Modern Creator
Joseph | Video Editing · YouTube

How to Edit a Viral Talking-Head Using ONLY AI (2026 Style)

A six-step Claude + Higgsfield pipeline that turns a raw two-hour podcast into a fully captioned, fully animated viral short — without a single manual cut.

Posted
4 days ago
Duration
Format
Tutorial
educational
Views
43.9K
2.2K likes
Big Idea

The argument in one line.

A trained AI system — Claude for judgment plus Higgsfield for image and video generation — can take a raw two-hour podcast through clipping, captioning, planning, designing, and animating into a finished short, but only after being fed the same courses, examples, and corrections a human editor would need to learn the craft.

Who This Is For

Read if. Skip if.

READ IF YOU ARE…
  • You already make or repurpose talking-head video content and want to cut the manual hours out of clipping, captioning, and animating.
  • You're comfortable experimenting with AI tools and connectors even though none of this is plug-and-play out of the box.
  • You have, or are willing to build, a library of your own past edits to train an AI on your specific taste, since the workflow depends on that training data.
SKIP IF…
  • You're looking for a done-in-five-minutes automated tool — this requires real setup time training Claude on courses, PDFs, and manually-cut examples first.
  • You don't already have raw long-form footage to repurpose — this is a repurposing workflow, not a from-scratch content generator.
  • You want free tools only — Higgsfield, Submagic, and Claude usage all carry their own costs, on top of the course being pitched throughout.
TL;DR

The full version, fast.

This tutorial demonstrates a six-step AI pipeline that turns a raw two-hour podcast into a fully animated, captioned short: Claude finds and cuts the clip after being trained on clipping courses and manually-annotated examples; Submagic auto-generates captions; Claude plans each animation using a 'crucial line vs. breather' rule learned from the creator's own past scripts; Higgsfield, connected via an MCP connector, generates branded design mockups and then turns them into motion using reference images and per-scene camera prompts; the pieces are assembled by hand in Premiere Pro. The core claim: 1-2 days of manual editing work compresses to under an hour once the AI has been trained on real editorial judgment instead of left to guess.

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.

00:0000:53

01 · Hook — "AI did everything"

Opening claim that the clip, captions, and animations were all AI-made; promises a complete beginner's guide and previews the six-step workflow using a Chris Williamson x Alex Hormozi podcast as the source.

00:5303:02

02 · Step 1 — Extract the viral clip

Claude is trained on clipping courses plus three manually-cut example shorts with voice-note reasoning, saved as a 'short form clipper skill.' An ElevenLabs JSON transcript lets Claude find three short-form candidates in the two-hour podcast, then FFMPEG cuts the chosen one into a rough cut.

03:0203:57

03 · Course pitch — AI editing masterclass

Brief reassurance that AI editing isn't as complex as it looks, into the first Ultimate Editors course CTA.

03:5705:30

04 · Step 2 — Auto-captions with Submagic

The rough clip is uploaded to Submagic, which auto-generates captions; the creator picks the 'Cali 2' style, sets font to Poppins, and exports the captioned render.

05:3008:45

05 · Step 3 — Plan the edit like an editor

Explains the editor's rule (crucial line = animation, breather = plain subtitles), then trains Claude on an inspiration-board PDF, a content-psychology PDF, and a PDF mapping past scripts to their animation choices — saved as the 'ultimate editor's skill.' Claude then produces a full beat-by-beat animation plan for the chosen short.

08:4511:15

06 · Step 4 — Design branded mockups with Claude + Higgsfield

Claude (the 'brain') is paired with Higgsfield (the 'machine') via an MCP connector. Roughly 15 Pinterest references lock in a white-and-blue brand look; Higgsfield's Nano Banana model generates a still mockup for each planned scene. A second Ultimate Editors CTA covers the full course catalog.

11:1515:06

07 · Step 5 & 6 — Animate scenes and assemble in Premiere

Each mockup is turned into a 5-second animation via Higgsfield's Seedance 2.0 model using a specific per-scene motion prompt (camera zoom, text write-on, paper-tear reveal, etc). The finished pieces are dragged into Premiere Pro on Claude's timeline map, with music set to -20dB under the voice.

