The argument in one line.
Claude Fable 5 now handles the full editorial pipeline — accurate cut selection, b-roll matching by description, and final render — reducing the creator role to recording and one prompt.
Read if. Skip if.
READ IF YOU ARE…
- A solo creator who shoots more than an hour of raw footage per week and is drowning in edit time.
- Someone already using Claude or AI tools for writing who wants to extend that into post-production.
- A YouTube or short-form creator with an existing b-roll library who wants Claude to match footage to script automatically.
- Anyone paying an editor or considering hiring one who wants to evaluate whether AI can replace that spend.
SKIP IF…
- You need frame-precise color grading or motion graphics beyond basic captions and callouts.
- You shoot in a team environment where editorial decisions require client sign-off — this workflow is built for solo autonomy.
- You have no existing b-roll library and do not film lifestyle or context footage alongside talking-head content.
The full version, fast.
The Fable 5 editing system runs two tracks in parallel: a cut skill that transcribes raw footage to the millisecond, detects waveform anomalies between takes, and exports only the keeper clips; and a b-roll library that auto-ingests new footage by screenshotting per second, color-grading, describing by mood, and sorting into categories. When you prompt Claude to edit a video, it reads the cut output and the library descriptions, plans which b-roll fits each sentence, checks your on-screen position to avoid text-over-face collisions, then produces a preview render. One natural-language feedback round triggers a full render with music and sound effects. Thirty minutes of raw footage becomes 13 edited clips with one prompt.
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 →Where the time goes.
00:00 – 01:08
01 · Hook + system overview
Claims Fable 5 handles cuts, b-roll, and animations. Password teased to lock watch-time.
01:08 – 02:48
02 · The cut skill
Raw take shared with Claude. Audio extracted, transcribed to millisecond, keeper takes selected, waveforms smoothed, clip rendered.
02:48 – 04:52
03 · Output inspection + b-roll track intro
Cut output folder shown. B-roll ingest process: SD card in, screenshot per second, color grade, describe, sort by mood.
04:52 – 05:31
04 · Full edit prompt
Single prompt fires the edit skill: b-roll plan, text placement, captions, callouts, preview render.
05:31 – 06:51
05 · B-roll library deep-dive
Frame check prevents text-over-face. Stage 2 planning reads all drive file descriptions to match b-roll to script sentences.
06:51 – 07:39
06 · Preview + feedback loop
Low-quality preview played. Creator rejects a b-roll clip via screenshot + natural language. Claude generates replacement.
07:39 – 08:36
07 · Full render with music + SFX
Timestamp-matched b-roll assembly shown. Same system for shorts and long-form. 30 min raw becomes 13 clips with one prompt.
08:36 – 09:23
08 · Final output + CTA
Final edited clip played. Password Fable revealed. Free skill link in description.
Lines worth screenshotting.
- Waveform detection between takes is what makes AI cuts sound clean — without it, audio level jumps between retakes break the illusion of a single continuous take.
- Describing b-roll files by mood category lets Claude match footage to script semantics rather than filename guessing.
- A screenshot-per-second ingest pass means Claude knows exactly what happens in each b-roll clip before the edit ever starts.
- Frame-checking speaker position before placing text prevents the most visible AI editing failure mode: captions burned over the host face.
- The password CTA mechanic — tease the free skill link at 0:07, withhold until 9:20 — is a clean watch-time retention play with a tangible reward.
- One prompt on 30 minutes of raw footage produced 13 fully edited clips, meaning the marginal cost of each additional clip approaches zero once the library and skills are set up.
- A low-quality preview render as the first output is the right design choice: it lets you reject bad b-roll before committing compute to a full render.
- Sorting b-roll by mood rather than topic makes the library compounding — every new clip filed makes every future video better, not just the next one.
- The cut skill picks the last of eight takes automatically, which means recording multiple attempts costs nothing — the AI handles keeper selection.
