18:29Matt Loui · Tutorial
21 BEST CapCut Tricks to Take you from a Noob to PRO!
An 18-minute desk-side walkthrough of 21 built-in CapCut features most editors never find.
December 15th 2025The argument in one line.
Claude can coordinate an entire faceless YouTube video pipeline — script, character design, 50 scene images, and thumbnails — inside one conversation for under seven dollars when connected to an image generator via MCP.
Read if. Skip if.
READ IF YOU ARE…
- You want to start a faceless YouTube channel but do not have on-camera skills or a large production budget.
- You are already using Claude and want to extend it with external tools via the MCP connector system.
- You are exploring AI-driven content production as a side income stream and need a concrete cost model.
- You make educational or entertainment content and are open to a stick-figure or illustrated animation format.
SKIP IF…
- You want live-action or talking-head video rather than AI-generated illustrated content.
- You are evaluating the ethics of AI-generated faceless channels rather than looking to act on this approach.
- You need a production system that does not depend on a $125/month third-party image generation subscription.
The full version, fast.
Connecting Higgsfield to Claude via MCP lets one conversation handle the entire faceless YouTube production stack. The host feeds Claude proven viral topic ideas from existing YouTube performance — not open-ended brainstorming — which is where the video argues the real quality ceiling sits. From there: character generation with a pose library for consistency, a YouTube-optimized script, 50 scene images generated in script order, CapCut assembly, and three thumbnail variants — all inside one session. Total image cost: $6.48. A reference channel with four such videos earned roughly $1,200 in two weeks on $26 of production spend.
Free for members
Chat with this breakdown — free.
Sign in and you get 23 free chat messages on us — ask for the hook, quote a framework, find the exact transcript moment, generate a markdown action plan. Bring your own key when you want unlimited.
Create a free account →Where the time goes.
00:00 – 00:45
01 · Proof of concept
Shows a live faceless channel's analytics and a clip from the AI-produced video to establish credibility before teaching anything.
00:45 – 01:10
02 · Claude + Higgsfield MCP setup
Walk-through of connecting Higgsfield to Claude via the custom MCP connector URL and permission configuration.
01:10 – 02:37
03 · Character creation
Prompting Claude to generate the Sticky stickman character with Higgsfield, then building a pose library for character consistency.
02:37 – 03:54
04 · Scripting with YouTube context
Feeding Claude viral reference ideas from YouTube, generating script concepts, selecting the poop history topic, and producing a YouTube-optimized script with a hook section.
03:54 – 05:09
05 · Storyboard and image generation
Uploading the pose library, prompting for scene stills, refining visual style to mixed-illustration, and scaling to 50 final images at 7-second max intervals.
05:09 – 06:08
06 · Assembly, voice over, and side use case
Importing 50 ordered images into CapCut, recording voice over, adding music. Side use case: product marketing imagery in two prompts.
06:08 – 07:23
07 · Cost breakdown and ROI
Exact credit math: 81 images x 4 credits = 324 credits = $6.48. Reference channel ROI: $26 spent, ~$1,200 earned in two weeks.
07:23 – 08:38
08 · Thumbnails and session summary
Claude generates three thumbnail variants from one prompt. Summary of everything produced in under an hour inside one conversation.
08:38 – 13:48
09 · The actual produced video
Full playback of the AI-generated faceless video on ancient city sanitation, demonstrating the illustrated stick-figure format output quality.
Lines worth screenshotting.
- Feeding Claude viral YouTube references before scripting produces better hooks than blank-slate prompting — the idea quality is the bottleneck, not the AI.
- A pose library generated before any scene images is the only reliable way to keep a stick figure character consistent across 50+ generations.
- Prompting Claude to generate images in script sequence means the downloaded folder drops straight into CapCut with no reordering.
- One follow-up style correction mid-session fixed all subsequent image generations for free — iteration cost on style is nearly zero.
- The first video session is the slow one; every subsequent video reuses the locked character and style as a reference anchor.
