08:59Luuk Alleman · Demo
AI Just Ended Manual Video Editing (It's Better, Faster and More Creative)
A creator live-builds an entire YouTube video inside a Claude Code chat, then sells the system that did it.
July 16thThe argument in one line.
Feeding two AI video-editing models the same raw footage, the same prompt, and a capped three-round feedback loop, the Anthropic-branded model produced more creative — if imperfect — cuts than the OpenAI-branded one across two separate head-to-head tests.
Read if. Skip if.
READ IF YOU ARE…
- You're evaluating AI video-editing tools and want a concrete side-by-side of how two specific models revise the same rough cut.
- You're building or piloting an agentic content workflow and want a real example of an iterative, timestamped feedback loop with AI.
- You create talking-head video content and are curious whether an AI editor can credibly handle a first-pass edit.
SKIP IF…
- You want a rigorous, blinded, statistically sound AI benchmark — this is one creator's subjective two-round test of his own product.
- You're looking for a tutorial on manual, hands-on video editing technique.
The full version, fast.
Luuk Alleman runs the same raw talking-head footage through Loop Studio, his own AI editing product, using two different underlying models he labels on-screen as Fable 5 (Anthropic) and Sol 5.6 (OpenAI). Both get the identical prompt and up to three rounds of timestamped feedback to fix issues like nonsensical tags, cut-off words, overlapping text, and clashing colors. Round one edits a short intro clip; round two edits a longer uncut recording. Across both rounds the models converge on solid technical execution, but Fable 5 edges out on creative choices, and Luuk calls it the overall winner — while noting Loop Studio works with either model.
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 – 00:19
01 · Cold open: which AI cut this intro?
Luuk reveals the video's own intro was AI-edited by one of the two models under test, and withholds which one until the reveal.
00:19 – 00:32
02 · Setting the rules
Same raw talking-head footage and the same prompt go to Fable 5 (labeled Anthropic) and Sol 5.6 (labeled OpenAI) inside Loop Studio, capped at three feedback iterations each.
00:32 – 03:11
03 · Round one: editing the short intro clip
Luuk runs the raw intro through Loop Studio's review tool, leaves timestamped notes on each version (a nonsensical tag, an awkward trim, a cut-off word, too much visual noise), and loops feedback through three versions before judging the results roughly even.
03:11 – 04:13
04 · Loop Studio goes public
Luuk pauses to pitch Loop Studio directly — now open to the public with a 24-hour refund guarantee — before starting a second, harder test.
04:13 – 05:20
05 · Round two: the full uncut talking-head test
A longer, uncut recording goes through the same three-iteration loop; notes this round cover a weak hook, a requested 3D timeline effect, overly loud music, and overflowing captions.
05:20 – 06:03
06 · The reveal and the verdict
Luuk replays the original intro, reveals it was cut by Fable 5, and declares Fable 5 the more creative model and the overall winner — while noting Loop Studio works with either model.
Lines worth screenshotting.
- Feeding two AI video editors the exact same raw footage and the exact same prompt is the only fair way to compare their editing judgment.
- Capping feedback to three iterations forces an AI editing workflow to prove it can converge, not just eventually get lucky.
- AI editors can over-literalize instructions — one model kept inserting a nonsensical 'editor hidden' tag across multiple cuts until it was told to drop it everywhere.
- Vague creative direction reliably produces overcorrection: asking an AI edit to feature the presenter more resulted in him appearing three times on screen at once.
- By the second and third revision, both AI models handled trimming, pacing, and audio sync competently — the gap between them showed up in creative choices, not technical execution.
- Specific, timestamped notes on a rough cut ('this word is cut off,' 'this tag doesn't make sense') get incorporated more reliably than general direction.
- Text overflow and clashing caption colors were the most common visual defects across both models' output, not editing or pacing mistakes.
- After two separate head-to-head tests on different footage, the reviewer judged the Anthropic-branded model more creative and the overall winner, even though both models worked inside the identical feedback loop.
AI video editors now converge on competent execution — the gap is creative judgment.
WHAT TO LEARN
Once you give an AI editor the same footage, the same prompt, and a few rounds of specific feedback, technical execution stops being the differentiator — creative choices are what separate one model's output from another's.
