting demonetized. And just to show you what's possible, take a look at this animation I made in like 15 minutes using the exact methods I'm about to show you. It's the 4th of March, 2023. A marble hall in downtown Dubai.

1:37A man walks onto the stage. He clicks a remote. The lights dim. Behind him, something appears on the screen. A ring of glass rooms. But they're not in the sky. [music] They're not on land. They're under the sea. So, why is one of the richest cities on Earth building down instead of up? Here's the thing. Dubai is running out

2:03of room. The towers can't climb any higher. All right, so the very first thing we need to do here is find the brain of the entire operation. Basically, you need a chatbot. Now, if you're trying to do this completely for free, you can use DeepSeek. It's open source, completely free, and the cool part is it doesn't really have a paywall. So, you've got unlimited chat tokens and can pretty much chat as much as you want. Now, if you want another option, you can also use ChatGPT by OpenAI, which I'm pretty sure most of you already know about. But for this course specifically, we're going to use Claude from Anthropic. So,

2:36just go ahead and sign up for Claude. The free plan works totally fine for what we're doing here. Now, once you pick the brain for your operation, the next thing you need is access [music] to the master prompt. You should find that either down in the description or inside my completely free community. So, here's what you're going to do. Scroll down to

2:53the section with the master prompt. Highlight [music] everything there and copy the whole thing. Then, head back to Claude and paste it in. Now, once you've done that, all you've got to do is choose which [music] Claude model you want to use. Different models use different amounts of credits, but right now I'm on Opus 4.7, so I'm just going to send the prompt. All right, so once you've done that, Claude is going to ask you for the source PDF that's needed to continue the process. [music] So, here's what you want to do. Head back to the master prompt and scroll all the way up to the top. see a section called source

3:24PDF. Now, if you click on that, it's going to take you to a Google Drive where you can download a full 45-page PDF that basically contains every single thing you need to understand this niche properly. So, just go ahead and download the file. Once you've got it, head back to Claude, upload the PDF into the chat, and then just type something simple like source PDF attached. Now, once you send that message, Claude is going to absorb the PDF, go through all the information inside it, and then tell you that it understands the task and is ready to continue. At that point, all you've got to do is type go. Now, when you type go,

3:59Claude starts generating branding recommendations for [music] your channel. So, it'll give you things like channel name ideas, a catchy description for your channel, logo prompts, banner prompts, basically everything you need to automate your branding. We're going to come back to that part later in the tutorial, but for now, let's keep moving and get into the next step. So, at this point, all I'm going to do is type next and hit send. Now, once you send that, Claude starts giving you a bunch of video title ideas to choose from. And if you look at the kind of topics it's generating, you'll notice they're actually really similar to the stuff you

4:30normally see on the model and reference channels. It follows that same documentary style theme they usually post. But here's the cool part. These are completely fresh ideas, so you're not just ripping content from other channels, which seriously helps when it comes to avoiding reused content issues and demonetization. What Claude basically does is absorb the source PDF, understand the style of content you're trying to make, and then come up with brand new ideas and concepts for you automatically. So, once you've got a bunch of topic ideas in front of you, the only thing left to do is just pick one you want to work on. All right, so

5:04for me, I think I'm going to go with idea number four, which talks about why Dubai is building an underwater city. So, I'm just going to type four and send that back to Claude. Now, once you do that, Claude is going to ask you how long you want the script to be. So, it'll ask if you want something like 1 minute, 10 minutes, 20 minutes, stuff like that. And all you've got to do is type in whatever duration you want your video to be. Now, one thing to keep in mind is that the duration you choose here directly affects how long the script becomes. So, obviously, a longer

5:33video means a longer script with more detail. For me, since this is just a quick tutorial, I'm going to keep it short and say I want the video to be around 1 minute and 30 seconds long. So, I'll type that in and send it. Now, once I send that, Claude basically figures out how many words are needed for that exact duration and then generates a script that properly fits inside that time frame. All right, so once Claude gives you your script, the next thing we need to do is generate the voiceover for it. Now, there are a few different platforms you can use for this part. For