- Higgsfield charges 4 credits per 4K generation — 81 images for a five-minute video cost 324 credits, or $6.48 at the $125/month tier.
- Claude generated three thumbnail A/B test variants with a single prompt from the same session that produced the video — no separate design tool needed.
- The demonstrated reference channel earned about $1,200 from four videos that cost roughly $26 total to produce.
- Voice over recorded under a blanket is functionally sufficient for faceless content — the illustrated visuals carry production quality, not the recording environment.
- The same two-prompt workflow applies to product marketing content, not just educational YouTube.
A $6 content pipeline is real — here is the constraint map.
WHAT TO LEARN
The cost ceiling on AI faceless content is already under seven dollars per video, but the idea quality remains the human-controlled bottleneck.
- Feeding Claude proven viral formats from YouTube before asking for script ideas produces better hooks than blank-slate prompting — the AI amplifies what is already working, it does not discover it.
- Building a pose library before any scene images is the only reliable way to keep a stick figure character visually consistent across 50 or more generations in one session.
- Prompting the model to generate images in exact script order means the downloaded batch drops into an editor with no reordering needed — the sequence is baked into the generation queue.
- Style corrections mid-session cost almost nothing — a single follow-up prompt fixed the black-and-white vs. mixed-illustration issue for all remaining images without regenerating earlier ones.
- The first video session carries most of the setup cost in time; once character, style, and script format are locked as a chat context, every future video starts from that reference anchor.
- Voice over recorded in an imperfect environment is functionally sufficient for faceless content — the illustrated visuals carry the production quality, not the audio environment.
- Generating thumbnails inside the same session that produced the script and images takes one prompt and outputs three A/B variants ready for YouTube test-and-compare.
Terms worth knowing.
- Higgsfield
- An AI image and video generation platform that connects to Claude as a custom MCP tool, letting Claude prompt image generations directly within a conversation.
- MCP connector
- A Model Context Protocol integration that links an external tool to Claude, allowing Claude to call that tool's functions within the same chat session.
- Pose library
- A set of reference images showing a character in multiple angles and poses, generated once and fed into future scene generations to maintain consistent character appearance.
- Nano Banana Pro
- Higgsfield's 4K image generation tier, priced at 4 credits per image at the $125/month plan.
- Faceless YouTube channel
- A YouTube channel that publishes videos without showing the creator on camera, typically using illustrated, animated, or stock-footage visuals with voice over narration.
Things they pointed at.
Lines you could clip.
02:35
“Your ideas are oftentimes the most important factor for success. See them as the backbone of virality.”
07:15
“$26 in a couple of hours, $1,200 made from their videos.”
08:24
“Half the time was just spent building that initial visual and channel direction. Now that that's done, we just hit play on mass video production.”
Word for word.
Read-along
Don't just watch it. Burn it in.
See every word as it's spoken — crank it to 2× and still catch all of it. The same dual-channel trick behind Amazon's Kindle + Audible.
analogystory
00:00Face's channels are blowing up all over YouTube. Take this channel for instance, which has only four videos, over 5,000 subscribers, and over 600,000 views in the last two weeks.
00:09It's easy to understand why they're blowing up when you see that you can now create a viral monetized content for practically nothing. The crazy part is some of these channels are generating thousands of dollars per month.
00:20There's a new system using Claude and Higgs Field that makes creating entire videos from start to finish as simple as a few basic prompts. I created this video in thirty minutes. Here's a small section from it.
00:31In ancient Rome, you could legally be sued for getting hit by a bucket of human poop. Because every morning in cities of a 100,000 people, citizens would walk to their windows, lean out, and just empty everything they made overnight straight into the street.
00:46I'll show you how to get set up, how to create videos on autopilot, and exactly how much this five minute viral video cost me to make. Because it's one thing to make thousands of dollars from these videos, but if they cost thousands of dollars to make, well, that just doesn't make sense. In a bit, you'll see that it definitely makes sense.
