AI Creates Videos with ZERO Human Touch
45sThe opening hook immediately challenges the viewer's perception of AI video creation by showcasing four diverse videos made entirely by AI, sparking curiosity and awe.
▶ Play Clip"Delivers exactly what the title promises—a full workflow with real examples, no fluff."
This video demonstrates how to build a faceless YouTube channel using AI, specifically Claude Code and Higgsfield. The creator shows a full workflow from research to finished videos, emphasizing the importance of detailed prompting over generic one-liners.
Connect Higgsfield to Claude Code via MCP URL to enable image, video, and voice generation.
Use Claude Code to analyze a competitor channel, pulling 14 videos, 64K subscribers, and 467K average views to find a proven topic pattern.
A vague prompt like 'make a video about coffee' produces a generic video. Adding style, voice, and structure creates a channel identity.
Generate one image to lock in the visual style before spending on a full video—easier to correct early.
Combine topic, title format, style, and voice into one prompt. Ask for a proposal before generating.
Fix a single scene by screenshotting the frame and prompting a change—only that clip is regenerated, voiceover stays aligned.
Four different styles (science, business, mythology, doodles) generated in one session without picking a model—Higgsfield routes automatically.
Faceless Studio is for template-based channels; Claude Code is for unique looks with custom research and scene control.
The key to standing out in AI video is not the tool but the prompt. By combining research, a locked visual style, a defined voice, and a structured pipeline, you can create a channel with real identity—and scale it to 10 videos at once.
What role does Higgsfield play in the Claude Code workflow?
Higgsfield provides the MCP connector that lets Claude Code send image, video, and voice generation jobs.
01:06
Why should you generate one image before a full video?
To lock in a visual style before spending time on a full video, generate a single test frame first.
06:39
What are the four pieces of the full prompt?
The four pieces are: the topic from research, the proven title format, the locked visual style, and the brand file for voice.
07:29
When should you use Faceless Studio instead of Claude Code?
Faceless Studio is for quick, template-based videos; Claude Code is for a unique look with custom research and scene fixes.
15:07
How can you scale the pipeline to 10 videos?
You can hand it a list of 10 topics and it will produce 10 finished videos in a folder.
12:34
Lock style on one frame
Saves time and money by validating the visual style before generating a full video.
06:39Vague prompt = video, specific = channel
The core principle that separates generic AI content from branded channels.
08:36Higgsfield routes models automatically
Claude Code describes the shot, Higgsfield decides which model renders it—no manual selection.
12:20Fix scenes without rebuilding
Screenshot a frame, prompt the change, and only that clip is regenerated—voiceover stays aligned.
08:52[00:00] So look at the four videos running on screen right now. Four completely different issues, four completely different styles, not a single frame of any of that was filmed, edited, or even touched by a human being.
[00:12] And all of it was made with clawed code and one good prompt. Now everyone is tying to bold videos, but the reason most AI channels still look like AI slop channels
[00:24] is that nobody spends extra time to actually prompt well. That's the part that I'm going to be teaching you today. So, in today's video, I'll show you how to build fantastic YouTube videos inside PlurCode,
[00:36] how to find a topic that's already proven, and the exact prompt structure. We're going to be doing four different video styles. We are going to be doing a start to finish, and every prompt that I'm going to be using in today's video is inside our free school community.
[00:50] The link to that is in the description below. But right now, let's get started. Now first things first, we need to connect Higgs field to Claude code. So Claude code can actually research, plan, it can write files and it can run commands
[01:06] on the machine that we are using. Now what it can't do is it can't generate visuals. It needs somewhere to send the actual generation jobs. Now that is where Higgs field comes in. And the reason why that's perfect for Claude code is because one connection involves your
[01:20] videos, your images and your voice over. Now the link to that is down in the description below so just click over there and then you can sign up and now once you are signed up just head over to Quillwards desktop app and then you're going to copy the Higgsfield MCP URL and also in the description below just
[01:36] to make life so much easier and then over here you're going to go to customize you're going to hit add custom connector and then you're going to paste the MCP URL right here and then you are going to sign in with your Higgsfield account and done.
[01:51] Okay, so from your end up, basically when code code is halfway through a bold and it needs an image, a video or a voice stack, then what it's going to do is it's going to reach into Higgsfield's mod task and then it's going to get exactly what we need. So basically,
[02:05] it isn't a tool that you have to switch to. Now Higgsfield isn't just a pile of models sitting behind this connector, it has its own faceless video tool that's actually built in and that That is called Faceless Studio.
[02:18] And that tool has its own flow of how a faceless video gets assembled together in the best way possible. So things like the scene breakdown, the images, the narration, how long each beat holds before it moves.
[02:32] So when Core Code asks for a faceless video, it isn't inventing that to a certain scratch. It's simply calling Higgs Field's tool, and then Higgs Field is going to build it through its own flow. Now, before we generate anything, there's one quick step that basically decides whether
[02:49] what we make was even worth making. Now I'm going to be pointing to a code at the channel in the niche that I want to be in, and that's basically going to be the whole prompt.
