Big Idea

The argument in one line.

An editor who couldn't get a developer to build his dream tool in 2018 built it himself with Claude Code in three months, choosing local analysis and a one-time purchase over cloud costs and subscription lock-in.

Who This Is For

Read if. Skip if.

READ IF YOU ARE…

You're a video editor or creator curious what a non-programmer can actually ship with Claude Code over a few months of evenings.
You're evaluating AI editing tools and want to understand the tradeoff between local, on-device analysis and sending footage to a cloud AI lab.
You edit multicam footage and are tired of manual sync workflows or paying monthly for a syncing tool.
You're weighing whether to build and sell your own tool versus staying locked into someone else's subscription software.

SKIP IF…

You're looking for a DaVinci Resolve editing tutorial rather than a story about building a tool for it.
You don't touch multicam footage or professional editing software and have no interest in editing workflow tools.

TL;DR

The full version, fast.

A video editor of fifteen years had wanted a tool that fixed his footage ingest and sync workflow since 2018, but a prior attempt with a hired developer went nowhere. After a few weeks of small experiments with Claude Code, he realized he could build it himself, and spent three months of evenings turning it into Cut Valet: automatic multicam backup and time-of-day sync into DaVinci Resolve, direct API control for speed ramps and graphics, and an AI assistant editor. The key decision was running 80% of footage analysis locally on his own GPU instead of uploading to a cloud AI lab, keeping metadata in plain markdown files beside the footage and selling the app as a one-time purchase instead of a subscription.

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 – 00:19

01 · The eight-year idea

A fifteen-year video editor explains he's wanted a tool like this since 2018, and a prior attempt to build it with a hired developer went nowhere.

00:19 – 00:57

02 · Discovering Claude Code

He starts experimenting with Claude Code on small editing tasks, then realizes it's capable enough to build the tool he'd been imagining for eight years, and spends three months of evenings on it.

00:57 – 01:40

03 · Meet Cut Valet

Introduces Cut Valet: it backs up footage automatically from a memory card, drag-and-drop, or a direct copy into the project folder.

01:40 – 02:25

04 · Time-of-day multicam sync

Cut Valet lines up clips from multiple cameras on a single timeline by time of day, syncing every other camera to a reference camera's audio waveform, similar to PluralEyes but automatic on ingest directly into DaVinci Resolve.

02:25 – 03:53

05 · Three core features

After a feature-creep phase he calls a dopamine-filled vibe-coding rush, he refactors the app into three pillars: file organization/backup/sync, tools and scripts that drive DaVinci Resolve's API, and smart footage analysis for an AI assistant editor.

03:53 – 04:53

06 · Scripting Resolve directly

Demonstrates controlling DaVinci Resolve's API in the background: one-click speed ramp presets applied as Fusion nodes, Remotion-generated graphics from a prompt, and pulling tracks straight from the Epidemic Sound library onto the timeline.

04:53 – 05:31

07 · Why analysis stays local

Rather than uploading footage to a cloud AI lab, Cut Valet runs about 80% of footage analysis locally on the editor's own GPU through eight free, open-source engines, handing off to a frontier LLM only for the hardest reasoning.

05:31 – 06:09

08 · Search and import

Every local engine's findings get written to a markdown file beside the footage, powering a fast text and semantic search across transcripts and metadata that imports results straight into Resolve's media pool.

06:09 – 07:00

09 · AI assistant editor

The second AI tool: pick footage or timelines to share, Cut Valet packages the markdown metadata with editing guidelines, sends it to an LLM, and executes the returned plan in the Resolve timeline, from one-shotting an edit to pulling selects.

07:00 – 07:39

10 · Buy once, own it

Closes with the pitch: Cut Valet is in beta at its lowest-ever price, sold as a one-time purchase with no subscription, because owning both the tool and the data it generates is the whole point.

Atomic Insights

Lines worth screenshotting.