01:03So first, let's get set up with Claude and Higgs Field. Open Claude on your computer, then go to the bottom left, click on settings, then connectors, and add custom connector.
01:13In the name section, you can put Higgs Field, and in the MCP URL section, you can just put what I put here. You can also find this link in the description. Then hit connect.
01:22It'll redirect you to your Higgs Field account, and all you have to do is hit allow. If you then click configure, you can change when Claude can use Higgs Field's tools.
01:32I just left them as follows and had no issues when generating. Claude will sometimes ask you for your permission before generating images and videos, which is great as we don't want images and videos just firing off randomly and using our credits. So we're now set up.
01:45Let's follow a step by step workflow to creating our full video. The first thing that I did was ask Claude to generate a stickman image character that we will use as the face of our new channel.
01:57I also attached a reference. By the way, if you'd like a document with all the prompts that I used in this video, go ahead to the description, and you'll find that document. You can use this to follow along or copy and paste these prompts if you wanna generate some things for yourself.
02:10While Claude and Hicksfield created Sticky, our main character for a brand new YouTube channel, I simultaneously started working on the script. In my prompt, I gave Claude some context of the new channel that I was creating. I then asked for a five minute script optimized for YouTube.
02:25I then gave Claude some initial catchy ideas from viral references I found on YouTube. I did this because your ideas are oftentimes the most important factor for success. See them as the backbone of virality.
02:37So giving Claude some necessary context from YouTube of what's already working will help you get that virality a whole lot faster. I then also explained that we'd be using our stick figure as the character and finally asked it to generate some script ideas and titles before we actually get into the script generation. From this, Claude generated some genuinely intriguing ideas.
02:57How did humans survive before electricity? What would happen if all the bees disappeared tomorrow? And my favorite, which is the video we created, how did the ancient cities deal with all the poop?
03:08Then just tell Claude which idea to generate the YouTube optimized script for. Let's go with idea number four. Going back to our character generation chat, Higgsfield has designed our sticky character.
03:18Now these are all great. I asked Claude to create our pose library as per his recommendation. Our pose library is something that's gonna help with different character angles or poses that our character is gonna make, ultimately leading to much better character consistency.
03:32After this, our script was now complete, and honestly, it was incredible. But most importantly, YouTube optimized. It added this hook section, which I thought was incredible.
03:41So with our script being done and our pose library being done, I went to the pose library chat and asked Claude to download all of those poses directly onto our computer. With our character done and our script done, this is where the magic actually starts.
03:57In the same Claw chat where we generated the script, I uploaded the pose library from my computer and asked Claw to use them in the stickman video. I also said add any environment or extra detail to the sick man to help explain the scenes and keep them interesting. Claude then asked me what I'd like for my first generation.
04:14It was trying to understand if I wanted to go full steam ahead or start slow to lock in the styling. I just told it to generate stills for each scene. In a few minutes, my entire storyboard was done.
04:25Check out how incredible these images are. So honestly, these were fantastic, but there was something missing. It only generated 16 images for me, and that's too few for an entire video.
04:35But now with the style confirmed, I asked Claude to create images for each shot of our script with a max duration of each image being only seven seconds. I also asked it to generate the images in the order of the script.
04:48This is so that I could download all those images, drag them into our editor, and the sequence would be done. Claude then generated the first scene's images and asked if it was on the right track. I really like this because there was one thing that actually needed changing.
05:01I didn't want some of the images just to be black and white, but maintain that mixed illustration visual styling like the storyboard. From here, honestly, it was too easy.
05:11Claude generated all 50 images while I worked on my next video. From here, Claude wasn't able to directly download all 50 images onto my computer. So what I did was I went to Higgs Field, batch selected the images that pertain to our first video, and downloaded them.
05:27I imported them into CapCut, and guess what? All perfectly in order of our script.
05:36From here, I just recorded my voice over under my blanket. In ancient Rome, you could legally be sued for getting hit by a bucket of human poop. You could do the same thing by just pasting your script into Higgs Field's voice over tool.