- This workflow separates shooting (human) from every post-production decision (Claude), which means the skill ceiling for production quality is now the quality of your b-roll library, not your editing speed.
How AI video editing actually works end-to-end.
WHAT TO LEARN
The real unlock is not a single AI cut — it is a compounding library where every clip you shoot makes the next edit smarter.
- Millisecond-level transcription is the foundation: without knowing exactly when each word was spoken, AI cannot make clean cut decisions or match b-roll to specific sentences.
- Waveform detection between takes is what separates a professional-sounding AI cut from an obvious one — ignoring audio level jumps produces harsh edits even when timing is right.
- A mood-sorted b-roll library compounds over time: the more footage filed with descriptions, the better every future video match becomes at no additional effort.
- Frame-checking speaker position before placing text is a prerequisite for professional output — skipping this step is why most AI video tools produce captions burned over faces.
- A low-quality preview render as an intermediate step is the correct design: it lets you catch bad b-roll choices before committing to a full render with music and sound effects.
- Giving feedback via a screenshot and natural language prompt is faster and more precise than timeline scrubbing — the AI knows exactly which clip you mean because it has the frame context.
- The same pipeline handles both shorts and long-form without modification — the difference is input length, not workflow complexity.
- Thirty minutes of raw footage becoming 13 edited clips with one prompt is a concrete throughput benchmark: it reframes the question from whether AI can edit to how many clips per hour this unlocks.
Terms worth knowing.
- Cut skill
- A Claude skill (structured prompt + Python scripts) that extracts audio from a raw take, transcribes it to the millisecond, identifies keeper segments, smooths waveform transitions between takes, and exports a trimmed clip.
- Waveform detection
- Analysis of audio amplitude patterns between retake segments to identify jarring level jumps; used here to smooth edits so cuts between takes sound natural rather than harsh.
- B-roll library
- A mood-sorted file store of supplemental footage where each clip has been color-graded and described by Claude so it can be retrieved semantically during editing.
- Frame check
- A pipeline step where Claude analyzes each video frame to determine where the speaker is positioned on screen, preventing text or captions from overlapping the subject.
- Edit skill
- A multi-stage Claude skill that reads the cut output and b-roll descriptions, plans the assembly, generates a low-quality preview, accepts feedback, then produces a full render with music and sound effects.
- Fable 5
- Claude model variant used in this workflow; selected in the Claude interface alongside a project folder containing the skill files and Python scripts.
Things they pointed at.
Lines you could clip.
01:06
“I basically never have to edit again.”
02:48
“We pushed that into one skill that can do everything the same level as an editor and even better on a larger scale.”
08:36
“I previously did this with thirty minutes of raw footage. And with one prompt, it gave me 13 different clips fully edited.”
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.
Professional video editing is finally possible with AI, and this always was a very hard task. But right now, with Fable five, it is really changing the whole industry. It can cut the clips insanely accurate, find my own b roll, and make cinematic animations.
Animations. And as someone who needs to edit more than four hours weekly, this is really good news. So make sure to stick around until the end of this video.
I will give you a password. And with that password, you will get access to all my Claude skills for video editing. So, let's dive in.
Claude Fable five is really insane. I basically never have to edit again. And in this video, I will show you everything you need to know to achieve this as well.
My last video about this went pretty viral, but this was just the start. It wasn't cutting the videos. It wasn't using any b roll.
And the animations were quite alright, but it could be better. And that is exactly what is now possible with Fable five. So in my videos, I basically have two types of content, talking and the b roll.
And I will explain you the whole system and how it works together. So it all starts with a raw messy take. So just like now, I'm sitting in front of the camera and talking.
Then we transcribe everything to the milliseconds. So every word gets a time stamp. Then it finds the storyline.
So, in all my retakes, in all my pauses, it cuts out exactly what we need for the video. So, what we have right here is we have our clip, our raw clip. So, we can open it and I can play it for you.