6:03example, if you want something completely free with a pretty decent amount of credits, you can check out Noise AI. It gives you around 2,000 free credits every month, which honestly isn't bad. But here's the thing, there is a trade-off. If you really want those super clean, high-quality voices that sound premium, then the best route is probably ElevenLabs. ElevenLabs gives you around 10,000 free credits every month. And that's usually enough for maybe two to three videos, depending on how long your videos are. But the biggest difference is the quality. The voices sound way more realistic and polished compared to Noise AI. Now, Noise AI is still usable. The voices are

6:38okay, but if you want that extra quality boost, we're going to go with ElevenLabs for this tutorial. So just head over to ElevenLabs, log in, and then click on the text-to-speech option. Now, the reason I decided to use ElevenLabs for this is because it lets you add these emotional pointers that tell the AI how the voice should sound. [music] So if you're using the V3 model, which is their newest model right now, you get access to really cool stuff like making the voice laugh, whisper, sound dramatic, or deliver certain parts in a specific [music] tone. So for this tutorial, I'm just going to pick a voice

7:12here. Let's say I go with Genette's voice. Then I'll head back to Claude and start working through the script. Now, one thing you really want to keep in mind when generating voiceovers is this: always generate in batches. So instead of copying the entire script at once, what I'm going to do is copy just the first chunk of the script, head over to ElevenLabs, paste that section in, and generate only that chunk first. Trust me on this, because if you generate the whole script in one go, ElevenLabs can start sounding glitchy or distorted in certain parts. So, if you want that cleaner, premium quality sound, it's way

7:46better to work in smaller batches. Now, another really cool feature is that you can add emotional toggles using square brackets on your keyboard. So, let's say I want a line to sound like a whisper. I'll just type whisper and then close the bracket. Once you do that, you'll notice that section of text turns purple. And what that basically means is Eleven Labs now understands that this specific part should be whispered [music] when the voiceover is generated, which is honestly super cool. So, let me preview it real quick so you can hear how it sounds. It's the 4th of March,

8:182023. In a marble-floored auditorium in downtown Dubai, a developer steps onto a stage. Now, as you can hear, that actually sounds really, really good. So, if I like the output, all I've got to do is download it. Then, I'll go back to my chatbot, copy the next chunk of the script, head back into Eleven Labs, paste that section in, maybe change the emotion to something different, and then generate that chunk, too. And honestly, I'm just going to keep repeating that same process again and again until I've generated every single part of the voiceover for my video. Now, once I'm done with all the voiceover generation,

8:54I'll head back to Claude and type next. [music] Now, what Claude does after this is really cool. It starts giving me super detailed character prompts for the entire video. And these prompts are what we're going to use with an image generator to actually bring the characters from our script to life. So, the next thing we need is a solid image generator. Now, I know a lot of you guys in the comments have been asking for a completely free tool that can do this

9:18properly. [music] And after spending weeks testing different platforms, one of the best free options I found is Meta AI from Meta. It gives really high-quality image outputs, [music] and surprisingly, there's very little quality loss for something that's completely free. So, all I'm going to do is open my browser, search for Meta AI, and then sign in either with my Google account, Instagram account, or Facebook account. Once you log in, you should land on a page that looks something like this. Now, the really great thing about Meta AI is that it gives you unlimited generations for both images and videos, which honestly makes it one of the best

9:55completely free AI models available right now. That said, I still want to give you an alternative, and that's Google Flow. Now, with Google Flow, you get access to the Nano Banana model along with unlimited image generations as well. So, once you sign up for Google Flow and click on the new project option, you should land on an interface that looks something like this. And as you can see, the Nano Banana model is already selected for the image generation phase, and it costs zero credits. So, both Meta AI and Google Flow give you unlimited image generations, which is honestly amazing

