Big Idea

The argument in one line.

Claude Opus 5.5, driven through Claude Code and connected directly to a video editor via MCP, can now execute an entire edit pass from one dictated brief, cutting silences, removing bad takes, adjusting layouts and zooms, and generating original motion graphics, getting a creator roughly 80% of the way to a finished video before any manual polish.

Who This Is For

Read if. Skip if.

READ IF YOU ARE…

You publish talking-head YouTube videos regularly and want AI to take over the mechanical parts of editing, like silence removal, bad-take cuts, and zooms, so you spend less time in the timeline.
You already use Claude Code for other work and are curious whether it can be pointed at creative tools through MCP instead of just writing code.
You're weighing a per-scene recording-and-editing tool like Borumi against a traditional NLE, especially if you want AI to eventually take over editing.

SKIP IF…

You edit multi-camera, narrative, or heavily produced video — this workflow is built around simple talking-head-plus-screen-share tutorials, not complex cuts.
You don't want to pay for both Claude's higher usage tiers and a separate paid editing tool at the same time.

TL;DR

The full version, fast.

Paul J Lipsky now edits every YouTube video with Claude Opus 5.5 running inside Claude Code, connected to a screen-recording editor called Borumi through its MCP integration. He records each scene separately in Borumi, then tells Claude, via a saved reusable skill, to cut silences, remove bad takes, clean up the frame, and add zoom effects on anything he points at or clicks. Beyond those mechanical edits, Opus 5.5 can now build original motion graphics and even capture its own screen recordings from a single voice-dictated, scene-by-scene brief, work he used to split between separate models and manual dragging. The result gets him roughly 80% of the way to a finished video in one pass, with a short manual polish step still required, and he calls it the first model good enough to replace his old hybrid manual-and-AI workflow.

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 – 00:23

01 · Intro

Cold open claim: this video and recent uploads were edited by Opus 5.5, which he now calls the best editing model he's used.

00:23 – 01:03

02 · Claude Setup

Explains he edits via Claude Code, not just chat, with Opus 5.5's thinking level set to high, and that a $100/month plan covers a full week of heavy editing plus other work.

01:03 – 02:18

03 · Meet Borumi

Introduces Borumi, the record-and-edit app he uses, discloses the affiliate link, and creates a new structured-video project with a script tab for planning scenes.

02:18 – 03:12

04 · Record in Scenes

Lays out six scenes (Intro, ChatGPT, Claude, Gemini, Apple, Outro), then shows the record tab: camera, microphone, screen, and system-audio capture per scene.

03:12 – 05:49

05 · Manual Editing

Tours the editor's timeline and auto-transcript: deleting silence markers, cutting false starts, switching camera/screen layouts, and adding zoom or highlight effects by hand, before any AI is involved.

05:49 – 07:29

06 · Connect Claude

Connects Claude to the named Borumi project via its MCP integration, states the rule that AI executes while the human sets creative direction, and packages the repeated basic-edit instructions into a reusable skill.

07:29 – 09:23

07 · Motion Graphics

Uses voice dictation to walk scene by scene through exactly what motion graphics he wants, including a title card and a screen recording he never captured himself, then sends it as one prompt.

09:23 – 11:35

08 · Review the Edit

Returns to Borumi to inspect the result: silences and bad takes gone, layout changes and zooms applied on cue, and original motion graphics including an intro title card and a usage-percentage chart.

11:35 – 12:38

09 · Final Polish

Calls the result about 80% of the way to finished, still needing a short manual pass, and closes with the Borumi affiliate link, noting it's a one-time purchase with frequent free updates.

Atomic Insights

Lines worth screenshotting.

