Modern Creator
Stockton Walbeck · YouTube

The Truth About Claude Video Editing

Five things AI video editors still can't do, and the workarounds a 10-year editor uses to ship Claude Code edits.

Posted
4 days ago
Duration
Format
Listicle
educational
Views
5.7K
87 likes
Big Idea

The argument in one line.

AI video editors produce great work only when you supply creative direction, real assets and a revision budget, because they can't read your mind, fetch what you point at, or nail the first cut.

Who This Is For

Read if. Skip if.

READ IF YOU ARE…
  • A creator or coach publishing talking-head videos who is considering Claude Code or another AI tool to replace a human editor.
  • Someone who tried handing an AI editor raw footage and got generic stock visuals back, and wants to know why.
  • A team lead comparing AI edit costs and turnaround against freelance editors billing $25 to $500 per video.
SKIP IF…
  • You want a step-by-step Claude Code setup or prompt walkthrough; this is expectations and workflow, not a tutorial.
  • You edit vlogs, weddings or music videos, since the editor discussed is built for single-speaker talking-head content.
TL;DR

The full version, fast.

AI can edit a full talking-head video, but only if you set realistic expectations. The creator lists five limits: it can't read your mind, so give direction and reference videos; it can't invent real footage, so hand over anything you point at; it isn't free or instant, with long form eating about 25% of a $200 weekly Claude Code allowance and 5 to 12 hours; it can't edit every format; and the first cut is never final. His workaround is direction plus real assets, then revisions, cutting cost from roughly $500 and ten days to a fraction.

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 →
Atomic Insights

Lines worth screenshotting.

  • If you point at it in your script (this video, my Canva file, here's what others post), AI can't see it, so you have to hand the asset over.
  • The best results come from a middle path: give the AI a starting storyboard and direction, then let it fill only the gaps.
  • AI editors don't ask you for missing assets; whatever you supply is all it works with, and the rest becomes made-up stand-in visuals.
  • A human editor previously cost about $500 and 5 to 10 business days per long-form video, which is the real baseline AI is measured against.
  • On a $200 Claude Code plan, one short-form edit uses about 2% of weekly usage, while a long-form edit uses about 25%.
  • The first AI long-form edit took about 36 hours including revisions; improved rules and models cut it to around 12 and then under 10.
  • Rendering still happens on your own machine, so GPU, memory and disk space matter as much as the AI software.
  • An edit that takes 10 minutes on an M1 Studio can take an hour or two on an older computer.
  • AI editing works best today on a narrow content type, such as one person speaking to camera, rather than as a universal black box.
  • Claims that an edit was done in one shot with zero revisions are sensationalized; even the featured video took eight revisions.
  • A screenshot of what you want is a valid revision prompt, and the AI can analyze it and redo the layout to match.
  • Revision count tracks your taste: a particular creator might need eight rounds, an easy-to-please one maybe two.
Takeaway

Direction, assets and revisions decide quality

WHAT TO LEARN

AI editing quality is capped by what you feed it: clear direction, the real assets you reference, and a willingness to revise.

  • Give the AI creative direction and reference videos up front, then let it fill gaps, instead of leaving every visual decision to it.
  • Anything you point at while speaking needs to be handed over as a file or URL, or the AI substitutes generic stand-in visuals.
  • Budget real cost and time: about 25% of a $200 plan and 5 to 12 hours for long form, versus $500 and ten days with a human.
  • Match the tool to the format; the editor described is built for single-speaker talking head content, not vlogs or music videos.
  • Expect several rounds of revision, and use screenshots or example videos as feedback to steer the look toward your taste.
Glossary

Terms worth knowing.

Storyboarding
The AI step where the transcript is split sentence by sentence and each line is matched to a visual that fits its meaning.
Talking head content
Video of one person speaking to a camera, usually educational, with or without a script.
Real asset
A specific file, URL or video you actually own or reference, as opposed to a made-up stand-in graphic the AI generates.
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.