10:30for free tools. Now, let's head back to Claude and copy the first character prompt it generated for us. Then I'll paste that prompt into Meta AI, toggle on the create image option, and send the prompt. And just to make things a little more interesting, I'm going to do the exact same thing inside Google Flow as well, just to compare the outputs and see which one gives us the better result. Now, after a couple of seconds, we get images that

10:54look like this. [music] And as you can see, this actually fits the animation style of the niche we're trying to get into. Both Meta AI and Google Flow did a really good job here. The only difference is Meta AI generated the image in more [music] of a box aspect ratio instead of that wide landscape look. Now, one thing I've noticed that really helps when using Meta AI is actually specifying the

11:15aspect ratio you want in your prompt. Because Meta doesn't really give you much control over things like aspect ratio or design settings, so you have to be very clear with what you want. But the upside is it still gives really high-quality images if your prompt is specific enough. Now that I've got my first batch of characters, I'm going to go back to Google Flow and grab the character prompt from Claude. Then I'll paste that same prompt into both Flow and Meta AI, and just hit generate again. All right, so one thing I've noticed with Meta AI is that sometimes you drop in a prompt and it just refuses

11:46to generate the image on the first try. So you kind of have to retry it a couple of times before you actually get a solid result. But with Google Flow, it feels way smoother. It just takes the prompt and runs with it almost every time without issues. And since both of them give you unlimited credits anyway, and Flow is already running Nano Banana as its base model, honestly, I think it just makes more sense to stick with Flow for this whole process. But if you still want to use Meta for whatever reason, that's totally fine, too. It just needs a bit more patience sometimes. Now, what

12:16I'm going to do next is head back to Claude and grab the prompt for the next character. Then I'll just paste it into Flow and generate the next image. And actually, if you go back to Meta AI and open a new chat, you'll see it also has a create video option, which we're going to test later in this course. But for now, let's just focus on getting all our images done first. Right now, Google Flow has basically given me all the images I need, so I'm just going to go ahead and download everything. Once that's done, I'll head back to Claude and type next. Now, the next thing

12:45Claude asks you is whether you have a visual flow in mind. And what that just means is, do you already have a sequence in your head for how you want the video to play out? So if you read your script properly, you can kind of visualize it. Like maybe it starts on a city street at night. The camera slowly zooms in, then it cuts to someone walking out of a gym, then it cuts to a person on a laptop, then it jumps back to the street, then

13:08maybe transitions into a bakery scene. Basically, whatever order you imagine those scenes happening in, that's your visual flow. It's just the story sequence of your video. Now, if you already have one in mind, that's actually the best case because you get full control over how your video turns out. But if you don't have anything planned, that's fine, too. You can just let your chatbot design it for you. So, for me, I'm just going to say, "Design it for me." But, if I wanted to, I could still give my own visual flow for even

13:35better control over the final result. Now, what happens next is your chatbot starts breaking everything down into multiple prompts in chapters. And these chapter-based prompts are actually designed for multi-shot generation. The reason it does this is because most AI video models have strict duration limits. For example, SeeDance usually has around a 15-second limit, while models like VO3 can go as low as an

13:598-second limit. So, instead of trying to force everything into one long video, the workflow splits your story into structured sections. Now, once you upload your reference character images, the ones we just created, and then paste in the prompt from chapter one into any of these models, what it does is generate a single shot that actually contains multiple sequences blended together inside that short duration. And that's really the power of this kind of advanced prompting. You're not just generating random clips, you're controlling structured cinematic moments inside each generation. So far, we've ended up with around six different chapters, which is actually really solid for building a full storyline. Now, the

14:39next step is pretty straightforward. All we need to do is take these chapter prompts and plug them into a video {slash} image-to-video model, so we can generate the multi-shot sequences we need for the final animation. Now, the two models we're going to use side by side are Meta AI and Google Flow. And what we're going to do is take the exact same prompt, use the same parameters,

