Big Idea

The argument in one line.

Wiring ChatGPT's Codex to Higgsfield's Genjutsu video model turns object swaps, AI-avatar reels, and viral-trend recreations into single-sentence prompts, letting one person do work that used to need a camera crew and an edit team.

Who This Is For

Read if. Skip if.

READ IF YOU ARE…

A solo creator or small business owner already making short-form video who wants to multiply output without hiring an editor or VFX artist.
Someone comfortable typing detailed instructions to a chat agent and willing to pay per-generation API costs (roughly $12-$34 per clip in this video).
A business owner curious about AI avatars or trend-jacking as a lead-gen or ad-testing channel.

SKIP IF…

You want a no-cost workflow. Every edit shown here runs through paid Higgsfield API credits, not a free tool.
You're looking for organic, unedited talking-head content. This whole system is built around replacing raw footage with AI-generated versions.

TL;DR

The full version, fast.

Samin Yasar walks through connecting ChatGPT's Codex to Higgsfield's Genjutsu video model so text prompts, not manual editing, drive three levels of AI video production. Level one is a single object-swap or motion-transfer edit run through the Higgsfield API for about $12. Level two adds two plugins, HyperFrames and a custom Reel Engine, that strip pauses and mistakes from raw footage, insert motion graphics, and can replace the creator's face and voice with an AI avatar for about $34. Level three mines Higgsfield's console for pre-built viral-trend templates, strips copyrighted audio before motion transfer, and re-splices it after, so a trend recreation with celebrity reference images costs about $20 to produce.

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 →

Chapters

Where the time goes.

00:00 – 01:07

01 · Intro

Cold open showing the finished AI-edited output, then a roadmap of the three levels the video will cover.

01:07 – 03:08

02 · Why This Matters

States the problem with manual video production: every fix costs time or money, every edit needs a different skill, and big ad accounts already outpace solo creators on volume.

03:08 – 04:36

03 · Level 1: What Genjutsu Is

Explains Higgsfield's Genjutsu model: motion transfer versus object swap, input limits, resolution, and why running it through ChatGPT is the easiest access point.

04:36 – 06:11

04 · Setting Up Genjutsu

Walks through installing the ChatGPT desktop app, opening Codex, and connecting a Higgsfield account either via the MCP plugin or a pay-per-use API key.

06:11 – 09:57

05 · Your First Genjutsu Video

Live demo: films a raw clip, prompts ChatGPT to object-swap a bottle for an orb and change a shirt after a snap, and shows the finished before/after result for about $12.

09:57 – 14:28

06 · Level 2: Making AI Instagram Reels

Installs HyperFrames and a custom Reel Engine plugin so ChatGPT can cut filler from a raw take, add motion graphics and B-roll, and assemble a finished reel automatically.

14:28 – 17:00

07 · Swapping In Your AI Avatar

Extends the same reel workflow to replace the creator's face and voice with an AI avatar, for use cases like being camera-shy or running multiple presenter variations, at about $34.

17:00 – 22:25

08 · Level 3: Recreating Viral Trends

Pulls a pre-built viral template from Higgsfield's console, strips the source audio to avoid copyright flags, motion-transfers in reference images (including a Steve Jobs likeness), and re-splices the audio for about $20.

Atomic Insights

Lines worth screenshotting.

Higgsfield's Genjutsu model combines two capabilities no single other model does: motion transfer, which keeps a clip's camera and timing but rebuilds the person and setting, and object swap, which changes one element while leaving the rest untouched.
A Genjutsu edit accepts a 4-to-30-second source clip plus up to 30 reference images and outputs at up to 1080p, which sets the practical ceiling on what one edit can cover.
Connecting a Higgsfield API key to ChatGPT's Codex tool lets a creator drive video edits with a typed sentence instead of learning Higgsfield's own interface.
A one-sentence object-and-wardrobe swap on a short clip cost about $12 at Higgsfield's discounted API pricing.
HyperFrames and a custom 'Reel Engine' plugin let the same chat agent strip ums, pauses, and retakes out of raw footage while inserting B-roll and motion graphics in one pass.
Swapping a creator's face and voice for an AI avatar across a finished reel cost about $34 in generation fees, roughly triple the cost of a single object swap.
Higgsfield's console keeps a running library of trending templates with the exact prompt and reference assets already attached, so recreating a viral format means copying a known recipe instead of reverse-engineering it.
Stripping a source clip's audio before motion transfer, then re-splicing the original audio afterward, avoids the copyright flag that comes from AI-generating over someone else's copyrighted song.
A full trend-recreation video, including sourcing celebrity reference images and stitching multiple generated clips together, ran about $20 in generation costs.
Telling the agent an explicit end goal before handing off a task pushes it to keep iterating toward a finished result instead of stopping at a rough first pass.

