# I Built a $58,989/Month AI Finance Channel With Claude + ChatGPT

Channel: Lucas AI
Video: https://www.youtube.com/watch?v=ktJwCgwZcPM
Duration: 18 min
Language: English
Words: 3458
Transcript page: https://viewrankai.com/tools/youtube-transcript/ktJwCgwZcPM

---

[0:00] There's a faceless YouTube niche blowing up right now. And some of these channels are pulling hundreds of thousands, even millions of views with videos that look ridiculously simple. No camera, no actors, no expensive production team, just topics like the economics of owning a casino, the economics of owning an airport. But when I started researching

[0:17] this niche, I noticed something strange. It wasn't just one channel doing it. I found channel after channel using almost the same formula. simple economics topics, curiositydriven thumbnails, faceless narration, and explainer style visuals. So, I wanted to know, can we build this entire style of channel using AI and automate most of the workflow for free? I reverse engineered the process from start to finish, finding topics with viral potential, researching them, writing the documentary script, generating the voice over, creating every visual, automating image generation, editing the video, and building the thumbnail. and I'm going to build one from scratch using the wild economics of owning a cruise ship. Now,

[0:55] somewhere in this video, I'll reveal the complete resource pack with every prompt and tool I used. Let's build the channel. So, before creating anything, I captured the channel's entire blueprint and handed it to Gemini. Not to summarize what was obvious, but to uncover what everyone else was missing. Within seconds, it started breaking everything down. The psychology behind the thumbnail, the reason the titles generate curiosity without feeling like clickbait. Why every topic feels instantly familiar yet reveals a hidden world most people never think about. But then it uncovered something I wasn't expecting. It said the real opportunity wasn't copying this niche. It was copying the framework. Instead of

[1:30] competing in the exact same space, it identified several untouched macro niches built on the same psychology. Entire industries, everyday systems, and massive businesses hiding in plain sight that almost nobody is making documentaries about. That completely changed my approach. I wasn't just reverse engineering one successful channel anymore. I had discovered a repeatable blueprint that could be adapted to entirely new categories. And that's when I realized the niche wasn't the secret. The system was. The biggest mistake new creators make isn't bad editing. It isn't bad thumbnails. It isn't even bad titles. It's choosing topics that never had the potential to explode in the first place. Most people sit down and ask, "What should I make

[2:09] next?" That's the wrong question. Instead, I wanted a system that could continuously uncover ideas people were already curious about, but almost nobody was explaining. So, I built a research prompt designed around one goal. Not finding interesting topics, finding topics the YouTube homepage wants to recommend. Instead of chasing trends, the prompt searches for businesses, industries, and everyday systems hiding massive amounts of money behind ordinary things people see all the time. Things like cruise ships, data centers, golf clubs, self- storage facilities, hospitals, car washes, container ships, businesses that millions of people recognize instantly, but almost nobody understands. Within seconds, Gemini started producing ideas that already checked the biggest boxes for browse and

[2:52] suggested massive curiosity, strong visual potential, real business stories, and enough depth to build a 20 or 30 minute documentary. But what impressed me even more was that every idea came with a complete strategy, not just a title. It suggested the thumbnail concept, the curiosity gap, why someone would actually click, and even explained why YouTube would likely recommend that topic to a brand new channel. That's a completely different way of thinking about content. You're no longer brainstorming, you're filtering. Every weak idea gets rejected before you ever spend hours writing a script. And if none of the ideas feel strong enough, I don't start forcing one to work. I

[3:27] simply type one word, next. A few seconds later, 10 completely different opportunities appear. Entirely new industries, different business models, different audiences, same proven framework. Self- storage, trailer parks, emergency rooms, cruise ships, scrapyards, water companies, parking lot empires. Some of these are topics I'd never even considered. Yet, every one of them has the ingredients for a documentary people instinctively want to watch. At that point, content stops being the bottleneck. Instead of worrying about running out of ideas, you're building a library of videos that

[3:58] could keep a channel alive for years. Because once you understand the framework, you're not creating one documentary. You're creating an endless documentary machine. Now, we have the topic, the wild economics of owning a cruise ship. But this is where most workflows fall apart. They find a great idea. Open a chatbot, ask it to write 4,000 words, and get something that