15:01and then compare the outputs directly. So, the first thing I'm going to do is scroll all the way up to chapter one, copy everything from that section, then head over to Meta AI, paste it in, and upload the reference mannequin image we already created. Now, one important thing to keep in mind is that this prompting technique is specifically

15:19designed for multi-shot generation. [music] So, we need a model that can actually handle multiple sequences inside a single output without breaking the flow. So, I'll send the first chunk of the prompt into Meta AI first. Then, after that, I'm going to switch over to Google Flow, switch from the image model to the video model, and upload the same

15:38reference character there as well. Now, quick heads-up here. When you're inside Google Flow and you click on the video model, make sure you switch from frames to video to ingredients to video. And that part is really important because it lets you upload multiple reference images for a specific style. And for what we're doing, that's exactly what we need. Multiple shots of our characters in different places and different settings, [music] so the model actually understands what

16:04we're trying to build. So, once I switch to the ingredients option, I'm going to upload all the images we just created, set the output to video generation, and then just paste in my prompt. All right, now while Google Flow is still working in the background, I'm going to jump over to my browser and search for Gemini because it's also another video generation tool we can test. So, once you sign in and log in to Gemini, all you've got to do now is click on the tools toggle, then

16:28select create video. That will bring up the video generation interface. Now, not a lot of people actually know that Gemini can create videos, but it's honestly another really solid free model to work with. So, what I'm going to do is upload the same multiple reference images I used in the other tools, the ones we already generated in the image models, and bring them into Gemini as well. Then, I'll set everything up properly, and for the model selection, I think I'm going to go with the thinking model instead of the fast

16:56model. The reason is simple. It takes a bit more time, but it's usually better at understanding complex prompts and giving higher quality output. So far, we've basically run the exact same prompt across Meta AI, Google Flow, and now Gemini. And now, I just want to compare all three outputs and see which one actually gives us the best result. All right, now after a couple seconds, we've got our outputs from the different models. First off, Meta AI basically couldn't handle it at all. It failed to generate a proper output, and it even said that the sequence and complexity of the prompt were too much for it to process. It was even

17:32suggesting things like rendering only the first 2 seconds of the video, which honestly isn't practical or usable for this kind of workflow. So, for Meta AI, I'm not going to lie, it just didn't perform well here. Now, Google Flow did generate a video, so let's preview that.

17:52[music] And Flow was actually able to handle the multi-shot structure pretty well. Overall, the result looks solid. The only limitation is that Flow is capped at around 8 seconds per clip. But if you're working on a budget, it still produces something really usable. It understood the prompt and gave us a decent output. Now, I also tested Gemini, and it was able to generate a video as well. And honestly, Gemini did

18:16a good job, too.

18:24So, when you compare all three, Flow, Gemini, and Meta AI, I'd say Gemini handled prompt adherence the best overall. But Flow did a better job when it comes to handling multi-shot structure specifically. Now, I also ran the exact same prompts [music] on Higgs Field, which is a paid tool, just so you know. And this was the

18:42result I got.

18:57And honestly, Higgs Field performed the best out of all of them, which makes sense because it has access to more advanced models like SeaDance 2.0. So, if you have a bit of budget to work with, Higgs Field is actually worth considering. Because we used the exact same prompts across all models, and you can clearly see the difference in output

19:16quality. All right, so what I'm going to do next is run the next prompt across all four models again and compare the results. All right, now after a couple seconds we've got outputs from every single model again. Once again, Meta AI wasn't able to generate a proper video. It looks like Meta AI has a limitation when it comes to video length, and it also struggles with multi-shot sequences. So, it can still work fine for simple one-off generations, but for something as structured and complex as this

19:44workflow, it just doesn't perform well. Now, moving on to Google Flow, Flow actually did a really good job here. The multi-shot structure is solid, the character consistency is maintained, and overall the output looks quite good.

20:07That said, there is a bit of distortion in some parts of the video. So, with Flow, you might need to run the same prompt a couple of times before you get a clean, fully usable result. But once it works properly, the output is definitely strong.