Takeaway

Video editing collapses into typed prompts

WORKFLOW BREAKDOWN

Chaining a video-generation model behind a general chat agent replaces manual editing skill with the ability to describe, in plain language, exactly what should change in a clip and why.

02Why This Matters

Manual video production used to charge for every fix: reshoot for a wardrobe change, hire a VFX specialist for an object swap, repeat the whole process for every new ad variation.
Large ad accounts already run hundreds of creative variations a week, so single-operator content only competes by finding a way to generate that many versions without a team.

03Level 1: What Genjutsu Is

Motion transfer keeps a clip's camera movement and timing but rebuilds the person, place, and look from reference images, so the original performance survives with a new appearance layered on.
Object swap changes one element in a video and leaves everything else untouched, avoiding a full re-shoot or re-generation for a single wardrobe or prop change.
Source clips can run 4 to 30 seconds with up to 30 reference images and output at up to 1080p, which sets the practical ceiling on what a single edit can cover.

04Setting Up Genjutsu

Routing a video-editing model through a general chat agent's coding environment lets a creator drive edits with natural-language prompts instead of a dedicated editing interface.
A pay-per-use API key exposes the exact cost of every generation up front, which matters once edits are billed per clip rather than folded into a flat subscription.

05Your First Genjutsu Video

Two literal instructions inside one prompt, replace this object and swap this garment after a specific gesture, produced two separate generated clips that were then stitched together automatically.
A single-sentence object-and-wardrobe swap on a short clip cost about $12 at discounted API pricing, setting a real per-edit price point for testing the workflow.

06Level 2: Making AI Instagram Reels

Installing formatting and pacing plugins ahead of an edit lets a chat agent apply a consistent visual style instead of reinventing motion-graphics choices on every clip.
The same agent that edits video can also strip verbal filler, pauses, and retakes from a raw take, collapsing two separate editing passes into one instruction.
Stating an explicit end goal before handing off a task pushes the agent to keep iterating toward a finished result instead of stopping at a rough first pass.

07Swapping In Your AI Avatar

Because the same model can already swap a person into a video, extending that to a synthetic voice turns one successful ad into a template for many different presenters.
Replacing a creator's face and voice with an AI avatar across a finished reel cost about $34 in generation fees, meaningfully more than a single object swap.

08Level 3: Recreating Viral Trends

A running library of trending video templates, each with its original prompt and reference assets attached, turns 'recreate this format' into copying a known recipe instead of reverse-engineering it.
Stripping a source clip's audio before motion transfer and re-splicing the original track afterward sidesteps the copyright flag that comes from generating video over someone else's copyrighted song.
A full trend recreation, sourcing reference images and stitching multiple generated clips into one video, cost about $20, roughly midway between a single object swap and a full avatar reel.

Glossary

Terms worth knowing.

Genjutsu: A video model from Higgsfield that edits existing footage two ways: motion transfer (keep the camera and timing, rebuild the person and setting) and object swap (replace one element, keep everything else).
Motion transfer: An edit that preserves a source video's camera movement, timing, and performance while replacing the person, place, or overall look shown on screen.
Object swap: An edit that changes a single item in a video, such as a prop or piece of clothing, while leaving the rest of the footage untouched.
HyperFrames: A plugin installed into ChatGPT that gives it a consistent visual style and pacing to follow when assembling a finished video edit.
Codex: ChatGPT's agentic coding environment, used here as the interface for running Higgsfield's API from plain-language prompts instead of a dedicated video-editing UI.

Resources

Things they pointed at.