15:0616:27

08 · The finished viral edit

The completed short plays in full — a message about conformity, average outcomes, and choosing to be 'extraordinary' — followed by the closing thesis and sign-off.

Atomic Insights

Lines worth screenshotting.

  • Claude cannot watch or listen to video — every workflow here starts by converting audio to a timestamped JSON transcript before Claude can do anything with it.
  • Training Claude on three manually-cut example shorts plus a voice note explaining the reasoning taught it hook and story logic faster than generic clipping courses alone.
  • Saving a trained process as a reusable Claude skill means future jobs pull up the same learned judgment instead of re-explaining the task from scratch.
  • FFMPEG, a free code-based video cutter, is what actually executes Claude's timestamped cuts — no editing software required for step one.
  • The editor's rule for where to animate a script: ask if the line is crucial for the viewer to understand — if yes it gets an animation, if it's a breather it stays plain subtitles.
  • Claude has no native image or video generation — it functions purely as the planning 'brain' and needs a paired execution tool to produce any actual asset.
  • Concrete visual references, like roughly 15 saved Pinterest designs, locked in a brand look far more reliably than a text prompt like 'make it minimal.'
  • A still-image mockup pass is done before any animation generation specifically to avoid paying for motion on a design that hasn't been approved yet.
  • Animation quality lives in timing, camera speed, and motion style — the same design can look premium or like 'AI slop' depending only on how it moves.
  • The video's stated payoff: what normally takes 1-2 days of manual clipping, planning, and animating compresses to under an hour once the AI is trained on real editorial judgment.
  • The average American adult being overweight, likely divorced, and holding under $1,000 in savings is used inside the demo clip itself as proof that 'normal' isn't a safe default.
Takeaway

Six steps to a fully AI-animated edit

TRAINED-AI WORKFLOW

A talking-head short goes from raw podcast to captioned, animated final cut once Claude is trained on real editing judgment and paired with tools that can actually generate images and motion.

02Step 1 — Extract the viral clip
  • Claude can only read text, not watch or listen to video, so the podcast has to be transcribed into a timestamped JSON before Claude can find anything in it.
  • Feeding Claude real training material — clipping courses plus three manually-cut shorts with a voice note explaining the reasoning — teaches it what makes a hook and a story work, not just generic rules.
  • Saving a trained process as a reusable 'skill' means every future job pulls up the same learned judgment instead of re-explaining the task from scratch.
  • FFMPEG, a free code-based video cutter, lets Claude make frame-accurate cuts once it has exact timestamps, with no editing software required.
04Step 2 — Auto-captions with Submagic
  • Captions are treated as the single most important part of a talking-head video because they keep sound-off scrollers watching.
  • A dedicated captioning tool auto-generates subtitles from audio, then lets you pick a style and customize font, weight, and size in a few clicks.
05Step 3 — Plan the edit like an editor
  • A real editor's rule for animating a script: ask if the line is crucial for the viewer to understand — if yes it gets an animation, if it's a breather it stays plain subtitles.
  • Animations serve two jobs at once: they explain the concept better than words alone, and they refresh the screen often enough that nobody swipes away.
  • Training an AI to plan edits took three source materials: an inspiration-board PDF, a content-psychology PDF, and a PDF mapping every past script to the exact animations chosen for it.
  • A trained model doesn't just decorate a script — it can alternate value beats and rest beats across an entire short the way a human editor paces retention.
06Step 4 — Design branded mockups with Claude + Higgsfield
  • Every creator has different visual branding, so before generating any mockup the model needs the brand rules taught explicitly, not assumed.
  • Claude has no native image or video generation — it functions as the 'brain' and needs an execution tool paired in as the 'machine' to actually produce assets.
  • Concrete visual references, like roughly 15 saved Pinterest designs, lock in a look far more reliably than a text description like 'make it minimal.'
  • A still-image mockup pass happens before any animation generation specifically to avoid paying for motion on a design that hasn't been approved yet.
07Step 5 & 6 — Animate scenes and assemble in Premiere
  • Animation quality lives in timing, camera speed, and motion style — the same design can look premium or like 'AI slop' depending only on how it moves.
  • Each animation is generated per-scene with a specific motion instruction rather than one generic prompt for the whole video.
  • Assembly is the fastest step once everything upstream is prepared correctly — it's just dragging clips and animations onto a timeline map Claude already produced.
  • The bottleneck this workflow removes is the 1-2 days normally spent finding, planning, and animating a clip — here that compresses to under an hour once the AI is trained.
08The finished viral edit
  • The finished short argues that being extraordinary requires doing things an ordinary person would judge as excessive, since most people's comparison points aren't lives they'd actually choose.
  • It backs that argument with checkable numbers — the average adult being overweight, likely divorced, and under $1,000 in savings — to make 'normal' feel like a warning rather than a safe default.
  • AI is framed as a tool that amplifies a trained editor's judgment rather than replacing it — the payoff depends entirely on the taste and material fed into it beforehand.
Glossary