20:26[music] Now, for Gemini, the video output also looks decent. In a lot of the clips it holds up fairly well, and the structure is understandable. So, overall Gemini gives you a pretty balanced result as well. But just to be clear, if you're using Meta AI for simple, single-shot, or one-off generations, it's totally fine. It can handle that without issues. But for complex, multi-shot video workflows like this, it just doesn't really do a good job. Now, for Google Flow, Flow actually did a really solid job here. The multi-shot structure looks good, and overall the output is quite usable. And so far, it really feels like both Gemini and Flow are the only two

21:07free options consistently giving strong results for this kind of workflow. Now, for Higgs Field, we also got an output that looks like this.

21:29And honestly, it looks really good. I mean, this is just from a single prompt that we ran across all the different models, and somehow Higgs Field was able to produce something this polished. So, at this point, what I'm going to do is just repeat the entire process. Since we only have about five chapters, I'm just going to keep pasting each prompt into all the models one by one. Then, I'll simply download whichever output looks the best from each round and keep building the video step-by-step until everything is complete. And that's pretty much it. All right, so once you've got all your footage ready, the next thing we need is some really smooth

22:02background music. Now, normally, we just generate something with AI, but for this part, I want to show you a different method that actually works really well. So, if you've got a YouTube channel, just click on your profile and head over to YouTube Studio. Once you're inside YouTube Studio, scroll down until you see the audio library section. Now, in there, you basically have access to a huge library of music you can use for free. Since we're doing something documentary style, we don't want anything too soft or random. We want something cinematic, something a bit suspenseful, intense, [music] and that builds tension. And inside the audio library, you can

22:36actually filter the music, too. So, you can set [music] it to things like cinematic, dramatic, and intense to narrow it down. So, I'm just going to go through a few of these tracks, listen to them, and then download the one that really fits the vibe of the video. All right, so once you've got your footage, your background music, and your voice-over ready, the next thing we need to do is edit everything together. Now, for this part, I'm just going to keep it simple and use CapCut. I've already made a new project, and if you don't really know how to edit yet, don't stress. This part is actually pretty easy. What I

23:05usually do first is import my whole project folder into CapCut. The reason is I like keeping everything organized in one place. All my voice-overs, music, footage, everything goes into a single folder per project. So, I just bring that folder into CapCut, click on it, and make sure all my assets are there and ready to go. Now, the first thing I'm going to do is drag my voice-overs

23:26onto the timeline. Since we generated them in batches, I just line them up so everything flows properly from start to finish. After that, I'll drag in the background music and place it underneath the voice-over. Then, I'll trim the music so it matches the exact length of the voice-over. I don't want it going longer than the narration. Once that's done, I'll add a fade in and fade out so it doesn't just come in or cut out abruptly. Then, I'll lower the music volume so it stays subtle in the background and increase the voice-over volume slightly so it always stays clear and on top. Now, the cool thing is most

23:59of the footage we generated already comes with built-in sound effects. So, we don't really have to worry about adding extra audio for that. I actually like how a lot of these models now automatically generate sound effects per scene. It just makes everything feel way more complete. So, at this point, all you've got to do is start dragging your video clips onto the timeline in the same order you generated them because now the main job is just syncing everything together. You're going to do a bit of cutting here and there, trimming parts, cleaning things up,

24:25removing anything that doesn't fit. That's just part of the process. So, what I'm going to do now is play the video and start making those adjustments. And I'm going to speed this part up a bit because it's pretty simple. The main thing is just making sure the voice-over matches the visuals so everything stays in sync. Once that's done, the whole thing starts to feel like a proper video instead of separate clips. So, yeah, I'll just speed through this section. All right, so here's a quick recap of exactly what I did. It's

24:51the 4th of March, 2023. A marble hall in downtown Dubai. A man walks onto the stage. He clicks a remote. The lights dim. Behind him, something appears on the screen. A ring of glass rooms. But they're not in the sky. They're not on land. They're under the sea. So why is one of the richest cities on