03:09toolHiggsfield / Genjutsu ↗

05:00toolChatGPT desktop app (Codex)

10:10toolHyperFrames plugin

10:15toolReel Engine plugin (Samin's own)

00:00linkGenjutsu resources + prompts bundle (from video description) ↗

00:00linkWork with Samin / Skool community (from video description) ↗

Quotables

Lines you could clip.

00:02

“You no longer need a camera, an editor, or even your face to make content that gets hundreds of thousands of views.”

cold-open thesis, works as a standalone hook→ TikTok hook↗ Tweet quote

03:00

“It's basically your AI video editor and whole content team, and you never have to leave the app.”

clean summary line of the whole video's value prop→ IG reel cold open↗ Tweet quote

09:50

“These generations cost me about $12 with Higgsfield's new API pricing because everything is 50% off.”

concrete price point makes the claim verifiable→ newsletter pull-quote↗ Tweet quote

16:50

“That kind of a reel cost about $34 because it had to remake me, make my voice, do all the animations.”

second concrete price anchor for the avatar workflow→ newsletter pull-quote↗ Tweet quote

The Script

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.

metaphor

GPT -6 just changed content creation forever because you no longer need a camera, an editor, or even your face to make content that gets hundreds of thousands of views. You can take any video, maybe a one that you filmed or something that's already going viral, and swap out the person, the outfit, the product, anything in it without having to reshoot a single thing.

And I've actually been using this to edit my YouTube videos and post Instagram reels for myself and an avatar. And they're hitting hundreds of thousands of views and getting my business way more sales on autopilot. So in this video, I'm going to be breaking down how you can get started in three levels.

First, we're going to get set up. I'll walk you through the basics of creating content with AI, and I'll show you how to make your first video. Then we're going to be taking things further with motion transfer, object swap, dynamic voice.

I'll show you how I use each of them in my content and how to bring them all together and make videos for YouTube and Instagram. And then level three, we're going to copy videos that are already trending and going viral. And I'll show you how to turn that into views, ads, and even a service that you can sell.

And even if you've never touched an AI video before, you'll be able to do this because all we're doing is simply talking to ChatGPT. Let's get into it. Okay, before we get into anything, I need to give you some context on why I'm actually making this video right now and why this matters.

Because for the past few years, I've been deep in the world of content production and here's the thing I keep coming back to. Until now, the way you won in content and ads was just to throw more money at it or waste an enormous amount of time learning 15 different tools just to get a decent result. And the problem really came down into three things.

First, every step would cost you time or money. So imagine you've got some really good footage that you like, but you need to change something small in it or put something in it. And normally to do that, you'd need to reshoot the entire thing.

And that reshoot would cost a lot more time and money to have to do it all over again. Second, every step you would need to edit that video would need a different tool or a different skill. And swapping something out in a video, you'd need a whole VFX team or you'd have to be one.

And now you can just do the same things with just one sentence. And third, every new video now means repeating the same process over and over because the big marketing agencies are testing 300 ads a week while you may be still working on one draft. But now you can start from what's already working and make your own variations in just minutes.

And the fix is having all of the tools in one place and getting ChatGPT to do and use all these tools without you knowing how. And the best part is that everything happened in just one app. So all the images, the videos, the edits, and you can even deploy what you want or what you build from there.

So it's not just a chatbot helping you write stuff anymore. It's basically your AI video editor and whole content team, and you never have to leave the app. And that's why you don't need a studio or a big team anymore.

The new advantage is knowing how to use these tools the right way. And that's what this video is. I'm going to be sharing what I've learned after testing this stuff for hundreds of hours.

So you get a massive head starts and I get some subs. So with that, let's get into level one. All right.

So for level one, before we set anything up, I want to make sure you understand what we're actually doing here and the new tool we're going to be using called Jinjutsu. What Genjasu is, is this new video model introduced by Hicksfield and it lets you edit the videos you already have and it does that in two ways. So the things it's good at and it's the only model that can do this is one, it can do motion transfer.

So it keeps the motion, the camera and the timing from your video and it rebuilds the person, the place and the whole look from the images you give it. So think of it like an actor stepping into a new costume. And they'll make the same performance with a completely different look.

And I'll show you some examples later. And the second one is object swap. It changes one thing in your video and it keeps everything else exactly the same.