Terms worth knowing.

Claude skill
A saved, reusable set of instructions and training material that lets Claude repeat a specific judgment-based task, like clipping or animation planning, the same way every time it's invoked.
MCP (Model Context Protocol)
A connector standard that lets an AI model like Claude call out to an external platform to perform actions it can't do natively — here, connecting Claude to Higgsfield.
Higgsfield
An AI platform that bundles image, video, and audio generation models behind one interface, used as the tool that executes what Claude plans and designs.
Seedance 2.0
The AI video-generation model inside Higgsfield used to turn a still design mockup into a short animated scene with camera movement and sound effects.
Nano Banana
The AI image-generation model inside Higgsfield used to turn a text description and reference images into a still branded design mockup.
Submagic
A web tool that auto-generates and lets you style word-level captions for a video directly from its audio track.
FFMPEG
A free command-line program for cutting, converting, and processing video and audio files without a graphical editor.
Rough cut
A first-pass version of an edit — here, the podcast clip trimmed to its rough boundaries before captions or animation are added.
Resources

Things they pointed at.

00:05channelChris Williamson x Alex Hormozi podcast (source footage)
02:00tool11 Labs speech-to-text
02:50toolFFMPEG
04:15toolSubmagic
09:15toolHiggsfield (MCP connector)
09:29toolNano Banana (image model)
11:20toolSeedance 2.0 (video model)
13:20toolPremiere Pro
Quotables

Lines you could clip.

00:15
if you've never edited anything with AI before, this is a complete beginner's guide
reassurance hook, works as a cold open for a tutorial seriesTikTok hook↗ Tweet quote
09:20
Claude is the brain and Higgs field is the machine
tight, quotable framing of an AI-plus-tool pairingnewsletter pull-quote↗ Tweet quote
15:20
AI on its own cannot replace an editor, but AI trained by a good editor can create stunning results in a fraction of the time.
thesis statement of the whole video, stands alone with zero context neededIG reel cold open↗ Tweet quote
15:44
A lot of people want to see you fail because it justifies the risk that they chose not to take.
the finished demo clip's punchline — proves the workflow by being genuinely quotable on its ownTikTok hook↗ 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.