- Testing two AI tools fairly means holding the input constant — same raw footage, same prompt, same number of feedback rounds — so the comparison isolates the model, not the setup.
- AI edits tend to over-literalize vague instructions, so asking for 'more of me on screen' can produce three overlapping copies of the same shot instead of one better-placed one.
- Specific, timestamped feedback ('this tag doesn't make sense,' 'this word is cut off') gets fixed reliably across iterations; general feedback doesn't.
- Capping revisions to a fixed number of rounds is a useful constraint for judging any AI workflow — it tests whether the tool converges, not just whether it can eventually get there with unlimited tries.
- The most common defects in AI-generated cuts were visual polish issues — overflowing text, clashing caption colors — not structural editing mistakes like bad pacing or missed cuts.
- When two AI outputs are both technically solid, the deciding factor becomes which one made more interesting creative choices, not which one followed instructions more literally.
Terms worth knowing.
- Loop Studio
- Luuk Alleman's agentic AI video-editing product, sold at build-loop.ai, that lets a user upload raw footage and iterate on an AI-generated edit through written, timestamped feedback.
- Vibe editing
- An iterative editing workflow, as used in this video, where a person leaves timestamped notes on an AI-generated cut, submits them, and lets the AI apply revisions in the next version — repeated a capped number of times.
- Fable 5 / Sol 5.6
- The two AI models under test, shown on-screen as built by Anthropic and OpenAI respectively; the video never states their actual underlying model names, only these on-screen labels.
Things they pointed at.
Lines you could clip.
00:00
“The intro you're watching right now was fully edited by AI, either by Fable five or Soul five point six. And I am not going to tell you which one because I don't know yet.”
01:30
“I do think that the quality of the edits are pretty good. I do like the style. I just think that it needs a little bit more direction.”
02:40
“It's almost more about ideation and creativity than the actual skills that Fable five or Soul 5.6 actually have, because they are good enough.”
05:45
“I think Fable five is really the winner here.”
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.
metaphor
00:00The intro you're watching right now was fully edited by AI, either by Fable five or Soul five point six. And I am not going to tell you which one because I don't know yet.
00:11We'll figure that out in this video. I will be testing both models head by head, starting with the same prompt at a maximum of three iterations. All of this to answer one question, which AI is actually better at video editing?
00:24Okay. So we have the following video. And as you can see, I can go through it.
00:28I have a lot of retakes. I have a lot of pauses. Everything that you would expect from a raw take.
00:34So as you see right here, we'll figure this out in the rest of this video. So we have this file and we will give it to Soul five point six and to Fable five. So we can open a new chat, copy paste the exact same prompt as for Fable five, and then run it.
00:49I will now show you the first version. The intro you're watching right now was fully edited by Ai, either by Fable five or Soul five point six. And I'm not going to tell you which one because I don't know yet.
00:57I will be testing both models head to head, starting with the same prompt and a maximum of three iterations. All of this to answer one question, which AI is actually better at video editing? So let's get started.
01:05So I needed to leave a few comments right here. So for example, this one, I didn't like the editor hidden. Sometimes AI really gives these tags an an extra text that just doesn't make sense.
01:17Right here, it once again had this, so my note right here was just leave it out everywhere. Some notes about the trimming right here. I think the six is a little bit cut off.
01:255.6. And I am and everything that I just see, I can leave right here. I do think that the quality of the edits are pretty good.
01:33I do like the style. I just think that it needs a little bit more direction. So I will be testing.
01:38I will be testing. So we leave that as a comment. I do think that this is pretty interesting.
01:43I do really like the comparison between the two. It makes it really clear that we are editing two videos. And right here, this one was pretty fun.
01:50It was pretty creative, but I think we can do a little bit more. So what I do is I click here. So we now say feedback is done and we can now see v two.
01:58And this is the exact loop in Loop Studio that we need to follow when vibe editing our videos. And we then have our v two. So let me show it.
02:05The intro you're watching right now was fully edited AI, either by Fable five or Soul 5.6. And I am not going to tell you which one because I don't know yet. I will be testing both models head to head, starting with the same prompt and a maximum of three iterations.