Claude Opus 5.5, run through Claude Code with the thinking level set to high, is the first model good enough that one creator lets it edit videos in a single pass instead of a hybrid manual/AI workflow.
A $100/month Claude plan covers a full week of both heavy video editing and other work with no usage issues, according to the creator's own account.
The editor Borumi records and edits inside the same app, broken into separate per-scene takes, so a bad take only requires re-recording that one scene, not the whole video.
Borumi's MCP integration lets Claude see and directly manipulate a specific named project inside the editor, rather than just generating instructions a human has to apply by hand.
The creator turned his repeated basic-edit instructions (cut silences, remove bad takes, clean the frame, add zooms on anything he points at) into a saved reusable skill so he doesn't retype them every video.
AI still needs a human to set creative direction: the creator's rule is that Claude does the labor, but he decides what layouts, highlights, and zooms he actually wants first.
With one voice-dictated, scene-by-scene brief, Opus 5.5 generated original motion graphics matched beat-by-beat to the narration, including title cards and a data visualization.
Opus 5.5 can perform its own screen recordings, such as scrolling a product page, when the creator forgot to capture that footage himself, a capability he says he only recently discovered.
Before Opus 5.5, the creator split video-editing work across different models, using one for generating motion graphics and Claude separately for placing them; Opus 5.5 now does both steps itself from a single prompt.
The finished AI edit gets the creator roughly 80% of the way to done, with a manual pass still needed for small mismatches like an unwanted zoom or a layout he wants adjusted.
Borumi is sold as a one-time purchase rather than a recurring subscription, and the developer has continued shipping near-weekly updates since launch.

Takeaway

AI editing now runs end to end, not step by step.

WHAT TO LEARN

A coding-agent model connected to a screen-recording editor through MCP can now execute an entire edit pass, silences to motion graphics, from one dictated brief, as long as a human still sets the creative direction first.

02Claude Setup

Editing inside a coding-agent interface instead of a plain chat app gives the AI more direct control to actually execute edits, not just describe them.
Setting a model's thinking level to 'high' before an editing session trades a bit of speed for more careful editing decisions.
A single mid-tier subscription plan can cover a full week of both heavy AI-editing work and unrelated tasks without hitting usage limits.

03Meet Borumi

Recording and editing inside the same app removes the export/import friction of moving footage between a separate camera app and a traditional editor.

04Record in Scenes

Breaking a video into a per-scene script means a single flubbed scene can be re-recorded in isolation instead of restarting the whole take.
Recording camera, microphone, screen, and system audio as separate toggles lets a single scene capture a live browser demo with its own sound baked in.

05Manual Editing

Understanding the manual editing steps, like deleting silence gaps, cutting false starts, choosing layouts, and adding zooms, is what makes it possible to write a precise brief for an AI to replicate them.
An auto-generated transcript with clickable silence markers turns 'remove the dead air' from a scrubbing task into a few clicks.
Zoom and highlight effects applied as discrete, addressable timeline actions are exactly the kind of structured task an AI agent can be instructed to repeat.

06Connect Claude

An MCP connection lets an AI model act inside a specific named project in a third-party app, not just chat about it; the model has to be told which project to operate on.
AI is not yet reliable at deciding what to edit on its own; a human still has to define the creative direction, which cuts, which zooms, which layouts, up front.
Turning a repeated editing brief into a saved, reusable skill avoids re-typing the same instructions for every new video.

07Motion Graphics

A single scene-by-scene voice-dictated brief can be enough for a model to generate original motion graphics matched beat-by-beat to the narration, not just apply pre-made templates.
Being specific about structure, like frame size, what text goes where, and what footage to show, in a dictated brief is what lets a model execute without back-and-forth.

08Review the Edit

Reviewing the actual timeline in the editor, rather than reading the AI's own summary of what it did, is how you actually verify the quality of an AI edit.
An AI agent that notices a missing asset on its own, like recording a live browser session when a screen capture was never taken, is a meaningfully higher bar than one that just leaves a gap.

09Final Polish

Even a strong single-pass AI edit lands around 80% finished; a short manual pass to swap out an unwanted zoom or layout choice is expected, not a sign of failure.
A one-time-purchase pricing model for a niche creator tool, with frequent free updates, is a meaningful trust signal compared to ongoing subscription software.

Glossary

Terms worth knowing.

Borumi: A screen-recording and editing app that lets creators record and edit talking-head video in the same interface, broken down scene by scene.
MCP (Model Context Protocol): A connection standard that lets an AI model see into and directly control a third-party app, such as one specific editing project, rather than just producing instructions a human applies by hand.
Thinking level: A setting that controls how much reasoning effort an AI model applies before responding; set to 'high' here to get more careful editing decisions.
Skill (Claude): A saved, reusable set of instructions a user can trigger by name instead of retyping the same prompt every time.
Zoom effect: A punch-in on a section of a video timeline that draws the viewer's eye to whatever is being discussed or clicked on screen.