Yes, AI edited this entire video, and yes, it can edit yours too. Everybody knows what AI can do, but what about what it can't do? There are five main things AI video editors still cannot do, but if you know how to work around it, it will allow you to produce videos at the quality you're watching right now.
I've been a video editor for 10 years. I've hired and trained over 100 human editors, and I'm launching my own AI video editor, which you can see in the link in the description. But whether you use mine, someone else's, or make your own, here's five things AI won't do.
and how I get around them. So to start, AI cannot read your mind. Now, what do I mean by that?
Okay, you have to give AI editors, just like human editors, creative direction. You and I both have expectations, right? I have an expectation right now as I'm filming this of how this long form video is going to go.
And because you know what you want, you have to tell it what you want. For example, for this video here, I'm trying to give it some creative direction of how I want it to flow. It's a listicle.
I'm doing five things AI can't do. So that motion graph. you saw come up here, I gave it that idea because I have that preference.
I also gave it these two YouTube videos, one of them being mine, for it to analyze and get a feel for the types of frames and how the video flows because it's going to be very similar to this one. Now, don't be mistaken here. AI video editors, including mine, do storyboard videos.
Storyboarding is a process of taking what I'm saying verbally, transcribing the whole thing, and then taking every single sentence and word you say and trying to match it with a visual contextually. And that is a skill, right? AI has to decide, okay, hmm.
What could I show visually from this timestamp to this one? And it does create some exceptional stuff, but you should not rely on it 100 % of the time. I believe this Goldilocks zone is a combination of the two.
You should give AI a starting point, give it some direction, and then just have it fill in the gaps later. Versus having AI fill in all of the creative gaps, which if you do that, you're going to get a bunch of just made up visuals that you may not like, which is exactly what point number two is. Number two is it cannot invent footage out of thin air.
When it comes to footage, assets, graphics, graphics, visuals, you name it, whatever's being shown on the screen. I like to split into two types of those.
There are what I call real stuff and then made up stuff. Okay. I can't really think of a better name to describe those two, but essentially what happens is AI is going to transcribe your video.
It's going to storyboard it and come up with the visuals itself. But as a perfect example, okay, let's say that in this video, I say. I built over 100 framework templates on Canva.
Okay, that sentence of I built over 100 framework templates on Canva is something I built. It's a Canva template. I built my own hands, but where is it?
How is AI supposed to know what that Canva template is, what's in it, how to extract from it? Is it going to screen record that tool? Is it going to, you know, video record it?
What is it going to do? That would be an example of a real asset. If I don't physically give that Canva template to AI, the URL.
It would not have been able to show these visuals it pulled from it. And instead it might've showed a stand -in graphic, a made up graphic or visual like this. Do you see the difference?
They're not the same thing. Or maybe it would have had to use a Canva logo that's kind of generic like this. or maybe a cartoon looking version of what a Canva file might look like, like this.
You know, who knows? Like the point is, you know, and that's a real asset you have. And so you have to supply it.
And it shouldn't come to any surprise to you that a human editor works the exact same way. Do you know how many times I have had my video editors say, hey, Stockton, at five minutes and 34 seconds, you mentioned a Canva template. Can you please send that to me?
And it's like, oh, my bad. Like if I don't give it to you, what can I expect them to do? And AI, after it looks at your video, is not going to ask you for those assets.
Whatever you give it is what it's going to work with. So I fill in that hole by giving it what I already knew I want. But on the flip side of that, some of you like really just may not have real assets and you don't really care.
And you kind of just want AI to just do its thing. Go find some stock footage B -roll, right? Go find some cartoony looking animations to grab.
That's fine. And if you don't care, you don't care. That's fine.
It's your video. But another example, right? Let's say that I said.
This Instagram video right here got 3 million views. This Instagram video showing on my screen right now, I had to supply that because contextually, how would AI know which video on my hundreds got 3 million views? I could leave it up to it.
It's an option. You can do whatever you want. But by telling AI, hey, in the part where I say this Instagram video here, I want you to show this one, this URL, download it.
put it inside of my video file and then I want you to show how many views it has on it and then a motion graphic of the numbers going up. So do you see how I just did two things? Number one, I gave it the creative direction and I gave it the real asset.
Those two combinations is how I'm able to get my videos looking the way they do because I'm giving AI the direction and then I'm having it fill in the gaps where I don't give it much. You get the idea, okay? This is the biggest time sink if you do not do this step.
I have a lot of people who download and use AI tools like this, and they just give it a piece of footage and some audio and say, make me a video. It will make you a video. But the quality of that video is actually determined by your input and your direction.