And it's like swapping out one ingredient instead of cooking the whole meal again, where we had to in the past regenerate entire videos or we mess one thing up. Now we don't have to do that anymore. And all you would need is to give it the clip anywhere from four to 30 seconds, and it can pull up to 30 images of what you want.

And the resolution, you can get up to 1080p. But the biggest thing is how we should use it. And the easiest way to use this is straight from ChatGPT.

And once we connect these things together, you don't have to learn any new buttons. We just tell ChatGPT what to do, and it runs Jinjitsu for you. So for that, we're going to need the ChatGPT desktop app.

So what you want to do is make sure after you're in ChatGPT, you open the desktop app, get that downloaded, and then you will see a scene like this. And what I want you to do is go into Codex. And that is the most powerful way we should be using our ChatGPT.

And to do that, what we're going to do is go to the top left, hit Codex right here, and then we have Codex. Cool. Next thing we want to do is make sure we have a Higgs field.

account and you don't have to pay for any of the subscriptions all right but i'll show you both ways to get it so the first way is the easy one you can just install the hicksfield plugin if you don't know where that is if you just go hit mcp right here and then hit install this plugin and you will see chatgpt will authenticate with your hicksfield account now your chatgpt and your hicksfield account are connected and if you don't have a subscription to hicksfield What I want you to do is make sure you go to hit this button called API.

And what API allows you to do is use all these models, one, at a discount, two, in a pay -per -use way, such that any videos you generate, you can do it knowing all the costs associated with it. And you can do that by hitting API keys right here, creating a new API key, copying that, and then you copy this. You go back to your ChatGPT codecs, and then you paste this in.

make sure you use a low model something like gbt 6 soul and then hit go this will enable your chat gpt to use your higgs field account and now you should be able to create videos from within chat gpt using higgs field all right Cool. So now let's actually make our first video.

What we did now is made sure we're set up. So we connected ChatGPT to Higgs field and now we have access. So actually went out into my kitchen and filmed this quick video about me showing off this thing and then doing a little motion to show you when things will change.

All right. And let me show you that. Okay.

So I have this item right here. I don't know what it does. It's kind of cool.

And you know it's kind of magic where if i snap my fingers it can do that so try it out see what it can do and um yeah cool so probably you've noticed is i said i snapped my finger and then i changed my clothes and i just pre -meditated all the stuff and now what we want to happen is ideally this changes into like an orb or something and then after i snap then my clothes change to something let's say i'm trying to advertise so let me show you how to do that and doing it is really simple so first i have my test footage right here and i'm going to go to chatgpt then i will take this test footage and drop it in here okay then i want to make sure i have a much a little better model and then ask it hey we just connected higgs field api and plugin what i want you to do is take this test footage and then do two things i want you to use Object swap to replace the pink bottles that I have into like a purple magic orb and then when I snap and the video such that it changes my T -shirt into this random hoodie that I found online So just google this and then put it in and then what I want to say I Want you to also create another video that shows the before and after

side by side on the left put the before and then on the right put the after okay cool and that's it i want to make sure you understand what we just did where i asked it to use the api and then take the test footage and then do what i wanted using the right model all right so once this is done let me come back and show you how easy that was great so it just got done and here it uploaded it to my google drive because i have that as an automation but let's watch it okay so i have this item right here i don't know what it does it's kind of cool and you know it's kind of magic where if i snap my fingers it can do that so try it out see what it can do and um yeah so that was so cool right you see how after i snapped that's when i got the hoodie these are little things that actually hicksfield didn't do what happened was ChatGPT made two clips one with me wearing the full hoodie and one without and then it sliced it and did all the video editing to get it into one place where after I snap it's showing that part of the clip and that's why we need to be using ChatGPT with something like Higgs field and it makes everything so much easier when you're building these videos and essentially what you get to do is

You know, I said swap the bottle for an orb and then I add the suri. It did all those things and all those little steps that I didn't have to do. It got out and did it all for me.

And if you want to do something like this, make sure to use something like GPT -6 Sol or even Astra. All right. But this is just level one of us using genjutsu to do one small thing.

And this is super useful. But let me show you how to take this to the next level. And if you're curious.

These generations cost me about $12 with the HixFuel's new API pricing because everything is 50 % off. So I want to make sure you guys abuse this while you can, all right? All right, so next I want to show you how I actually make stuff like these kind of Instagram Reels that you see are getting tons and tons of engagement.