An editor who couldn't get a developer to build his idea in 2018 shipped it himself in three months once he had Claude Code.
Cut Valet syncs multicam footage by treating the first camera as the time-of-day source of truth and aligning every other camera to it by audio waveform, the way PluralEyes used to, but automated on ingest.
The app's footage timeline is indexed by time of day, so finding every 3PM shot across a shoot day is as simple as scrubbing to the 1,500 mark on the timeline.
DaVinci Resolve exposes part of its toolset through an API, which lets an external app control it in the background without it being the active window.
Speed ramping through the app applies a saved preset as a Fusion node in one click, replacing DaVinci's keyframe editor for that task.
Instead of uploading footage to a cloud AI lab for analysis, the app runs eight free, open-source engines locally and only hands off to a frontier LLM for the hardest reasoning.
Running analysis locally let the builder use hardware he'd already paid for, an expensive GPU and a MacBook Pro, instead of paying a data center to do it.
Every local analysis engine writes its findings to a markdown file saved next to the footage, so the editor keeps and can reuse that metadata forever, independent of the app.
Feeding an LLM pre-digested markdown metadata instead of raw footage cuts context and API costs enough to make an AI assistant editor financially sustainable to run.
The AI assistant editor can one-shot a full edit, pull selects, or cut a long interview down to its best minutes, and it tells the editor upfront whether a request is possible before attempting it.
The product is sold as a single one-time purchase with no subscription, explicitly framed as owning your tools and your data rather than renting them.

Takeaway

Local-first analysis beat cloud AI on cost and ownership

WHAT TO LEARN

Running most footage analysis on hardware you already own, and saving results as plain files instead of locking them in a subscription, is what makes an AI editing tool both cheaper to run and safer to depend on long-term.

01The eight-year idea

A tool idea can sit unbuilt for years not because it's a bad idea, but because it depended on finding and funding the right developer.

02Discovering Claude Code

Small, low-stakes experiments with an AI coding tool are how you find out it can handle a much bigger, previously-out-of-reach project.

03Meet Cut Valet

Automatic footage backup that works from a memory card, drag-and-drop, or a plain folder copy removes friction at the exact moment editors are most likely to skip a manual backup step.

04Time-of-day multicam sync

Indexing footage by time of day, not just by camera or clip name, turns 'find the 3PM shots' into a scrub-to-timestamp action instead of a search.

05Three core features

A sprawling feature set built during an excited late-night build phase is a sign to refactor down to a small number of clear pillars before shipping.

06Scripting Resolve directly

Editing software that exposes an API turns repetitive manual tasks, like keyframing a speed ramp, into a one-click preset.

07Why analysis stays local

Running roughly 80% of AI footage analysis locally on owned hardware avoids per-video cloud processing costs and keeps raw footage off a third party's servers.
Reserving the cloud LLM for only the hardest reasoning steps, after local processing has done the bulk of the work, is what keeps the AI feature affordable to run.

08Search and import

Saving AI analysis results as plain markdown files next to the source footage keeps that metadata usable and portable even if the app that generated it disappears.

09AI assistant editor

Feeding an LLM condensed, pre-digested metadata instead of raw footage or transcripts cuts the token cost enough to make an AI assistant feature financially sustainable to offer.
Letting a tool state upfront whether a requested task is even possible, before attempting it, sets a more trustworthy expectation than silently failing or guessing.

10Buy once, own it

Positioning a tool as a one-time purchase instead of a subscription is itself a selling point to a professional audience wary of losing access if they stop paying.

Glossary

Terms worth knowing.

Cut Valet: The DaVinci Resolve companion app built in this video: it backs up and syncs multicam footage, scripts Resolve's API, and runs local AI footage analysis.
DaVinci Resolve API: A set of functions DaVinci Resolve exposes so outside programs can control parts of the editor, like adding timeline items or triggering renders, without the editor being the focused window.
Multicam sync by time of day: Lining up footage from several cameras on one timeline based on the real-world clock time each clip was shot, rather than manually matching clips by hand.
PluralEyes: A well-known third-party tool editors have historically used to automatically sync multicam footage by matching audio waveforms across cameras.
Fusion node: A building block inside DaVinci Resolve's Fusion compositing page; effects like a speed ramp preset can be applied to a clip as a saved Fusion node.
Local footage analysis: Running AI processing on footage using the editor's own computer and graphics card, instead of uploading it to a remote AI company's servers.
Frontier lab: One of the companies building the most advanced, largest AI models, typically accessed through paid cloud APIs.
Markdown metadata file: A plain text file saved alongside a video clip that stores transcript, description, and other analysis data in a simple, human-readable, portable format.
AI assistant editor: A feature where the app hands selected footage and its metadata to an LLM, which returns and executes an edit plan, like pulling selects or building a rough cut, directly in the DaVinci Resolve timeline.
Epidemic MCP: An integration that lets the app search the Epidemic Sound music library directly and drop a chosen track onto the timeline without leaving the app.