05:49Then I added some music, and our video was done just like the references that are blowing up all over YouTube. So before I dive into exactly how much this entire video cost me to make, This is only one use case. I took this product image, put it into Claude, and asked Claude to develop a full marketing plan for this product image.
06:08From here, I said use this marketing plan to develop content in imagery and video for our product. I had all my marketing content for the next few weeks with two prompts. So the actual cost to generate this five minute faceless YouTube video.
06:22Higgs Field charges 4 credits per Nano Banana Pro four k generation. We generated four character hero images and nine pose library images. This is 15 generations in total.
06:33We then generated 16 storyboard or kind of directional images. And finally, we generated 50 final stick figure animations that we used in the video. 81 images in total, which times four is 324 credits.
06:47At a $125 a month, you get 6,000 credits, which equates to 2¢ per credit. When we times 2¢ by 324, this comes to a final price for all of the images and the entire video of $6.48.
07:01So let's quickly check this channel with four videos. We can confirm that, yes, ads are displaying and they are monetized. They've got four videos on their channel, which costed roughly $26 to make.
07:11And with the viewership that they've gained, they've generated about a thousand $200 in the last two weeks. $26 in a couple of hours, a thousand $200 made from their videos.
07:21But it doesn't actually end here. We still need thumbnails for our videos. We want the entire thing to be packaged pretty much on autopilot.
07:29So what I did was, now that Claude had the context for the script and it knew the visual direction of our video, I simply wrote a prompt, Claude, please generate me some thumbnails for our video.
07:40With a simple prompt, three optimized thumbnails ready to go. So let's summarize exactly what we made in less than an hour all inside of Claude. We generated a brand new character and developed his pose guide.
07:51We then generated a YouTube optimized script, which is actually interesting, and used that script to generate our entire content library. We brought that library inside of CapCut, and our images were perfectly in sequence, so we didn't need to reorder them.
08:03And then from there, we generated three thumbnails that we could use using YouTube's test and compare feature. If you're not signed up for Higgs Field yet and you wanna get involved in this automation faceless YouTube game, check out Higgs Field.
08:15There's a link in the description. With our character set up, Claude understanding exactly how we like our visual direction and our YouTube scripts, and Higgs field prep for mass production, creating an entire new video would take a fraction of the time. Half the time was just spent building that initial visual and channel direction.
08:31Now that that's done, we just hit play on mass video production.
08:39In ancient Rome, you could legally be sued for getting hit by a bucket of human poop. Because every morning in cities of a 100,000 people, citizens would walk to their windows, lean out, and just empty everything they made overnight straight into the street.
08:54And Rome was the clean one. Here's the problem nobody warns you about when you build a city. The average human produces around a 150 grams of waste per day.
09:04Multiply that by the population of ancient Rome, roughly 1,000,000 people at its peak, and you get 150 tons of human waste every single day. Now imagine, you have no flushing toilets, no sewers in the neighborhoods, no bin collection, no idea germs exist, and summer is coming.
09:23So how did ancient cities not just drown in it? For most of human history, this wasn't a problem. When you live in a group of 40 people, you walk into the bushes, you dig a hole, and you're done.
09:34The earth handles the rest. But around five thousand years ago, something broke. Humans started living in cities.
09:42Thousands of people, then tens of thousands, then hundreds of thousands, all eating, all drinking, and all processing.
09:52Suddenly, the bush and hole strategy collapses. There are no bushes, there is no soil, and there's just stone, your neighbors and a rising tide of human output.
10:02Every great civilization had to solve this, and the solutions ranged from genius to insane to genuinely tragic. First stop, the Indies Valley, around February, modern day Pakistan.
10:16These people did something nobody else would do for another four thousand years. They built private toilets in every house connected to the covered drains running through every street flushed with water from their own wells.
10:30To put that in perspective, Victorian London, four thousand years later, was still dumping raw sewage into the Thames. The Indies Valley had a better sewage system than most of Europe did in the industrial revolution. Next, Rome, famous for the Cloaca Maxima, the great drain.