And it's with AI doing a lot of the heavy lifting and editing, right? And you can see all these motion graphics that are happening, swapping me in and out of the screen. AI can do all of that, and I'm going to be showing you how.

And you can even swap yourself out for an avatar in level two. So let's get into that. And to build those kind of reels, this is the tool set we're going to be using, right?

So first, obviously, we have to film a script on camera. So that's the raw performance. So right here, you can see I filmed this video, and it's just a raw take where I have my mistakes and all that stuff.

So let me just show you a quick clip.

So you can see I'm making a lot of mistakes, pauses, retakes, all these things, right? We need AI to get rid of that. We need to make sure it's adding motion graphics and stuff and all the animations so it's way more engaging, right?

So let me show you how we can do that. So first things first, what we want to do is make sure we give this to ChatGPT. So I'm going to give this drive link to my ChatGPT.

We want to make sure we give and install this thing called HyperFrames. And HyperFrames, you can install it by using this thing right here. You can go in and say, make sure to install HyperFrames.

Okay. And then And then I have this plugin and I call it some in real engine plugin.

And I use this to give ChatGPT the skills to edit all my videos. So all I did is just, I pasted all these things in and let's use a little better model for this. So it does a little better.

Okay. So right now I gave ChatGPT all the inputs I needed to, then now I need to describe what it should be doing with each of these inputs and how it should be editing the reel all right so let's get this done hey what i want you to do is use hyperframes install it i've put down the skill and then also install the skills from my reel engine and what i want you to do is using hyperframes and the reel engine that's how i like my videos edited i want you to edit the drive link i gave you make sure to cut out all the ums ahs pauses retakes and also put in the relevant B roll, the transitions, everything based on how I edit my videos in the real engine plugin.

Right? So that's your goal. Make sure you do that and send me the output and put it inside Google drive.

Okay. So I'll send this out and then I'll show you the result in my Google drive. Notice I did a couple of different things.

I gave it all the different things it needs. And if you need any of these skills, guys, links will be in the description below and. Make sure I've put down all these skills.

And I said, that's your goal. This little thing that I just said makes it loop until it's good and ready. All right.

So make sure you always, whenever you make your prompt, say stuff like this, and this will make your generations way better. Right. But with that, let me come back with the finished result and show you.

Okay, so it opened up here, but I'm just gonna open in my Chrome, but okay, this is a little better So let's look at it. This map gives cloud code 600 ,000 real designs to steal ideas from it's called mob in and instead of asking Claude to guess what a good website or app should look like you can connect it to a library of UI screens from real products people actually use every day So you see it did a lot of different things it basically made a lot of these motion graphics it made the full screen motion graphics it took me out it cut out all my ums ahs pauses retakes and it made a full reel and it's using the skill that i showed you guys that's where bulk of the effort went into and building that took a long amount of time but i want to make sure you guys have it so make sure you grab these things because they're super useful whenever you're editing your videos use Hyperframes so you can actually give your ai some sort of taste and it knows what it should make the end output look like.

All right But there's something even cooler you can do with this now with this What I can do is if I go back to chat gpt and start a new chat and then paste this in And what I can say is right now because jinjutsu can actually switch people in avatar It should also be able to switch my voice and my character. So if you have an ad that did really well, what you can do is you can swap out with many different avatars and then run those on autopilot.

And let me show you how cool this looks. So, hey, what I want you to do is use Higgs field Jinjutsu to create an avatar. Maybe it looks like Jaden Smith or something, but what I want you to do is put him as the main character in the video, swap me with him, and then also make his voice different from mine.

Okay. then open it up in my google drive cool as simple as that remember i told it exactly what to do and because i'm telling it which models to use and how to use it it's a lot better at actually getting this stuff done so after this is done let me show you what this looks like great so it got done and you can see already it has someone that looks like me what is not me and it even has the 100 million dollar leads behind it but let's take a look at a couple seconds Okay, really cool, right?

Now, even if you're scared to make content and stuff like that, if you really are, then you can even use avatar that looks and sounds somewhat like you and you can start putting in variables that sounds more and more like you to actually start producing content and getting some results and leads for your business. So this was a really cool hack that I found on how you can actually build these kind of reels and level it up using something like hyperframes and Hicksfield Jinjutsu.