Resources

Things they pointed at.

01:03toolBorumi ↗

00:30toolClaude Code

00:01toolClaude Opus 5.5

Quotables

Lines you could clip.

00:01

“This video and the YouTube video you're watching right now were both edited by Opus 5.5, and it is now the best model I have ever used for editing videos.”

headline claim, no setup needed, states the whole premise in one line→ TikTok hook↗ Tweet quote

11:44

“This gets me like 80% of the way there... compared to any other model that I've used, this gets me much, much closer to the final result.”

honest, specific payoff number that frames the whole demo's conclusion→ IG reel cold open↗ Tweet quote

11:58

“It's not free though. It's a one-time payment though. So you're not gonna be paying an ongoing subscription.”

clear pricing distinction that lands the CTA without sounding like a pitch→ 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.

This video and the YouTube video you're watching right now were both edited by Opus 5 .5, and it is now the best model I have ever used for editing videos. It cuts out the silences, removes bad takes, add effects, and even creates motion graphics all by itself.

Today I'm going to show you my whole setup and how I edit all my YouTube videos now using Opus 5 .5. So first, I edit all my YouTube videos using the Clod desktop app, and even though I primarily use Clod what I do inside of Clawed, for editing videos I use Clawed code because it gives Clawed a lot more control to be able to help me better with my edits.

If you've never used Clawed code before but you have used Clawed co -work, I promise you what you're going to get with Clawed code is going to feel pretty much exactly the same. I also of course use Opus 5 .5 and I set the thinking level to high. And despite doing a lot of editing every week and a lot of other work inside of Claude, my $100 a month plan lasts me all week with no issues.

Now, there are many different tools that Claude can use to edit videos for you. The one that I use is called Barumi. I really like the software because it allows me to record and edit directly into the same software.

And it's really good for talking head style tutorials like the ones that I make on YouTube. So let me give you a quick tour of how Barumi works. They're not sponsoring this video, but if you want to try it out, there is a link for it down below and that will be my affiliate link.

But I've been recommending them long before they even had an affiliate. program so if you want to use my link or not it really doesn't matter to me so this is where you see all of your projects that you've been working on I'll create a new one by clicking right here where it says structured video And then down here at the bottom, you'll see three tabs, script, record, and edit.

So the script tab is where you can put in the script for your video. This is completely optional. You don't need to do this, but it's also where you can set up your scenes.

So I might have my intro here and then I might have another scene. So let's pretend this is a video where I'm doing a recap for the week of all the AI news from the week. So I have my intro and then I might first talk about what has been happening with ChatGPT and then Claude and then Gemini.

and then some Apple stuff, and then we'll have the outro. So we have six scenes in this video. The next tab down here is the record tab.

And this, of course, is where you can record directly into Burumi. Now you notice here that each of the recordings is broken down by scene, which means that you don't have to record the entire video at once. You just record each scene.

And if you mess up one of them, you can just start over that single scene. Over here on the left, you can choose what you record. So we have my camera.

my microphone right here and my screen you can choose to record your entire screen or just a single app and you can also record the system sound so that's helpful for instance if you want to record something on your browser like a video playing you can record that and it'll also record the audio from it and then when you're ready to record you just click right here where it says record now i'm just going to pause the video right here i'm going to record all of these scenes so i can then show you how we edit them using the editor which is the last tab down here all right i'm now finished recording the video and you can see that all of the scenes now have a video in them i also changed my shirt for the video so it's easier for you to tell this paul from that paul just as you're watching this.