[03:03] So here's the link, and now let's analyze to make sure everything's good. Okay, this thing looks fine. And then it goes, it gets data itself, and then it fixes the pages, it goes through them, And if it hits a wall, maybe because YouTube changed their format or something, and then it's going to rewrite, it's going to rework, and then it's going to run it again.
[03:21] Immediately from that, 14 videos were pulled directly for us. And that's 64,000 subscribers, 467,000 views on an average video. And literally found the type of pattern that 8 of their videos share, and what that pattern is actually worth.
[03:37] It even caught a growth number that looks incredible. and then at hand and leave the subjects that nobody in this niche has ever touched. Now that's my topic you and I didn't have to touch a single line of it. Now that we
[03:49] got all of that let go Now here the thing this is where most people are going to start okay Now they gonna put in one line they gonna say something like Make a faceless video about the history of coffee Now that basically it There no style no direction nothing
[04:06] And now just watch what it does with that, okay? It writes the script, it breaks it into scenes, and then it's straight over into Higgs field, it's then generating the images, one for every single beat, then it's turning each of those images into a moving clip, then a narration on top of it,
[04:22] and I didn't have to ask for a single one of those tips. Now that's Higgs field running flow end to end. I gave it the sentence and then gave you a finished video. Your morning coffee began with dancing goats.
[04:36] High on an Ethiopian hillside, a herder saw his flock gobble red berries and skip wheat. Legend calls the herder Kaldi some 1,200 years ago.
[04:48] His goats grew so restless after the cherries they danced until dawn. Curious, a local monk boiled those wild berries into a bitter brew, then prayed right through the entire night, never once nodding off.
[05:04] The beans sailed across the wide Red Sea to Yemen, where farmers first grew them, and earned a fresh name, Kahwa, meaning wine. Sufi mystics sipped it to stay alert through midnight prayers.
[05:17] By 1414, it reached Mecca, and buzzing coffee houses spread across Cairo. From one Ethiopian hillside, coffee conquered the whole planet, and today the world drinks over 2 billion cups a day, thanks to ghosts.
[05:33] I mean, look, it's great, it's a good video, but it's more important to build a specific channel identity, style, and branding, because otherwise, if you don't do that, then your videos are just going to look like anyone else's.
[05:46] else's it's definitely nothing that's going to differentiate your channel from your competitors so this is a prompting problem but i'm going to show you how to fix it
[06:01] so the first thing you're going to do is you're going to add in the visual style okay so code cannot read your mind i literally can't guess what your taste is so if you leave the look unspecified then what it's going to do is it's going to default to the generic
[06:14] because generic is always the safest guesswork to make. So, I want you to watch this. I'm asking for one frame with the specific style that I want. Now, take a look at what it comes back with.
[06:26] It's very colorful and also, more importantly, it's memorable. Now, the next thing is lock it on one frame before you spend.
[06:39] Now, just notice this, okay? Notice what I just did. I didn't generate a video to find out whether the style was right. I generated one image. Now that is important because it's easier to lock in your style than correct it when generating an image rather than generating an entire video only to realize that you made a mistake.
[07:02] Now the next thing is set the tone of voice. The look is now locked in. Okay, we got that. we got that it still doesn't sound like me though so I'm just gonna prompt it with my specific tone
[07:15] of voice that I want to keep and then I have a prompt template for that down in the free school community so you can simply just go over there copy it and then you can keep following along and now that basically folds in the tone of your voice
[07:29] Now for the full prompt, okay? So, now all four pieces on one run, on the topic that the research picked for us, the proven title format, the style is now locked on a frame, the brand file for voice,
[07:45] and one more instruction that's in there, and it's basically just to show me the direction before you generate anything, now comes back with a real proposal this time a specific loop a specific pacing decision and the league field does all the rest And I mean just let take a look at the results okay Goats found coffee first Twelve centuries ago an Ethiopian herder named
[08:08] Kaldi watched his flock chew small red berries off a wild bush. They stopped sleeping. The goats danced until dawn, so Kaldi ate a berry himself and spent the
[08:20] entire night wide awake. A local monk boiled the cherries into a bitter dark drink, stayed awake through prayer and accidentally invented the magic. That is what you want. Minimal. Content that keeps people coming back. A vague prompt gets you a video.
[08:37] But if you just specify what you want, then you are going to get a channel. So now let's take a look at a way of enhancing our video because sometimes something can
[08:52] look off or a scene doesn't quite fit your style but the rest of the video is exactly how you like it. So, just take a screenshot of the frame that you don't like and then you're going to prompt the change that you want. So here I'm going to say, regenerate the scene, keep everything else exactly as it
[09:09] is. And that's basically all it touches. One clip swapped out, the rest of the video untouched, the voiceover still lined up. Perfect. Let's see if this actually works across anything.
[09:23] Okay, so four videos, four completely different styles, moving fast here because you already know the process. So first one is a science channel. Okay, I want something that's realistic, cinematic look, natural lighting, slow push-ins because
[09:36] that's always so important, muted greens and blues. Okay, ran it. is the video, okay? ...between dead. It just doesn't know it yet. And something inside is quietly rewriting every instruction its body follows.