Resources

Things they pointed at.

00:22toolClaude Code

00:57productCut Valet

02:20toolPluralEyes

05:11toolRemotion

05:24toolEpidemic Sound (via Epidemic MCP)

00:00toolDaVinci Resolve

Quotables

Lines you could clip.

00:04

“I've been a videographer and more importantly, a video editor for well over fifteen years now.”

credibility-establishing cold open→ TikTok hook↗ Tweet quote

00:58

“Park your footage, pick up the cut.”

tight product tagline→ IG reel cold open↗ Tweet quote

07:13

“I have an expensive graphics card. I have an expensive MacBook Pro. Why can't I use my own graphics card on my own compute?”

sharp rhetorical case for local-first AI tooling→ TikTok hook↗ Tweet quote

08:51

“With this program, when you buy it, you own it. That's it.”

clean anti-subscription punchline→ 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.

I've been a videographer and more importantly, a video editor for well over fifteen years now. For the past eight years or so, I've had this tool in mind that would just supercharge my workflow and make everything so much easier. I actually met a developer in 2018 and tried to get get this project up off the ground, but it never went anywhere.

Enter Claude Coat. This software is actually wild. It lets you control and utilize computers in a completely new way.

And I started off with this the same way I do with most things, experimenting. Claude, can you organize my footage for me? Claude, can you do this edit for me?

After a few weeks of applying small band aids to my video production workflow, I realized that with this tool, I could build exactly what I've been dreaming of for the past eight years. So I got to work. Over the past three months in the evenings and in between shoots, I've been picking away at developing this program.

And today, I'm finally ready to share it with you guys. This is the Cut Valet. Park your footage, pick up the cut.

At its core, it does one main thing, and then everything else kinda builds on top of that. It backs up your footage. You can plug in a memory card and it will automatically see it or drag and drop anywhere onto the app or just simply copy footage into the project files folder directly, and it will copy and back up that footage for you.

Now I know there's tools out there already that exist that do this, but I wanted to take it a step further. I wanted to take clips from multiple different cameras and throw them all right onto a timeline based on the time of day that it was shot at. Then the first camera becomes the truth for time of day, and all the other cameras sync to it based on the audio wave forms.

Kinda like what PluralEyes used to do, but automated and automatic on ingest and sync directly into DaVinci Resolve. So you get a time of date timeline for every day that footage was shot.

So if you wanna find 3PM, you just go to 1,500 on the timeline, and there's all your 3PM shots.

At this point, I started to fall into a dopamine filled feature building rush of late night vibe coding and bug fixes, and at a certain point, I got a bit too excited about what I could build and sorta lost the plot. Fast forward to the present, and I've refactored it into three main features. And it works on Mac and Windows, by the way.

Check it out. File organization, backup, and DaVinci Resolve sync, still here, still works great.

Tools and scripts, these control DaVinci Resolve in the background. So powerful, so handy, so awesome.

And smart footage analysis, powering an AI assistant editor, but we'll get more into that shortly. DaVinci Resolve is nice enough to expose a few of their tools to us through an API.

What that means is we can control DaVinci Resolve from our program without having to have it as the active window or clicking anything. Super handy.

It took a long time to crack this one, but we can speed ramp. Instead of using that finicky keyframe editor, you can just add segments and save those presets and then apply them as a fusion node to add speed ramps in one click.

Super helpful if you do a lot of speed ramping. We can create a graphic using Remotion. Now this is something you've probably seen online watching those Claude edited my videos for me clickbait videos, but it is a super powerful tool to create graphics from prompts, and it deserves its own video about it in its own time.