metaphorstory
00:00This viral talking head video, the clip, the captions and the animations was all created entirely by AI and I didn't touch a single thing. And in this video I'm going to show you the exact AI workflow inside of Glau that I used to create this viral Instagram reel from scratch without knowing a single thing about editing.
00:17And don't worry, if you've never edited anything with AI before, this is a complete beginner's guide. I'm gonna take this raw podcast that Chris Williamson did with Alex Hermosy and run it through this workflow to get this viral talking head edit. The entire workflow is just six simple steps and with these six steps you can turn any clips that you have into viral talking head edits.
00:37But it all starts with step one extracting the short form clip. Every viral edit starts with one thing a really good clip and traditionally finding that clip is brutal. You have to watch through the entire two hour podcast take notes on every hook and every story and piece together the sections that could work as a short.
00:54And this alone can take hours. So before you start saying, oh, Joseph, now I have to go and watch this whole podcast. No.
00:59Claude can now do all of that. The only problem is out of the box, Claude has no idea what makes a clip go viral. So first, we have to teach it.
01:06So first, I went into the clipping courses that we have in my community, Ultimate Editors. These are the exact courses that we use to teach editors how to clip long form videos and I basically just gave them to Claude to study. But we didn't stop there.
01:17I took a podcast myself, cut out three shorts manually and then I gave them to Claude with a voice note explaining why I chose that hook and why I built the story that way and what I cut out. By studying this Claude learned how to pick hooks how to link clips together into valuable stories and what to leave on the table.
01:33And once Claude was a master clipper, I asked it to save everything it learned into something called a skill so that now anytime I need clips cut Claude pulls up this skill and instantly knows how to do it just like I would. We called it the short form clipper skill. Best part is you guys don't have to train Claude.
01:48You can just download this skill and it's gonna be able to cut the clips for you too. All the skills that are gonna be covered in this video are gonna be inside my school community which is linked down below. Now here's how we actually use this.
01:57There's one problem. Claude can't watch or listen to videos. It can only read.
02:01So we have to turn the podcast into text. So we head over to a tool called 11 Labs, and we go to the speech to text, click transcribe, and upload the entire podcast. Now we just wait a little, and we get this, a full transcript of every word said in the podcast.
02:14Click export here and export it as a JSON file. Why a JSON file? Because a JSON doesn't just show what is being said, it shows exactly when it's being said down to the second.
02:25Now remember that because those timestamps are about to matter. So I upload the JSON into Claude and I tell it this is the transcript of the podcast. Use the short form clipper skill to extract three high quality short form videos from it.
02:35And just like that using the hook psychology and story logic that we taught it, Claude reads through the entire podcast and hands me three short form videos that can work really well with the full script for each one. Now after reading through them, I was a huge fan of short one, so now Claude needs to cut it out of the podcast.
02:50I upload the podcast file in Claude and since it has the transcript with the exact timestamps, it knows exactly where every sentence of short one lives inside that two hour video. To make the cuts Claude uses a tool called FFMPEG, a free program that cuts video files using code, no editing software needed.
03:06Claude makes every cut at the right timestamp and hands me this, which is a rough cut of the short. Now real quick, later in the video, we're gonna show you how you can create six stunning animations with Claude. But if any of this AI video editing stuff sounds complex, don't worry.
03:19That's completely normal. AI is evolving fast, and video editing is about to change because of it. That's why if you're new to this and you'd like to learn more about how you can edit viral videos with AI, we've created a full a to z masterclass on the fundamentals of AI editing.
03:32Inside Ultimate Editors, we'll teach you everything from the editing programs to the four viral editing styles of 2026 and now how to edit viral videos with AI. If any of this sounds interesting, click the link down below to check it out. But enough of that and let's get back to the video.
03:45Now just like that, Claude just listened through our two hour podcast and turned it into a raw clip ready for us to edit. This checks off step one bringing us into step two creating the captions. Now that we have a raw clip that's actually a good clip, it's time to start editing.
04:00And the first thing every talking head edit needs is high quality subtitles. Captions are the single most important part of a talking head video. They keep people watching even when they're scrolling with the sound off.
04:09For this, we use a tool called Submagic, and just check out how easy it is. Search for Submagic in your browser, click create project, click generate captions here, select the files from my computer, and add in the short that Claude just cut.
04:21Then click this button. Submagic now listens through the clip and turns the audio into subtitles automatically. Once it's done, click the style button next to AI captions, and this is where you choose from a whole bunch of caption styles.
04:32Personally, I'm a huge fan of the Cali two style. I click on customize, and I change the font to Poppins. Use the lightweight and set the size to 28, and that gives us these amazing cleaved captions.
04:43And just like that, we have high quality captions in a matter of minutes. Then just click export up here and render the video with the captions on it. This checks off step two, and we now have a high quality clip with amazing captions.
04:54Now stay tuned because later in the video, we're going to create these amazing animations using Clog, which is where you'll truly be mind blown. And, again, please don't forget, if you feel like we're moving too fast and you'd like to learn AI video editing step by step, we're making a full AI video editing course. It's gonna be linked down below.
05:09Check it out. But now that we have a high quality edit with captions, let's move into step three, planning out our edit. Before I show you how Claude planned this edit, you first have to understand how a real editor plans a short form video.
05:19The visuals on the screen do two very important things. First, they explain what the speaker is saying. When a viewer sees a concept, they understand it way better than with words alone.