As an easy rule of thumb, if you point at it, hand it over. So anytime you say this or my or here's, you're pointing at something, right? This Instagram video, my Canva file.
Here's what other creators are posting. AI cannot see what you're pointing at or what you're trying to reference. And so you got to give it to it and supply it.
Okay. Next thing, and this is probably the best reality check for a lot of people. This is not free and it is not instant.
Right now, at the time of recording this video, I use cloud code to edit all my videos. Maybe Codex will do it in the future. Maybe there'll be other AI tools that do even better.
Who knows? We'll always adapt. But AI tokens cost money.
I pay $200 a month for my cloud code. And then in that allotted usage, I can only edit so many short form videos or long form videos. And on top of that, it takes a couple hours right now to edit those videos.
But let me give you the starting. point, right? For context, my human editors that I've hired over the last couple of years would take on average about five business days to edit a YouTube video like this one.
Seven to 10 if it's like a 15 plus minute video. So that is what we're comparing against. It would cost me about $500 and five to 10 business days to get it back.
And then for short form, it would take anywhere from one to three days to get the quality of short form Instagram content like this one back. But that's not including revisions, right? Every time I send back some feedback, that might take a couple of hours.
or a day or two. And again, it would cost me about $25 to $50 each. So I always tell people, it's like, no, AI edits are not instantaneous.
It's not like this, plug it in and it poops it out in 15 minutes and it's perfect. It takes a long time and a lot of usage. But when I very first had AI edit my long form videos, it took about 36 hours for it to go all the way through, including revisions.
And then I got it down to about 12 hours and it continues to go down and get better and better as models improve, as my rules get better and I train it how to be more efficient. It's a learning process. So again, from this, and onwards, AI models and AI video editing is only going to get cheaper and better and more efficient.
However, right now, from a usage standpoint, I'm on that 200x cloud code plan, costs me $200 a month. One short form video for me right now, on average, takes about 2 % of my weekly usage. If you're on the 100x plan, that would be 4 % of your weekly usage.
And if you're on that $20 per month plan, I think that'll come out to like 15 to 20 % of your usage. Long form, however, is a totally different animal. Again, as of right now, it takes about 25 % of my 200X usage plan and still takes about five to 12 hours to get a video done at this quality.
That is the expectation to expect right now. However, again, it's going to improve. But again, understand like what are we comparing it to?
I was spending 500 bucks and waiting 10 days to get back the same long form video. From an ROI standpoint, it is absolutely worth it. And my goal is, again, to get this down as much as humanly possible, both from a usage standpoint and a time standpoint.
And the cool part about my AI video editor that I'm releasing very soon here which you can join the waiting list for below, is that you're grandfathered in. All future updates I push making it better and more efficient and more creative, you get access to.
But that is just the usage side of things. We haven't even talked about the hardware side of things on your computer. This to me is an investment thing.
It's like if you care about putting out tremendous amounts of content for your personal brand, for your coaching business, for your course, for your digital products, whatever it might be, you have to invest in a good computer and something like Cloud Code so that once you get your hands on a good AI video editor, the two of those things create a beautiful combination.
AI video edits are just like normal human edits, okay? Just like if you were to edit on Premiere Pro or DaVinci Resolve or Final Cut Pro or CapCut, your computer still has to render videos. Rendering.
takes processing power on your computer. It still has to build your graphics, right? All the stuff you've been seeing on the screen and throughout this video needs a good memory to make those.
Got to have a GPU. And yes, all this footage, all these files take up space. So you got to have general disk space on your computer to handle that.
So again, that's a step a lot of people don't really understand is that the hardware is actually just as important as the AI software, which is cloud in this case. And I found that the newer Apple chips right now are built for exactly this kind of video editing work. PCs can still do it.
They're not as good. And yeah, even older computers can still technically edit videos.
It just does everything the slow way. And an edit that takes me 10 minutes might take yours an hour or two. So when you see these benchmarks I'm giving you, just understand that's from my machine.
I'm on a M1 studio right now. But if you want to get your video edit content done for you in the most cost efficient way possible. I would highly recommend you invest in a good computer with good hardware.
The AI revolution is here, folks. You got to have the hardware that can actually use it well. The next thing AI can't do very well is it cannot edit every type of video.
My AI video editor, for example, I have built for one type of very specialized kind of content and it's called talking head content. Talking head content is very simply just a person speaking to a camera like I'm doing right now. It's one person.
Generally speaking, it's educational in nature. There is a script oftentimes that you're reading. You don't have to have a script.