You didn't start a business just to get a job. As you can see, this is a very raw clip of me making mistakes, having pauses, and this really needs an edit.
So what we do is the We can share this clip directly with Claude. So we go to Claude, we open a new chat, and we choose fable five.
What we then do is the following, is we add a folder. So I click here on add folder, I go to fable five, and I open it. I then say the following, please use the video cut up skill to trim this video in a professional clip, cut out the bad parts, etcetera.
We can then add the clip name, so that will be ads dot m p four, and we can run it. Right here, we have the video cut scale. So it has knowledge, it has some Python scripts for the actual transcribing and cutting, and it has the actual skill.
So I can now open this one. And as you see right here, and once again, you can get the skill, the link is in the description, it has an exact description of every single step it needs to take. So it is extracting the audio, it is transcribing to the milliseconds, milliseconds, it is picking the keeper takes, and as you can see, it's a pretty short scale.
I was really surprised by the quality of the output. So it sees exactly when I say what word, and it will be able to cut out my mistakes. But what it will also do is it will detect waveforms, waveforms.
And that is one thing that is very important as when you have sound that is changing a lot between takes, it sounds very harsh and we don't want that. So what we basically did is the workflow of an exact editor, we pushed that into one skill that can do everything the same level as an editor and even better on a larger scale.
Okay. So it is finally done. And as you can see right here, the first sentence is you didn't start a business just to get a job.
It took the last of the eight takes. Yes. I did a lot of retakes there.
And it created a new folder right here, and I can show you exactly what it looks like. So it has the actual video rendered for me. It has the audio.
It has all the words and everything you need to get a great context for AI, and it has enough context to basically edit the whole video for me. So I can now show you exactly what it looks like. You didn't start a business just to get a job, but somewhere you became the hardest worker.
You cannot be sick, you cannot take the afternoon, and you just cannot leave. You built the whole thing. As you can see, the pacing is really good and this helps a lot, especially when you have, for example, one hour long content.
Then this is saving me a lot of time, but this is one part. So very quick, if you're enjoying this video, make sure to like and subscribe. And now, let's go on with the video.
When I got new footage, I just put in my SD card, Claude reads all the files, and it can add it as b roll. And I will explain how that goes in a second. But what is even more important is how everything comes together, and my videos become really good without any effort.
I'll explain everything in this video. So what basically happens is when we have new content, it makes screenshots per second. So then we have an image like this so Fable five can read the image and understand exactly what is happening in what order.
What then happens is we cut out the important parts. It can be one clip per long video or maybe five. It does color grading and it describes the file.
So what is happening in this specific video? Then it gets sorted by mood, and this is a library that will only keep growing. So we have details, we have freedom, we have grind, we have machine, we have people, place, and products.
And based on this, whenever Claude needs any footage, it just reads all the descriptions and it can match the appropriate clip with the right sentence. So it plans the whole video. So, for example, we have some text.
We add some b roll. We have aesthetic. So this is really how a professional editor also works.
And we have all these components we can choose for based on the video that we need to edit. So we have captions, callouts, and way more. I can now say the following.
Please fully edit it with the correct b roll, animations, music, everything. I can now run this and I will now explain you the fun part about this. So in my finder, I have my Google Drive and I have my b roll library.
So what I really just did is the following. I shot some content, put the SD card in my laptop, and I said, please get the b roll. So the result was the following.
I have my b roll library right here and I have all these files from ping pong to carts to my dog, for example, to kite surfing to whatever. All the b roll that I had was cut up into correct parts and it was also color graded.
But with all these files, there came a description as well. So Claude can now go through all my b roll and find the correct footage. But in the background, it is doing way more.
It is checking exactly where I am. So as you can see right here, it has the frame check. So it knows exactly where I am positioned.
So when it is adding text, it is not putting text on my forehead, for example. So right here, I have the video edit skill. And this is the exact skill from my last video, but then better.