[4:17] sounds like it was written by a chatbot. So, we're not doing that. For the script, I'm going to use Claude Opus 4.6. And as promised, I'll show you the method I'm currently using to access it through GenSpark. Open GenSpark, click new, go under content creation, and select AI chat. Now, I'm pasting the script prompt I built specifically for this documentary format. And this prompt is doing much more than saying, "Write me a YouTube script." I'm forcing the model to first find the story underneath the topic, the hidden system, the contradiction, the money, the risk, and most importantly, the question that keeps someone watching because they need

[4:50] to know how this thing actually works. I'm also telling it how to handle retention. Instead of dumping information for 5 minutes straight, the script has to keep introducing new complications, reversals, and payoffs throughout the story. And there's another rule in here that's extremely important. Don't invent anything. If a number can't be supported, don't make one up just because it makes the script sound better. Now, replace the topic placeholder with the wild economics of owning a cruise ship. Set your target length. For a roughly 20 to 25 minute documentary, I'm aiming around 3,500 to 4,000 words. Before you generate, open the model selector. Genspark gives you access to different models here. For

[5:27] this script, I'm selecting Claude Opus 4.6. I'm also turning on web search because I don't want the model relying only on what it already knows. Now, hit generate and watch what happens before it even starts writing. It's searching for cruise ship construction costs, operating expenses, revenue models, dry dock maintenance, debt, financing, and realworld failures. Then it starts opening and reading sources. That's exactly what I want. Research first, script second, because the quality of a documentary isn't determined by how sophisticated the writing sounds. It's determined by whether there's actually a story worth telling underneath it, and the research finds one. At first, cruise ships look ridiculous in the best

[6:05] possible way. A single mega ship can generate millions of dollars in revenue while carrying thousands of paying customers every week. Sounds like a floating money printer until you look at what it costs to keep that machine alive. A modern mega ship can cost around $2 billion to build. Then you have fuel, thousands of crew members, food, maintenance, financing, and an asset that needs to keep generating cash for decades. Then the script finds the real twist. The ticket isn't the whole business. Once passengers are on board, there are drinks, casinos, specialty restaurants, excursions, spas, and upgrades. So, the ship starts looking less like a hotel and more like a

[6:40] floating marketplace where thousands of customers can't exactly walk across the street to a competitor. And that's where the documentary gets much stronger because now we have a system and systems can break. The script eventually connects all of this to what happens when these enormous ships stop generating revenue. then uses a real cruise industry collapse to show how billions of dollars in assets can become liabilities incredibly fast. That's what we're looking for, not 4,000 words of filler. A story where each answer creates the next question. Once Claude finishes, don't immediately record the voice over. Read the script. Check the major numbers and claims. Cut anything repetitive. Rewrite anything that

[7:16] doesn't sound like you. The model should handle the heavy research and first draft. You still make the final editorial decisions. And for the exact Gen Spark setup I'm using and the resources from this workflow, I'll put the details in my Discord community. Now, we've got the topic, we've got the research, and we've got the finished documentary script. But right now, it's still just words on a screen. Now, we have the script. Before we generate a single visual, I want the voiceover finished first because the narration controls the pacing of everything we're going to build next. Open your favorite voice generation tool. If you're not ready to pay for one yet, I've already

[7:48] covered 10 plus free AI voiceover tools on the channel that you can use while you're getting your channel off the ground. For this project, I'm using 11 Labs. And don't make the mistake of grabbing the first deep voice you hear and calling it cinematic. The voice needs to match the content. This is a business documentary about billion-dollar ships, hidden economics, risk, and failure. I want something mature and authoritative, but still conversational enough that it doesn't sound like a movie trailer. So, I'm going with Adam. Now paste in the script. Before generating the entire thing, test a small section first, especially the opening. Listen for three things. Does the hook feel natural? Does

[8:23] the voice make the numbers easy to follow? And would you actually listen to this person for 20 minutes? If the answer is no, change the voice or settings. Now, don't try to fix a bad narrator with editing later. Once I'm happy with the tone, I'll generate the narration. 11 Labs gives me two variations here. So, I'm listening to both and keeping the version that fits the documentary better. then download the audio. But we're not dropping it straight into the editor yet. There are usually small dead spaces between sentences and paragraphs. And across a long documentary, those little pauses add up. They can make an otherwise strong script feel painfully slow. So,