[09:52] We have an even descent by the single spore lands on its back, dissolves through the shell, and spreads fine threads between its muscle fibres. The fungus takes over the muscles, but never the brain,
[10:05] and within days, the infected ant leaves its colony and wanders off alone. Now, there's even a thumbnail to match it. Now the second one, a business explainer. You're gonna need something that's clean, corporate, that has some sharp cuts, a lot
[10:20] of white space, confident blue and white palette. Okay, now, generated it. Now, completely different feel from the last one, okay, but sold just as impressive. The price was set in 1985.
[10:33] Inflation would put it near $4.50 today, but Costco built its own sausage plant. Now the third one is a mythology channel, and this one goes fully illustrated. So it's going to have a paper cutout look, it's going to have visible layered edges,
[10:48] it's going to have a muted telecaster and deep indigo, and everything moving in slow parallax like a pop-up book. Okay, now we're going to add that. And there we go, it even has a nice beautiful thumbnail to match.
[11:01] Six seeds were all she ate. and for that the world above answered with a frost that has returned every single year since. Persephone was gathering flowers when the ground opened
[11:15] and Hades took her down into a kingdom no living thing had ever left. Her mother Demeter searched the whole of the earth and while she looked the fields went grey and every harvest failed.
[11:29] Okay, and for the fourth one, y'all went as far from realistic as I could get. So, fun facts channel with the doodles. I have the big, blunt things nobody can explain.
[11:42] A stranger yawns, and yours follows moments later, entirely unexplained. I have a much-fummy. Deja vu hits about two of people and it strikes without any warning So laboratories can barely study the real thing There is a the vicars have no known purpose and the leading guest is a leftover reflex inherited from ancient ancestors that breathed
[12:06] through gills. So now there you can see it we have four completely different looks in one session and I never picked a model once that's Higgs field underneath doing the routing. ProWord code describes
[12:20] the shot, Higgs field decides what renders it and then all of that. The link is in the description below if you want the same setup and if you want 10 instead of 4 then all you have to do is hand it a list of 10 topics so you can walk away
[12:34] you can come back to a folder with 10 finished videos. It's the same pipeline but it's just running without you sitting there. But now there's one more
[12:47] thing that I want to show you okay so remember earlier on in the video when I said crew word code was calling his fields own tool now that tool is a faceless studio and you can just use it directly there's no connector no
[13:01] palm structure so facial studio is for people who want to work on the web okay so the first thing that it wants is a category okay so education history kids or even storytelling now for this one I'm gonna be picking storytelling and
[13:16] And then one line about what the channel is. Okay, so not about a video topic. What the channel is. And there's an upload to the button if you've already written something. And a reference if you want it matching a look that you've already got.
[13:30] Okay, now we got that. Now let's move on. Okay, so now watch this. Because this is honestly something that I didn't really expect. It's asking for the topic of the first episode.
[13:42] And it's already written five of them for me. but I didn't have to give it any of these. I took the channel idea and it came back with five episode titles that fit and if none of them land, then there's also a generate new button
[13:54] and it just simply goes again. Now down here is the aspect ratio, the length, okay, 16 by 9 for long form and 9 by 16 if you want it to be short. So the next step is choosing your style, okay. So it has a bunch like editorial, motion graphics,
[14:08] stickman cartoons, fairy tale, myth and a bunch of others. So, every one of those is an entire channel's look. So, you can basically kind of get the idea right now. Now, step three is just picking the narrator that you want.
[14:22] So, you can preview different voices, and then all you have to do is just hit generate. Now, episode one is generating right there, and when I want episode two, then all I do is I just hit create new episode, and it's already in the same file, the same voice, the same category, and I don't have to specify anything.
[14:39] They erased her name from every page. Yet, in a forgotten tower, a discarded queen counted black sand toward midnight. Years before, a blind seer had whispered a promise.
[14:54] When the last grain fell, a stolen crown would return to its rightful head. So here's how I'd actually pick between the two, okay? So Facebook Studio is when you want a channel that looks like a channel, okay? Like boss, and it's using one of those styles,
[15:07] and it's close enough to what you have in mind, then it's perfect. Now, code is when you want a look that nobody else has. Your own voice out of your own brand file, the real research behind the topic,
[15:19] files on your machine that you get somebody to point at, and the ability to fix scenes without rebuilding the entire video. Now, anything that I use today, the exact prompts, the brand file template, the research prompts,
[15:32] the pipeline set up, all of it is inside our free school community. The link to that is down in the description below, so go and grab it so you can get started today. And also, if you want to bold this yourself,
[15:45] Higgs field is what powered every single generation that you just watched us do in today's video. And now, if this video helped you, then give this video a like, subscribe to the channel, turn on those notifications, and as always, I'll catch you on the next one.
⚡ Saved you 0h 15m reading this? Transcribe any YouTube video for free — no signup needed.