10:48A sewer so big, you could sail a boat through it. But here's the twist. The sewer only served the rich.
10:55If you lived in an apartment block, which most Romans did, you used a public toilet. A long stone bench, no walls, no privacy, 20 strangers sitting in a row having a conversation.
11:06And toilet paper didn't exist. You used a stick with a sponge tied to the end, shared by everyone, rinsed in a bucket of vinegar between uses. This is, by the way, where the phrase getting the wrong end of the stick may come from.
11:22Now we go to medieval Europe, and everything Rome figured out gets thrown away. When Rome fell, the sewers fell with it.
11:32For the next thousand years, most European cities had one strategy. It was called Gardelieu, from the French Gardelieu, meaning watch out for the water.
11:42You'd shout it from your window, then you'd throw your chamber pot into the streets. This is genuinely why high heeled shoes were invented. Not for fashion, to keep your feet out of the river of human waste flowing past your door.
11:56Streets sloped down the middle into a central gutter. When it rained, that gutter became a brown stream. When it didn't rain, it became something much worse.
12:05And this, this is where the black death came from.
12:10But here's the part nobody talks about. The poop wasn't actually the thing that killed people. The poop was just the messenger.
12:17The real killer was something invisible, something every ancient city was accidentally doing, and none of them realized until one doctor in one city finally connected the dots.
12:271854, a cholera outbreak is killing hundreds of people in a single neighborhood. Everybody blames bad air.
12:35One doctor, John Snow, doesn't buy He walks the streets. He asks every grieving family one question, where did you get your water? Every death traces back to one water pump on Broad Street.
12:47And underneath that pump, just a few feet away, a leaking cesspit full of waste from a baby who had died of cholera the week before. The poop in the streets, in the gutters, in the rivers, it had been leaking into the drinking water for thousands of years.
13:02That's what killed Romans. John Snow ripped the handle off the pump. The death stopped, and modern sanitization was born.
13:10Today, we flush a toilet and we never think about it again. But every time you do, you're using technology that took humans five thousand years.
13:19A few 100 plagues and one stubborn doctor with a map to figure it out. The Indies Valley got it right first. Rome got it half right and forgot.
13:28Medieval Europe paid the price. And the modern world only works because somewhere along the way, somebody finally asked where does it all go. So next time you flush your toilets, appreciate it.
13:39It's the most underrated invention in human history. If you wanna know how humans survived a problem even bigger than this one, watch this next.
The Hook
The bait, then the rug-pull.
The video opens not with a promise but with proof: a real competitor channel with four videos, 5,000 subscribers, and 600,000 views in two weeks. By the time the host explains what he built, you already know it works.
Named ideas worth stealing.
00:00concept
Proof-First Hook
Open with a real competitor result before explaining what you built, so the viewer's trust is anchored to evidence rather than a claim.
Steal forAny tutorial where the output can be shown before the process is explained
03:19model
Pose Library First
- Generate hero character images
- Generate pose library (pointing, walking, surprised, sitting)
- Lock hero reference image
- Feed pose library into all subsequent scene generations
Before generating a single scene image, build a pose reference library so the model has a character anchor for every generation.
Steal forAny AI image workflow requiring consistent character appearances across multiple generations
04:25concept
In-Order Generation
Prompt the AI to generate images in the exact sequence they appear in the script, then download in batch, so the editor sequence is already done.
Steal forAny AI image-to-video workflow where assembly order matters
How they asked for the click.
VERBAL ASK
08:10link
“If you're not signed up for Higgsfield yet and you wanna get involved in this automation faceless YouTube game, check out Higgsfield. There's a link in the description.”
Soft mid-video CTA for Higgsfield placed after the most convincing data point — $26 in, $1,200 out — which is smart placement.
MENTIONED ON CAMERA
FROM THE DESCRIPTION
PRIMARY CTAWhere the creator wants you to go next.
OTHER LINKSAlso linked in the description.
Visual structure at a glance.