But now that I'm done recording, I can click down here on the bottom where it says edit. and this is the editor and you'll notice that right here we have a timeline this looks somewhat similar to a lot of other video editors out there and on the top right we have a transcript that was automatically generated so before i started using ai to edit my videos this is how i would start editing my videos because you could highlight any of these dots right here these dots meant or mean that there is a silence in the video that's little bit too long so I could just click delete here and it deletes that silence from the timeline down here at the bottom I also would read through this transcript and remove all the false starts that's when I would start talking kind of like mess up stop and then go back and say what I tried to say again so here you can see in the intro I say hey what's up everyone welcome back then I pause and then I start again so I'm gonna get rid of all of that by highlighting the silence delete that

and take this one sentence right here and delete that and all that is deleted from the timeline so i'd go through the entire thing and do that the next thing i would do is change some of the layouts so by default if i record both my screen and my face it would always look like this with me in the bottom corner but i could change that layout to something like just a screen to just my face or i can turn myself into a circle if i wanted to or move the position of it so there's a lot of different choices in here or you just leave on an auto and it automatically puts you in the bottom right hand corner I would often add zoom effects as well.

So right here in the video, I'm talking about the new ChatGPT model. So I would add this zoom effect. Click on add.

And now when you play it, you see it zooms in where the cursor is. I could also add a highlight effect. So that's this button right here.

So I would select this. Let's say I just want to highlight this section of it like that. And then as I play it, you see it just highlights.

that one section and darkens everything else. Those are the basics. There's a bunch more you can do with this, but I'm not making this video as a full Barumi tutorial.

I just want to kind of show you the basics of how to edit things and what you can do inside of Barumi. So you all understand what Opus 5 .5 is doing when it's editing the video for me. Okay, back in Claude, I've already set up the Barumi MCP, which allows Claude to see what's happening in any of my Burumi projects and to make edits inside the video.

And you can set that up by coming to your Burumi settings, clicking on the left here where it says AI, and there are setup instructions for ChatGPT, Claude, and other AI apps that are pretty easy to follow. So once you have that set up, you can come over to Claude and tell it the name of the project you're working on. So the name in Burumi is called AI News Demo, so that's the name of the project that I gave it.

Now, you can't just tell Obis to edit the video. And that's it. Because AI is not very good creatively.

You as a human need to drive the creative direction. The AI is just doing the actual work for you. So what I always start by doing is just describing the basic edits that I want Claude to make for me.

Those are cutting out the silences, removing the bad takes, and just cleaning up the outside of the video. If there's like a menu bar exposed, I want that cut out. And just adding the zoom effects anywhere I'm clicking or anything that I'm pointing out in the video, I want it to zoom in on that area.

And because those are instructions that I repeat every single time I have edit videos for me, I just created a skill for that. This is it right here. It's just called edit Burumi video, and it just does everything I just told you.

It's really nothing special there. Then what I would sometimes do is give it more specific instructions as to different layouts I want or where I wanted to create those highlights or zooms in. But usually I just let it do its first pass and then tell it to make any changes that I want after that first pass.

And particularly, I have found that Opus 5 .5 is very good at these basic edits. The next thing that I do is I ask it to create for me motion graphics. where i need them in the video this is something that used to be a lot more work because i used to have gpt 6 astra or 5 .6 edit my videos and i would have claude make the motion graphics and then i would have to drag them in myself but opus 5 .5 can do all of it with just a single prompt so i will use a voice dictation tool and go through the timeline and describe what motion graphics i want and any specific changes that i want as well Okay, so in scene one, I'm just a talking head.

So I need you to create a motion graphic for this that matches beat by beat what I'm talking about. So for instance, you can have something put me on the right hand side. It's like a nine by 16 frame.

And on the left, I want some words that say like weekly AI news recap. Underneath that says all the most important news. And then I want you to show a screen recording or screenshot of my YouTube channel where people can see I've posted a lot of videos there.

So that's everything for scene one. In scene two, there's a section where I talk about how much usage the different models have burned through. So I want a motion graphic for that representing all of that sort of data, so to speak.

And then finally, in scene five, I'm talking about the new iPhone, but I didn't do a screen recording for that, so can you screen record the Apple website, the photos of the new iPhone Duo, and just kind of click through each of the photos as I'm talking, just to have something visual on the screen while I'm talking. Okay, so I think that's everything.

There we go. And so now all of that is typed out. Let's just send it along and then we'll give it a few minutes and see what Opus 5 .5 turns this video into.

All right, it is now finished. We have everything that it did listed out for us right here. I usually don't read this, to be honest.