So if I open it, you can see that it's a very long skill with a lot of information for specific parts of the editing process. So in stage two, it is planning the b roll. So it can access all my drive files, read the descriptions of every single b roll, and come up with a plan on what b roll makes sense for a specific video.
And how this flow works is first, it makes a preview, so a low quality render of the actual video. I can now show you exactly what it looks like. You didn't start a business just to get a job, but somewhere you became the hardest worker.
You cannot be sick, you cannot take the afternoon, and you just cannot leave. You built the whole thing, and it just built a cage around you. Everyone says do more.
More agents, more tools, more hustle. But I'm saying the opposite. Build a business that doesn't need you.
Okay. So this is already a really, really good short. And I think, like, there's obviously some feedback.
This b roll is just very bad. So what I will now do is I will remove the b roll, get a better one. But the fact that it is able to choose the correct b roll, add nice text to it, have some zooming on my face, and everything, that is already so cool.
And what we can now do is we can now start a full render. So it will add sound effects, music, and better quality.
So I will do that right now. I will also give feedback to Claude for this file because I think this is a really bad example. So what I then do is I make a screenshot of this clip.
I go to Claude and I say, I don't like this clip. Please remove it and find a better one. I run that.
It will come up with a new version and I can now say the following as well. Also, make a full render now, please. Add music, sound effects, everything.
So, I will run that and I will then show you the final version. And as you can see right here, it is very specific in the time stamps. So, at the end, we have b roll that fits what I'm saying.
So, for example, when I say, so I just dropped it and kept moving, then we pick something where I'm actually moving, and it assembles the whole thing. And it is amazing. I'm not hyping this up.
This is gonna be really big. And the same system is for short and long form. So, also, vlogs, explainer videos, like, everything.
As long as you have this base with b roll and the skills that you can get, the link is in the description, you can make the most amazing videos. So let me show you exactly what the end result is. You didn't start a business just to get a job, but somewhere you became the hardest worker.
You cannot be sick, you cannot take the afternoon, and you just cannot leave. You built the whole thing, and it just built a cage around you.
Everyone says do more. More agents, more tools, more hustle. But I'm saying the opposite.
Build a business that doesn't need you. Fire yourself from the boring work. Just being busy was never the point.
The freedom was. So this is something that looks really good. And the fun part is this was only one video.
I previously did this with thirty minutes of raw footage. And with one prompt, it gave me 13 different clips fully edited. So, if you want to get all of these skills for free, the password is Fable.
Make sure to check the link in the description and you can get started with your own clips very quickly. Thank you for sticking around until the end of this video. I hope you enjoyed it and learned anything from it.
If so, make sure to like and subscribe, and then we'll see each other in my next video.
The Hook
The bait, then the rug-pull.
The title promises a full team replacement. The first eight seconds deliver the thesis plainly, then pivot to a password tease that keeps viewers watching to the end — a clean retention mechanic built into the opening breath.
Named ideas worth stealing.
01:08model
The Cut Skill Pipeline
- Extract audio
- Transcribe to millisecond
- Pick keeper takes
- Detect waveforms
- Render clip
A Claude skill + Python scripts that handles the full rough-cut workflow automatically.
Steal forAny creator with repetitive talking-head content who does multiple takes per sentence
03:20model
Mood-Sorted B-Roll Library
- SD card in
- Screenshot per second
- Color grade
- Describe file
- Sort by mood: details, freedom, grind, machine, people, place, products
A self-building asset library where each new clip filed improves all future video matches.
Steal forAny creator who shoots lifestyle or context footage alongside talking-head videos
How they asked for the click.
VERBAL ASK
08:36link
“the password is Fable. Make sure to check the link in the description and you can get started with your own clips very quickly.”
Well-structured — password withheld for 9+ minutes creates genuine incentive to stay. Free skill download lowers the conversion barrier and seeds the creator into a funnel (build-loop.ai call booking).
MENTIONED ON CAMERA
Visual structure at a glance.