So you can have an avatar with animated reels that you can use. to make content on autopilot and this is really mind -blowing to me when i first did it and i recommend you try it out while this is so cool because it's only just going to get better from here and if you're curious doing that kind of a reel cost about 34 because it had to remake me make my voice do all the animations all this stuff and that's the cost if you don't want to be on camera and want an avatar to be making reels and videos like that for yourself at a very high quality All right, so now what I want to show you is if you've been scrolling your timelines and seeing these kind of memes, I want to show you on how you can actually put yourself in it, recreate these memes, and even make ads out of them, advertise any product, and actually capitalize on all the attention these kind of trends are getting.

So let's get into that. But the thing is, the problem with these kind of trends, there are always going to be some copyright issues. And I want to show you some hacks on how you can actually skirt those.

So you can make these for yourself. And to show you the hack, if you actually head into Hicksfield's console, you can see they're always having all these little trends and you don't need to pay for any of them. Let's say recreate this one.

You can always hit recreate and you will see they have their prompt right here on how you can remake them. And then the assets right here, right? Okay, this is cool.

But I want to show you is the inputs that we take and we're going to be giving to ChatCPT so we can recreate it from us in one place. And notice how this is the source video. So this video is where we are going to be doing our motion transfer onto these characters.

And that's how we actually make our video. Right. So obviously what we need to do.

is we can download this video right here you can see if you needed to find a video you can find it from higgs field and then drop in your video that you just downloaded and actually queue up some releases right so what i'm going to say right now is first i need to make sure i am in this video right and let's say we also want to put steve jobs because i just bought a mac ultra and it was very expensive so what i'm going to do is i'm just going to go google a picture of steve jobs ideally you look for one that's pretty high resolution and you can copy and paste that in right here and then and then what i did was made a reference sheet of myself kind of just like this it's not hyper accurate but you know it'll do for now and you can see what i can do is i will put this in here and now what you want to do and this part is really really important right really understand the steps i'm about to take because what i'm going to say is this is the reference video but i'm going to say remove the audio and then create

the video with these characters the reason we're going to do that is because whenever you have the audio inside and it's a popular song or something they're going to be copyrighted so to avoid getting copyrighted what you want to do is remove the audio and then only do the motion transfer Download that video and then stitch them all together.

Okay, so that's exactly what I'm gonna be telling my prompt to do Hey, what I want you to do is I've attached a video and the pictures of two people What I want you to do is have the image on the first person to be the singer Replace and use Higgs field Jinjutsu and use motion transfer to get this done Okay, I want you to make sure they're both wearing suits and not what they're wearing in the pictures and also put the man in image two as the second person on the right inside the video but it's very important for you to take these steps to get me the final video and what i want you to do is first remove the audio from the test video and when you're using Hixfield Jinjutsu upload the video without the audio and then upload all the reference images then when you generate the motion transferred video and it's done i want you to re slice in the audio and give me the final product and upload that into google drive and also um make two variations of this video actually cool so this is what i have and it's a little heavier of a task so i'm going to do this on astra high and then i'll show you what this looks like once it's done but

i basically outlined all the steps it should need to take to get me that final finished product into google drive so this will probably take a while but i'll be back after it's done great so it got done so let me show you okay so that's pretty cool right now you can actually have these things inside your video and you can notice there's actually like a little Apple watch right here and say we're advertising a watch or a piece of clothing or something of the sorts you can always upload those as references and then start building these videos and using them and posting them as ads and linking them down below that's a really cool way where you can trend jack and start pitching to other companies like hey i can make these things for you and it becomes super easy to get new clients because they know you're on top of your stuff and you know where and which tools to use like in hicksfield where you can find these but making this thing was very expensive making this video cost about 20 or so so if you do want to learn how to make videos like this but for cheaper using all the optimizations i know watch this video right here

The Hook

The bait, then the rug-pull.

Samin Yasar opens with a claim built to stop the scroll: GPT-6 just changed content creation forever, no camera, no editor, not even your own face required. What follows is a live demonstration of exactly how, three levels deep, with the dollar cost of every generation shown on screen.

Frameworks