[8:55] I'm using this free remove silence from audio tool. Drop the narration here. Choose how much silence you want to preserve and process the file. Don't remove every pause. That's important. Natural speech needs breathing room. We're cutting unnecessary dead air, not turning the narrator into an auctioneer. Now, there's one more pass. I'm taking the cleaned recording into Adobe Podcast Enhance to polish the audio and make the narration sound cleaner and more consistent. Upload the file, process it, listen to the result and download it if it actually sounds better. Don't automatically assume more processing means better audio. If the enhanced version sounds overly compressed or

[9:30] unnatural, keep the cleaner original. Now, open your video editor and place the final voice over on the timeline first. This becomes our backbone. Every visual, animation, sound effect, and transition from this point forward has one job. Support what the viewer is hearing at that exact moment. And this is where the workflow gets much more interesting. Because we're not going to throw random stock footage over a 20-minute narration. We're going to break the script into visual beats and generate scenes that actually explain the story. Voice over is done. This is where most people make the biggest mistake. They start generating random images that look good individually, but

[10:03] when you put them together, they don't actually tell the story. So instead, we're going to make the visuals directly from the script. Open chat GPT. I'm going to paste this storyboard prompt here. But before hitting generate, go back to GenSpark, copy the entire finished script, and paste it into the script section of the prompt. What this prompt does is important. I'm not asking Chat GPT to simply describe every sentence. I'm asking it to translate each sentence into a visual idea. So if the narration says a company is losing money, I don't necessarily want a

[10:31] businessman standing there looking sad. Now, hit generate. Chat GPT takes the first section of the script and gives us 15 visual prompts at a time. Look at the first one. A large cruise ship can generate an estimated $2.7 million in revenue per day. Instead of simply drawing a cruise ship, the visual turns the ship into a giant money machine, spitting out stacks of cash. Next, when we're talking about the ship approaching a billion dollars in annual revenue, we get a ship climbing an enormous mountain of money toward a $1 billion flag. Then the story changes. That's what I mean by visual storytelling. We're not decorating the narration. We're making

[11:05] the narration easier to understand. And because the prompt stops after 15 images, I can review each batch before moving forward. If something doesn't make sense visually, fix it now. But we have another problem. The output contains the original sentence, the visual concept, and the final image prompt. For the automation I'm about to show you, we only need the prompts. And I'm not manually copying hundreds of them one by one. Open chat GPT in another tab. Paste the entire batch we just generated and type extract only prompts from 1 to 15. Hit generate. And now look at the result. Prompt one, prompt two, prompt three, all the way

[11:38] through prompt 15. No original narration, no visual concept explanation, just the clean prompts we actually need. And this is where the workflow is about to get dramatically faster because generating 150 or 200 images manually, copy prompt, paste prompt, generate, save, repeat, is a terrible use of your time. So next I'm going to show you the free automation workflow I built to take these prompts and handle the repetitive generation process for us. Copy all the image prompts we extracted from chat GPT. Then head over to Google Flow and switch to agent mode. Paste the entire batch of

[12:10] prompts here before you hit generate. Open the image generation settings and configure the output you want. Since our documentary uses one consistent handdrawn explainer style, keep those settings consistent across the entire batch. Save the settings, then hit generate. Now watch what happens. Instead of copying prompt one, generating an image, downloading it, then repeating that process again and again, the agent works through the batch for you. And look at the results. The $2 billion construction cost becomes this massive price tag weighing down the cruise ship. The fuel expense becomes an

[12:40] oversized fuel pump feeding the ship. The daily operating cost becomes a calendar, literally throwing money away. The 30-year financial burden becomes this endless timeline draining cash year after year. And the profit comparison becomes a giant balance scale between the cruise ship and a hotel. More importantly, the visual language stays consistent. Beige background, handdrawn outlines, simple characters, exaggerated scale, and financial concepts you can understand almost instantly. So, instead of manually creating dozens or eventually hundreds of images, we've turned the storyboard into a batch visual generation workflow. Now we have the script, voice over, and visuals. The next step is turning all of this into the final documentary. Now that all our

[13:19] visuals are ready, it's time to turn everything into the final video. Download all the images we generated. Open your favorite video editor and drag the visuals onto the timeline above your finished voiceover. First, arrange them in the same order as the script. This part matters because every visual was created for a specific line of narration. So, when the voiceover talks about the $2 billion construction cost, that exact visual should appear. When it moves into fuel costs, switch to the fuel visual. If you ever forget which image belongs to which part, just go back to chat GPT. Since we generated every prompt directly from the script, you already have the sentence to image