05:29Second, they keep retention. Every animation makes the video feel fresh so nobody swipes away. So a high level editor takes the script and goes line by line asking one question.
05:38Is this line crucial for the viewer to understand? If yes, it gets an animation that explains it. If it's a breather moment, it stays as pure subtitles so the viewer can rest and connect with the speaker.
05:47That is the skill I wanted Claude to have, to look at any script and instantly know where the animations go and what each one needs to show just like a high level editor. So here's how I trained it. First, I grabbed the ultimate editor's inspiration board PDF, the one that we use in our course to teach editors how to create animations and designs from scratch based on a script.
06:06And then I told Claude to study it. Then I grabbed another PDF from our community that breaks down content psychology, where animations need to sit in a video and how they emphasize the value.
06:15And I gave that to it too. Then the most important one, I gave it a PDF with every long form video we've made in the last few months and the exact animations we planned for every single line of every single script. This showed Claude how I, as an editor, actually make these decisions.
06:30Now with those three assets studied, I told Claude to save it as a new skill called the ultimate editor's skill. Again, if you guys want all these AI skills, it's gonna be inside the new course in the community. Now to actually test it, I gave Claude the script of short one and told it, use the ultimate editor skill and plan out exactly how we're going to edit this entire short.
06:48It processed for a little bit, but then it sent me this. And honestly, just take a look. For the hook, it kept pure subtitles, so the viewer locks in and connects with our speaker.
06:56The first statement talks about being extraordinary. So Claude planned an animation of a person sticking out of a crowd. The next line talks about the actions that you take.
07:04So it planned a scene of a lonely person focused in a room. Then a subtitle only section because Claude knew the viewer needs a rest. The next section talked about a bad lifestyle and its values.
07:13So it planned the design with a fat person and text. The scene after is about not caring about their opinion, a main character questioning everything. Then the script lists out three bad habits of most people, and Claude planned to show all three in a torn paper layout.
07:26Then back to subtitles for another breather, then the section about people loving to see you fail, a crowd pointing at our main character, and for the final lesson of the video, a focus scene with the lesson written out on the screen. This was mind blowing. Claw didn't just decorate the video.
07:39It planned exactly like a high level editor would, animations on the value and rest where the viewer needs it. Retention the whole way through. Again, if you guys want access to these skills, it's all gonna be inside our new AI editing course at Ultimate Editors.
07:51The link is down below. But now that we have a complete plan of how we want to edit this video, step three is complete, and it's time to begin creating the animations starting with step four, the design. Now before we animate anything, we need to design what each scene is going to look like.
08:04Why? Because animating is the heavy expensive part of this process and I don't want to generate a full animation before we've confirmed that the designs actually look good. So we gotta do mock up first and then animation after.
08:15And here's the thing about design. Every creator has different branding. Some are minimal.
08:19Some run UI style animations. Some do realistic three d. If the animations don't match the creator's branding, the edits gonna feel pieced together and cheap.
08:27So Claude needs to learn the branding first. Here's exactly how. Remember that PDF that we gave it for planning?
08:32It also covers branding and design. And on top of that Claude already understands good design because it has something called Claude design built in. This is a brain that knows how to design high quality visuals for things like websites and presentations.
08:45But here is where it gets good. Claude has one weakness. It can't actually create images or videos.
08:50It's a brain. It's not a machine. So we pair it with Higgs Field, a platform that connects AI image models, video models, and AI audio models all in one place.
08:58Higgs field can make things but it doesn't have the brains to know what to make. So Claude is the brain and Higgs field is the machine and if we put them together we can create high end branded designs and animations. And connecting them is stupid simple.
09:10You just go to Higgsfield Claude MCP. An MCP is just a bridge that lets Claude talk to another program. You copy this, and then in Claude, you click customize, connectors, and add the Higgsfield connector.
09:20Now for this edit, I wanted a minimal white theme with a blue accent, and I think it fits this video really well. But I can't just tell make it minimal and hope for the best. It needs references.
09:29So I booted up Pinterest, searched white and blue graphic designs, and spent about five minutes saving almost 15 designs that I really loved. I uploaded them all into Claude and told it to study these references because this was the branding for the client. Then it was time to design.
09:43I prompted Claude to use Hicksfield to create the design mock ups for animation one, the person working alone focused in the room. Hicksfield used an image model called Nano Banana to generate it and this is what came out. Realistic, minimal, perfectly on our white and blue theme.
09:56Then animation too and it created this person sticking out of line with the word extraordinary on top. This was looking good. So I told it to create the mock ups for everything else and it delivered all of these.
10:06The fat person scene, the questioning scene, the descriptive scene, the pointing scene, and the final concluding scene. Every single one was on brand and every single design looked impressive which meant that the animations were also going to look just as good. Now let me just pause you here real quick.
10:21The only reason I was able to get such amazing results with AI was because I spent so long training it on the material that we had in ultimate editors. You see ultimate editors was built for beginner editors that wanna learn how to edit viral videos and make $2,000 per month in the next ninety days as video editors. That's why we have the Premiere Pro and After Effects master classes to teach them the editing programs.
10:41We have four viral editing style master classes where they can learn the cinematic editing style, the minimal animation style, the SAS motion design style, and the viral animation breakdowns. And that's why they have the $2,000 per month editor blueprint where they can learn the secrets to making 2 k per month as a video editor.