25:20Earth building down instead of up? Here's the thing. Dubai is running out of room. The towers can't climb any higher. The fake islands are almost full. So where do you go when there's nowhere left to go? You go under. For the voiceover, I split everything into multiple chunks and intentionally left small gaps between them. Now, the reason I did that is because I wanted to build suspense. If you look at channels like Fern, you'll notice they often use these natural pauses that create

25:58anticipation and keep viewers locked in. That's exactly the effect I was going for here. So most of those gaps you see in the timeline are actually intentional. They're there for pacing, not mistakes. Now, the fast cuts you see in the video were actually generated by the model itself. So I didn't really need to manually edit those. They were

26:17already built into the prompt structure. I also went through and adjusted the audio levels. Some clips were a bit too loud or didn't sit well with the narration. So I just reduced their volume where needed. After that, I stretched the background music to make sure it covered the full duration of the video smoothly from start to finish. And that's basically it. Now, Fern does also use a mix of realistic footage and

26:38screenshots from vlogs in their videos. But that pipeline is a lot more complex to set up properly. That's actually part of the reason I made this tutorial in the first place. To simplify everything down to a system anyone can follow. In the next course, we'll go deeper into how to integrate real footage, screenshots, vlog style clips, and all that extra polish to take things to the next level. But, once you're done assembling your video, the final step is

27:05simple. Just export it. All right. Now that we're done with the video creation process, the next step is making our thumbnail. So, if you go back to Claude, right after it gives you the video animation prompt, all you need to do is type next. Once you do that, Claude will generate thumbnail prompts for you. And the way this works is pretty useful. It actually gives you multiple variations, usually around five different thumbnail concepts to choose

27:29from. Now, to make things even easier, Claude also recommends which model you should use for generating the thumbnails. And since we already have access to a model that gives us unlimited image generation, that's the one we're going to stick with. So, we're going to use Google Flow for this part. So, I'll head back to Google Flow, log in again, and create a new project. Then, I'll go back to my chatbot, scroll up to the first thumbnail prompt, copy everything, and use that to generate the first batch of thumbnails. Now, as you probably know, most image models actually perform way better when you give them a reference

28:02image. So, if you go back to the source PDF in the master prompt and scroll through it, you'll see I added a couple of thumbnail references in there. What you can do is just take screenshots of those references, and then attach them inside Google Flow whenever you're generating each thumbnail variation. And the reason we do this is to keep everything consistent. So, all your thumbnails have the same style and don't feel random or disconnected. [music] For example, this is what we got when we

28:29used the prompt without any reference. Now, I'm going to take that exact same prompt again, but this time, I'll attach a reference image from the source material file and send it along with the prompt. [music] And you'll notice something interesting. The output completely changes, even though the prompt is exactly the same. That's basically why references are so powerful because they guide the model into the exact visual direction you

28:53want. Now we've got the second output after attaching a reference. And as you can see, this one looks noticeably better. At first glance, you might think both results look the same, but if you look closely at the details, you'll notice the version with the reference image looks more polished and more professional overall. Now what I'm going to do next is take every single thumbnail prompt variant that Claude gave me and run them one by one. And for each one, I'm going to attach the same reference image inside Google Flow so we can keep the style consistent and also compare which thumbnail stands out the most. So, I'll

29:25just keep copying each prompt and pasting it into Flow one after the other. Now, since we've got unlimited generations here, there's really no issue with testing multiple variations. So, I'll just run through all of them and see which one gives the best result. All right, so now the thumbnails are done and honestly, they look really really good. I like the consistency across all of them, the style, the text, the overlays, everything feels clean and intentional. This is the kind of thumbnail you could realistically download and upload straight to YouTube without needing much extra editing. So, at this point, all I've got to do is just download the best one and that

30:00completes the video production process. Now for the final step, channel creation. If we go back to our chatbot and scroll all the way to the top, we'll see a list of channel name ideas and branding suggestions. So, I'm going to scroll up, review the options, and pick a name. All right, so right now I think I like idea number five the most. So, I'm just going to