[13:52] mapping. You don't have to guess. Next, adjust the duration of every image to match the voiceover. Don't leave one static image on screen for too long. Change the visual whenever the narration introduces a new idea, number, comparison, or reveal. Now, generate your captions. Choose a clean caption template that matches your documentary style or create your own. Keep the text readable and don't let oversized captions compete with the visuals. At this point, the video technically works, but static images alone can still make

[14:20] it feel AI generated. So add movement. Use subtle zoomins, zoom outs, fades, and directional movement. Then use key frames to manually control the framing. you can slowly push toward an important detail, move across a larger illustration, or pull back when the narration reveals the bigger picture. Don't animate everything aggressively. The goal is to make a still image feel alive without making the viewer notice the animation itself. If you don't know how to use key frames yet, search YouTube for a key frame tutorial for your specific editor. You can learn the basics in a few minutes. Between selected scenes, add transitions like slides, pull outs, or mixes, but use

[14:56] them selectively. If every cut has a flashy transition, the editing starts feeling cheap. Most cuts can simply be clean cuts. Finally, add an adjustment layer across the entire video. Make small corrections to exposure, contrast, highlights, and vignette until all the AI generated visuals feel like they belong in the same video. Watch the entire video once from beginning to end. Fix any awkward timing. Check your captions, balance the audio, and that's it. Topic, research, script, voice over, visuals, and editing. You've just built

[15:25] the entire documentary from scratch. We're almost done, but there's one thing we absolutely cannot ignore, the thumbnail. Go back to Gemini, and I'm giving you one final prompt. This time, make sure you include both your video title and complete script. Hit generate, and Gemini turns the script into a complete thumbnail prompt, headline, main subject, layout, financial callouts, arrows, icons, and the overall visual style. for this video. It created, "So, you want to own a cruise ship?" With the cruise ship dominating the center and the business model broken into simple callouts: fuel costs, onboard spending, ticket economics, dry dock losses, and fixed debt. And this is exactly what we want. Now, copy that

[16:03] thumbnail prompt, go back to Google Flow, paste it in, and generate. I generated two versions here. Both follow the same visual identity, but the second version is stronger. The ship is larger and cleaner. The call outs have more breathing room and the composition is easier to understand at thumbnail size. Don't automatically use the first generation. Create multiple variations and choose the one with the strongest hierarchy and readability. And that's the complete workflow from finding the niche and researching topics to scripting, voice over, automated visuals, editing, and finally the thumbnail. I'll put the prompts and resources from this workflow in my Discord community. If this helped you,

[16:38] subscribe and I'll see you in the next one. A large cruise ship can generate an estimated $2.7 million in revenue per day. On a single 7-day voyage, Royal Caribbean's Icon of the Seas, the largest cruise ship ever built, may pull in somewhere around $19 million. Across an entire year, that one vessel alone could approach a billion dollars in revenue. The profit margins for major cruise lines run between 16 and 19%, nearly double those of comparable hotel chains. If you just glance at those numbers, you'd think a cruise ship is one of the greatest money printing machines ever devised. A floating ATM with a water slide. But here's what

[17:13] those numbers don't tell you. Icon of the Seas cost approximately $2 billion to build. It burns through roughly 250 tons of fuel every single day. That's more than 80,000 gallons. Its estimated operating costs run close to $900,000 per day. And that meter is ticking whether the ship is full of passengers or sitting empty at a dock. A cruise ship doesn't wait for you to decide whether the economics work. The moment it exists, it starts costing you a million dollars a day, and it will keep

[17:40] costing you that for the next 30 years. The cruise industry looks like a business built on sunshine and buffets. Underneath, it runs on a financial logic so aggressive, so leveraged, and so dependent on nothing ever going wrong that even a few empty weeks can trigger a crisis worth billions. Understanding how that works reveals something genuinely strange about one of the world's most popular vacations. Let's start with what it takes to bring a

[18:04] cruise ship into existence.

---

About this transcript
Read from YouTube's own caption track and laid out by ViewRank AI (https://viewrankai.com).
ViewRank AI finds the videos already beating a creator's own average on Instagram, TikTok and YouTube Shorts, transcribes them from the audio itself in more than 60 languages, and turns what worked into new ideas and scripts.
Free transcript tools, no account needed: https://viewrankai.com/tools
How to read any video this way: https://viewrankai.com/llms.txt