I'm kind of going ad lib right now, but AI is able to take one speaker's content of whatever you're talking about or teaching about. It can transcribe the whole thing, make sense of what you're trying to do, and then storyboard all your content and use it for either short form or long form videos. There are just so many types of video edits out there.
There's vlogs, you got wedding videos, real estate videos, music videos, podcast videos. I do believe AI is going to be able to edit all of that content moving forward. But right now it's not this black box where you can just simply give it some footage and then it magically.
just gives you whatever you want. But again, you can train it, I think, to do pretty much anything. If you're willing to put in the time, the investment, the hundreds of hours it probably will take you to do that.
I think having a foundation like this AI video editor I'm releasing pretty soon will give you a starting point to then be able to branch off and create all sorts of types of videos down the road. But finally, the fifth and final thing to be aware of when it comes to AI video edits is that the first cut is never the final cut.
This again, I think is just a overblown expectation people have, partly because there's a lot of content out there where people say, this edit was done in one shot with cloud code, zero anything. It just was perfect.
It's just over sensationalized. That does not exist. It never existed with human editors who, again, I think are right now even more creative than AI is.
Revisions are a creative process. Again, we have expectations. We've got tastes.
We've got preferences, things we like and don't like. I'm always giving tweaks. to any video edit and sometimes i just ask it to redo the entire storyboarding process altogether as an example last week i posted this youtube video here came out really well many of you guys were blown away at what cloud code was able to produce at the quality that it did But this was nowhere near the first revision.
I took this one through eight revisions. Now, again, some of you think, what the heck? That's crazy.
Again, I'm a particular person. I happen to have a very high preference or particularity scale of how I like my edits to go. You may not have any preference at all.
Everybody is just different. And so where I might take a edit through eight revisions, you might take it through two. But for this video, right, I would leave it feedback.
Like, for example, I don't like how you are composing this part of the video. So I gave a screenshot of what I felt like it should look like and said, yours is too boxy.
I don't like this black thing around it. It's too thin as well. Please redo it and make it look like this.
And this to me is like the real beauty of AI, right? I can just screenshot something. I could give it a YouTube video.
I could give an Instagram video. I can give it a Dropbox link. I can give it whatever I want and say, make it look like this.
And when you plug that into your cloud code subscription, again, a language learning model, an AI tool just goes through, it can analyze that visual and say, oh, okay, I see what you want. And then it came back, revised. it and it came back looking like this instead of this.
But making that tweak, right, was very simple. I just had to explain what I wanted. And it's a revision process.
The cool part to me is that it's a one size fits all tool. If you've got a very creative brain and you feel like you want to coach AI every step of the way to get exactly how you want it, knock yourself out. Your edits are going to take a long time, whether it's done by AI or a human.
That's just the nature of having a lot of preferences. But if you don't. You're a very easy to please person.
And so you'll probably like the first version or two that comes back to you and you won't have very much feedback. But with that, over the last couple of months, editing very deeply with AI. Those are the five things I have learned.
AI just isn't quite very good at, but it's not that big of a deal once you know how to get around it. If you set the right expectations, you give it really good input. You give it the assets, the footage, the stuff that you actually want it to go from.
You prepare yourself to give revisions. It really is amazing to just like be working at my desktop. I've got cloud code running in the background with my AI video editor.
I don't really... care if it takes four hours or five hours. Once that's done, I leave my feedback.
I can wake up in the morning, have it done the next day, and I post it. And this is precisely how I'm able to pump out so much more content. I'm moving at a 10x faster rate.
And perhaps more importantly, I'm spending 10 to 20 times less money to get my videos done. It's been an absolute game changer for me. And if you want to get your videos edited by AI and you want to join the waiting list below, or if you're watching this video by the time it's launched, you can click the link in the description below.
That's it for this video. Go wild, enjoy, and we'll see you guys in the next one.
The Hook

The bait, then the rug-pull.

The video you're watching was cut by Claude Code, and the host says that's exactly why he can tell you where AI editing breaks. His five limits come with workarounds, and the cost numbers are specific.

Frame Gallery

Visual moments.

One-click upgrade to your Google

Get more breakdowns in your search results

Add Modern Creator as a preferred source and Google shows you more of our breakdowns in Search, Top Stories, and AI Overviews. It only changes what you see, and you can undo it in your Google settings anytime.

Add to Preferred SourcesOpens your Google source preferences with us pre-loaded. Tick the box and you're done.
Watch next

More from this channel + related breakdowns.

48:50
Samin Yasar · Tutorial

5 Insane Claude Fable Use Cases (Last Chance)

Five ways to turn Claude's Fable model into a paid service — website builds, cold outreach, claymation ads, video editing, and a Jarvis-style command center — before flat-rate access to it disappears.

July 17th