30:20copy that name and head over to YouTube. Then I'll click on my profile, hit create channel, paste in the name I just got, and click create. And boom, just like that, the channel is ready. Now we need to brand it. So, we're talking logo, banner, and channel description. For this part, we're going to use Canva. But first, if you go back to the chatbot, it already gave you a couple logo and banner prompt ideas. So, I'm going to copy the logo prompt, jump back into Google Flow, paste it in, set it to

30:49a square 1:1 format, and generate it. While that's running, I'll go into Canva, sign in with my Google account, and hit the create button in the top left. That brings up a menu. Then, I'll search for logo and pick a YouTube channel logo template, so it opens a fresh workspace. Inside Canva, you can basically design your logo however you

31:07want. Now, let's check Flow real quick. And yeah, this is what it gave us. But, honestly, it's a bit too minimal for my taste, so I'm probably not going to use it. So, I'll just build the logo from scratch in Canva instead. To make it a bit fun, I'll go to elements, search for

31:22vault, and look for a vault-style icon. Once I find one I like, I'll drop it in, stretch it, and adjust it until it looks right. And yeah, if you like what Flow gives you, you can definitely use it. But, I just want something a bit more customized. Once you found the design you like, the next important step is creating a consistent brand color identity. This part is actually really important, because it's what makes everything feel professional and recognizable. It's very hard to keep your thumbnails, logo, and banner consistent if you don't lock in your colors early. So, for this course, I'm going to go with a green and black

31:57theme. That's the brand direction I want to build around. Inside Canva, I'll click on the color wheel and select a strong green tone. Then, combine it with black as the secondary color. Once that's done, I'll just download the logo. Now, we can move on to the banner. I'll go back into Canva, open the search bar again, and this time search for

32:15banner. Then, I'll select the YouTube banner template, which opens a fresh canvas. Now, this part is a bit tricky, because YouTube automatically crops [music] and adjusts your banner depending on the device. So, what that means is someone watching on mobile will see a different portion compared to someone on TV or desktop. Because of that, you need a proper guide to know where to place your key elements, so nothing important gets cut off. Now, I already have a template for that, so I'm going to upload it into

32:40Canva and place it into my workspace. Then, I'll set it as a background image so I can clearly see the safe zones and structure everything properly inside them. Now, the key thing you need to understand is that you should always place your most important elements inside that red underlined area. That area is basically the universally visible zone, meaning it's what shows properly across all devices in YouTube, whether someone is on mobile, desktop, or TV. So, for anything important, that's where it has to go. Now, for me, I'm just going to add a simple text

33:11element that says "New videos weekly." After that, I'll change the font to something clean and visually strong, just to make it stand out a bit more. Once that's done, I'll add a few extra design elements to the banner to give it more personality. So, I'm going to go into the elements tab inside Canva and search for a loading bar, since that gives a nice

33:30sense of progress and motion. I'll pick one that fits, place it carefully within the safe zone, and adjust it so it aligns properly with the layout. Then, I'll click on the banner, open the color wheel, and change the loading bar color to green so it matches the overall brand identity. That way, everything stays consistent. [music] So, now we've got the logo, the banner, everything is coming together really nicely. It looks realistic, but still has that clean, professional touch. At this point, I'm just going to export everything. Once that's done, I'll go back to my YouTube Studio and click on customize channel. That takes me into the studio where I can upload both the

34:07logo and the banner I just created. Now, if I go back to Claude, I can see it also generated two different channel [music] description variants. I like variant B the most, so I'll copy that, head back into YouTube Studio, scroll down to the description section, and paste it in. I might add a small emoji just to make it feel a bit more engaging, and then click publish. Now, if I go back to the channel, you can see everything is is Logo, banner, description, everything is in place. And just like that, we've built and fully branded a YouTube

34:37channel from scratch.