I just go over to Burumi and actually take a look at what it did. So let's hop over there and look at the edit. Okay, so obviously the timeline now looks a lot different.

All of the silences have been removed. Up here in scene one, which is where most of those bad takes were, I can see that all those have been deleted and we just have the final nice take. And we have a lot of layout changes.

So over here, for instance, it just shows my face. Here it shows me in the corner, and they added a bunch of zooms, so let's take a look at some of these zooms. New model, you can see that right, see it zooms in when I'm talking about the model picker there, and even uses the highlight function as well.

All that looks great. Here it zooms in again when I'm showing, when I'm typing into Gemini. So let's look at some of the other stuff that it did, like the motion graphic.

So let me just actually play the intro because the intro has quite a bit of motion graphics in it that I asked for. Welcome back to the channel, everyone. My name is Paul J.

Lipski. And in this video, we're going to do a weekly AI news recap. We're going to tell you all the important AI news from the past week.

If you don't know who I am, I make videos every week about AI and different AI news. So if that interests you, make sure to subscribe. And now let's get into the important updates.

from this past week. Yeah, I think that looks really, really nice. Let's look at the other motion graphic.

This is it right here is when I'm talking about some numbers. However, it is getting me through the entire week. GPT -6 Sol would burn about 25 % of my weekly usage in just a single day, but now GPT -6 Sol burns less than 10%.

So it is much more usable and it'll go much further than... Yeah, it looks really nice. I like the way that one came out.

And then finally over here, this is a section where I was talking about the iPhone Duo and I didn't do a screen recording. So it actually did a screen recording for me, which is really cool. I didn't even know that this could do this until recently.

I just asked Claude, hey, can you... actually just do the screen recording for me and put it in and it did which is really cool to see so this gets me like 80 of the way there i still have to go through and do a final polish there's always small mistakes that it makes there's always sections where i'm like you know what i want this to look a little bit different or maybe i don't want to zoom here after all but overall this does a fantastic job and compared to any other model that i've used this gets me much much closer to the final result is incredibly impressive.

So that's it. That's how I edit my videos now using Opus 5 .5 and Burumi. If you wanna try out Burumi, there is a link for it in the description down below.

That's my affiliate link. And you also get a free trial of it if you wanna try it out. It's not free though.

It's a one -time payment though. So you're not gonna be paying an ongoing subscription. I've paid for it once.

There was never any upcharge after that. And the developer continues to update it Pretty much it feels like every single week he's adding updates these days.

So definitely check it out if you want and you like this setup. Otherwise, thanks so much for watching and I'll see you in the next video. Bye for now.

The Hook

The bait, then the rug-pull.

Paul J Lipsky opens by claiming the video you're watching was cut entirely by Claude Opus 5.5, then spends twelve minutes proving it: touring the recording tool, showing the manual editing basics AI needs to replicate, and revealing exactly what Opus 5.5 built from one dictated brief.

Frameworks

Named ideas worth stealing.

05:49concept

The two-pass AI edit

Pass one: a saved skill triggers the mechanical cleanup — cut silences, remove bad takes, clean up the frame, add zooms on anything clicked
Pass two: a scene-by-scene voice-dictated brief for original motion graphics and any specific layout or highlight requests
Manual review: swap or adjust the handful of edits that missed the mark

A repeatable structure for handing an entire edit to an AI agent connected to a creative tool via MCP: automate the mechanical pass with a saved skill, dictate the creative pass scene by scene, then do a short manual review.

Steal forany repeatable video production workflow using an AI agent connected to a creative tool through MCP

CTA Breakdown

How they asked for the click.

VERBAL ASK

11:58link

“If you wanna try out Borumi, there is a link for it in the description down below. That's my affiliate link... you also get a free trial of it if you wanna try it out. It's not free though. It's a one-time payment.”

Disclosed as an affiliate link twice (once early, once at close), softened by noting he recommended the tool before any affiliate program existed and that it's a one-time purchase, not a subscription.

MENTIONED ON CAMERA

01:03toolBorumi ↗

FROM THE DESCRIPTION

PRIMARY CTAWhere the creator wants you to go next.