10:56And now with the rise of AI, we have decided to create the AI editing fundamentals course where we will be teaching editors how to use AI to create stunning short form viral edits just like the ones that you saw today. If you wanna learn the editing programs or learn the editing styles of 2026 or learn how to use AI to edit content in 2026, Ultimate Editors is what you need.
11:16Editors join our program as complete beginners and graduate as professional editors, and a lot of them make thousands of dollars per month with editing. If this sounds interesting, click the link down below and check out Ultimate Editors two point o. The link is going to be down below.
11:29But now that the high quality branded designs were complete, step four is crossed off and now it was time to bring them to life with animations in step five. With the designs confirmed, the hard part is pretty much done. So now we need to turn these still images into moving animations that we can drop off into the edit.
11:43But just like designs, animations have style. The timing, the speed, and the way that the camera moves, that's what separates a premium animation from AI slop. So how do we teach Claude our style?
11:54Before making this video, I spent time going back and forth with Claude, generating animation after animation, and every time I corrected it, the timing, the speed, the flow, the camera until I could nail the high quality animations over and over again. Then just like before, I told it to save everything it learned as a skill, the short form animator skill.
12:13And the beauty of this is Higgs Field already has animation skills built into it. So Claude and Higgs Field speak the same language here. So now the process is that simple.
12:20I upload the mock up of the scene. I tell Claude to use C Dance two point o, that's the AI video model inside of Higgs Field, to create a five second animation with sound effects but no music, and I describe the motion that I want. For scene one, I asked for a minimal write out animation on the text and a constant dynamic camera zoom as the character steps out of line.
12:38This is usually the part that takes the most time, but after generating, this is the result that we got. And honestly, I'm pretty satisfied with it. So now we repeat the process for every design.
12:46Scene two, a slow dynamic camera movement zooming into the character and typing, and this is the result that we got, which I'm very happy about. Scene three, a dynamic zoom while the text writes out smoothly behind the character. Looks pretty solid.
12:57And now this is where it gets interesting. Scene four, I asked for a vox paper style scale in that reveals our character as the text writes out with the blue lines highlighting the special blue word, and this is what I got. It looked pretty nice.
13:08Scene five, I asked for a Vox paper style animation where the character tears through the white paper and reveals the content inside. The whole scene feels like a rough paper animation straight out of a Vox video, and this is the result that looked awesome. And scene six, a dynamic camera zoom in with a text write out.
13:22It looks clean. And finally, scene seven, a paper style animation introducing the pointing fingers while the character types in the middle of the screen, and the text writes out, which also looked amazing. And just like that, every design became a high quality animation ready for the edit.
13:37So let's recap where we are. Claude extracted the clip, Submagic built the captions, Claude planned the edit, and Hicksfield designed and animated every scene. There is one thing left which is putting it all together, and it's the easiest step of them all.
13:49Everything comes together inside of Premiere Pro. And if you've never used Premiere, don't worry. This is literally just dragging and dropping.
13:55First, I drag my subtitle short into the timeline. Then remember those timestamps from step one? Claude gave me a timeline table showing exactly which animation goes over which section of the clip.
14:04So I spend a few minutes dragging each animation onto its exact spot. And to top everything off, I drag in a music track that I liked and set the audio levels to minus 20 so it sits under the voice instead of fighting it. And just like that, everything is officially complete.
14:18This is the six step workflow that creates viral talking head videos for your Instagram. We started with a raw two hour podcast, and in step one, Claude cut out the clip. In step two, Submagic created the captions.
14:28In step three, Claude planned out the entire edit. In step four, Field generated the branded designs. And in step five, Higgs Field turned those designs into animations.
14:36And finally, in step six, we combined it all together in Premiere Pro. Doing this manually, like watching the podcast for the clips, planning the edit, animating the scenes like these, that usually takes one to two days of work. With this entire workflow, it all came together in less than an hour.
14:50Because here's the real secret, AI on its own cannot replace an editor, but AI trained by a good editor can create stunning results in a fraction of the time. And with this six step workflow, this is what the final edit looks like. The average person will always try to keep you average.
15:04It makes sense that if you want to be extraordinary,
15:07you will do things that an ordinary person would see as extra. We
15:11yearn for the approval of many people who don't have lives that we want.
15:17And so if they have a specific life, then it means that that is what they think is valuable. And if we don't want what they have, then why would we value
15:28their weight on our decisions? I think the average American adult is obese, likely to be divorced, has less than 1 k in the bank. And doing what everybody else does sounds like a great idea, but it's actually a reliable route to a life that you're probably not looking for.
15:42A lot of people want to see you fail because it justifies the risk that they chose not to take. We always have to think about listening to the people who are closest to our goals, not closest to us. So ladies and gentlemen, thank you guys so much for watching all the way till the end of the video.
15:54If you guys are enjoying this new AI content, we've got a lot of AI editing videos that are gonna be rolling out very soon. If you're a video editor and you're still fighting AI, you're playing it completely wrong. Whether you like it or not, AI is developing, and it's going to get good enough to the point where it's going to take over video editing.
16:11It's not going to replace video editors, but it's going to be the new thing that you need to be learning instead of the editing styles and just the editing programs. And I hope this video was a good guide and a good tutorial on showing you how you can use AI to create short form edits. If you guys enjoyed it, let me know down below, and stay tuned for our next video on the
The Hook