We can find a song using the epidemic MCP, searches the library directly, find what we're looking for, and it automatically saves it to our project file, lands on the timeline, add our playhead on a separate music track. We can build a rough cut, color grade, make a new timeline version, find sound effects, create thumbnails from markers straight out of Resolve, automate uploading to YouTube, and even build and manage our own tools to add in any functionality that you want, which brings us to the AI assistant editor, which is where some of the crazy stuff happens.

Now my goal isn't to replace me as the editor, it's to give me better tools to help do my job. Pull selects, index my archive, cut down that twenty minute interview to the best three minutes, that kind of thing.

But in order for any AI to meaningfully help you do this, it needs to understand your footage, footage, when things are happening and what's happening and what's said and all this advanced metadata. So a lot of the other tools require you to upload your footage, send it to a big AI lab, and have them analyze it all through their data center, which costs a lot of money.

But I'm a video editor. I have an expensive graphics card. I have an expensive MacBook Pro.

Why can't I use my own graphics card on my own compute? Why do I have to send it off to a data center? What I found was that I could do 80% of the local analysis on my own machine, and then when I need it to do the advanced really difficult stuff, we can just hand over that information to a Frontier lab without giving it all this footage and making it do all this extra work, and it can still output a really similar result.

So that's where we left off. I built in eight different metadata engines, let's call them.

And these are all free to run locally on your machine, open source, good to go.

No cost whatsoever. And they all save all of this information to a markdown file that's saved right beside your footage, so you own your data. You can keep it.

You can use it however you want forever. To me, that's the most important part while building this. Not getting locked into some proprietary subscription once a month, and if you leave, you lose all of that progress you've made.

So with this program, when you buy it, you own it. That's it. With all of this information, we can power two main tools that are really awesome.

The first, pretty simple, search and import. You can search any of that transcription, description, metadata.

It's all searchable. It's all quick, and it automatically imports right into Resolve once you found what you're looking for. I keep building this out and keep adding new things because I find this to be one of the most useful things I've ever built.

You wanna create a match cut shot of your dog growing up. Boom. Search dog.

It's It's all there. So handy. And the second one is using AI as an assistant editor.

And this is definitely still in beta, but it's getting better every time I use it, which is very cool. But basically, you choose what footage or timelines you wanna share with it, and then CutValet packages all of that information from those markdown files that we talked about earlier with guidelines and rules, and it sends it over to an LLM to help make decisions.

And it returns with a plan, executes it in your DaVinci Resolve timeline. And this can do anything from one shotting a full edit to just pulling selects or cutting down an interview.

Whatever you ask it to do, it'll tell you if you can do it, and it'll try. And since we're providing the LLM with all this information in a really digestible format, we're saving a ton of money on context and API costs, making this actually a sustainable tool that you can use.

If you made it this far in the video, thanks for watching. You're obviously interested in this kind of thing, so check out CutValley in the description below and, uh, think about giving it a try.

It's in the cheapest it will ever be right now because it's in beta, but it's pretty awesome.

I'm really excited about it. I'm gonna keep refining it. I'm kind of obsessed, so I'm gonna keep building it constantly because I use it every day already on all these projects.

And it's it's not gonna be subscription. It's a buy once, you own it. You own your data.

I'm all about that kind of thing. So no subscriptions here, and yeah.

Thanks for watching. I'll see you guys on the next one.

The Hook

The bait, then the rug-pull.

He'd wanted this tool since 2018 and even hired a developer to build it once. It went nowhere. Eight years later, Claude Code let him build it himself, in the evenings, around his actual editing work.

Frameworks

Named ideas worth stealing.

03:35list

Cut Valet's three pillars

File organization, backup, and DaVinci Resolve sync
Tools and scripts that control Resolve in the background
Smart footage analysis powering an AI assistant editor

After an early feature-creep phase, the builder refactored the sprawling tool down into three organizing categories.

Steal forstructuring any multi-feature internal tool pitch or product page

CTA Breakdown

How they asked for the click.

VERBAL ASK

07:15product

“check out Cut Valet in the description below... it's in the cheapest it will ever be right now because it's in beta”

Soft, credibility-first pitch after the full feature walkthrough: frames the current beta price as the floor, then closes on the one-time-purchase, own-your-data positioning rather than urgency or discounting.

FROM THE DESCRIPTION

PRIMARY CTAWhere the creator wants you to go next.