The bait, then the rug-pull.

Every viral talking-head edit starts as raw, unglamorous footage — a two-hour podcast nobody wants to scrub through by hand. This breakdown follows a six-step AI pipeline, powered by a trained Claude and a design-and-animation tool called Higgsfield, that turns that raw file into a fully captioned, fully animated short without a single manual cut.

Frameworks

Named ideas worth stealing.

00:30list

The Six-Step AI Editing Workflow

  1. Extract the clip
  2. Generate captions
  3. Plan the edit
  4. Design the mockups
  5. Animate the scenes
  6. Assemble in Premiere Pro

The full pipeline the video walks through, from a raw long-form recording to a finished animated short.

Steal forrepurposing any long-form recording (podcast, interview, webinar) into short-form content
05:41concept

The Crucial-Line Test

Go line by line through a script and ask whether the line is crucial for the viewer to understand — if yes, it gets an animation; if it's a breather moment, it stays plain subtitles.

Steal fordeciding where any explainer or ad script needs a visual versus where it needs to just breathe
CTA Breakdown

How they asked for the click.

VERBAL ASK
03:18product
we've created a full a to z masterclass on the fundamentals of AI editing... click the link down below to check it out

Soft CTA delivered mid-tutorial as reassurance ('don't worry, this is normal'), repeated more fully around 9:29-11:14 with the full course catalog ($2,000/mo Editor Blueprint, four style masterclasses) before returning to the workflow.

MENTIONED ON CAMERA
FROM THE DESCRIPTION
PRIMARY CTAWhere the creator wants you to go next.
Storyboard

Visual structure at a glance.

open
hookopen00:00
course pitch
ctacourse pitch03:02
Higgsfield MCP
valueHiggsfield MCP09:15
finished short
valuefinished short15:06
Frame Gallery

Visual moments.

Chat about this