02:15All of this to answer one question, which AI is actually better at video editing? So let's get started. Is just way too much.
02:20This is way too much. I think it now really overdid the fact that I need to be in this video. So I see myself three times everywhere, and I think that is a little bit much.
02:28That is also the feedback that I have been giving here, too much noise, not clean anymore. We can now go back again and we say feedback. Done.
02:36And we now see the final version. As you see, it's almost more about ideation and creativity than the actual skills that Fable five or Soul 5.6
02:45actually have because they are good enough. So we now have the last version. Let me check it out.
02:49So the intro you're watching right now was fully edited by Ein, either by Fable five or Soul 5.6, and I am not going to tell you which one because I don't know yet. I will be testing both models head to head, starting with the same prompt and a maximum of three iterations. All of this to answer one which AI is actually better at video editing?
03:03So let's get started. Stuff like this is not really that great, but, like, obviously, one prompt to fix it. I do think this looks pretty alright.
03:09I don't like that this is overlaying.
03:11Okay. So very quickly, I've made Loop Studio public. So if you create videos, this is really worth a try, and there is no risk.
03:18If you don't like it, you will get a refund within twenty four hours. So make sure to check the link in the description, and now let's go on with the comparison. So we have the following input.
03:27So we say we have a new video. There's an uncut talking head recording here. So we just start.
03:32As you see, it is going through every single step and then giving us a URL. So we can click it right now and see the first version. So this is what it came up with.
03:40The intro you're watching right now was fully edited by either of Fable five or Soul five basis. And am not going to tell you which one because I don't know how to figure that out in this video. I will be testing both models at by having starting with this.
03:50All of this to answer one question, which AI is actually better at video editing? Perfect. So, is the first version.
03:55As you see right here, I have been leaving notes. And you will see that I think the hook here is pretty annoying. It doesn't really catch attention, and that is my feedback.
04:04Then I went on, and right here, I have the following. And it is that we want some kind of a three d timeline right in the screen. So as you see, I'm really this video and just leaving feedback.
04:15What I can then do is I can click send to Claude. So right here, I say feedback is done.
04:22That is all. Okay. Perfect.
04:23So we now have v two. So let's play it. The intro you're watching right now was fully edited by AI, either by Fable five or Soul Fighter six.
04:30And I am not going to tell you which one because I don't know how to figure that out later. I will be testing both models and by having starting with this, all of this to answer one question. Which AI is actually better at video editing?
04:38Perfect. So, as you see, I have been leaving notes again. For example, I think that the music is a little bit too loud, but also I think we want it even more creative.
04:47If I go right here, I think this looks a little bit weird and I think overall it doesn't really hit. So, as you see, there's way more feedback. Also here, the text is overflowing.
04:57Stuff like this tend to happen. And then right here, I don't like the yellow in combination with the orange and in combination with the blue. So, I think this should just be white.
05:05I can then once again say feedback done. Okay.
05:09So v three, let's see what it looks like. This is the last version and then we go on and can compare it to the OpenAI version. So the intro you're watching right now was fully edited by either by Fable five or Soul 5.6.
05:22And I am not going to tell you which one because I don't know yet. We'll figure that out in this video. I will be testing both models head by head, starting with the same prompt at a maximum of three iterations.
05:32All of this to answer one question, which AI is actually better at video editing? I think this is actually super good. So I think when I compare this one to the Fable five version, I think Fable five is a little bit more creative.
05:45I think overall, Fable five is really the winner here. Loop Studio works with both, so you can just test it out.
05:51See what works best for your use case. Thank you for sticking around until the end of this video. If you've enjoyed it, make sure to like and subscribe, and make sure to check out Loop Studio.
05:59The link is in the description. And then we'll see each other in my next
The Hook
The bait, then the rug-pull.
The video's own cold open was cut by one of the two AI models under test, and Luuk Alleman refuses to say which — until he reveals it at the very end.
Named ideas worth stealing.
01:28model
The Loop Studio feedback loop
- Upload raw footage + one prompt
- AI produces v1
- Leave specific, timestamped notes on the cut
- Mark feedback done
- AI produces v2 (repeat up to 3x)
Luuk treats an AI edit like a code review: instead of one giant prompt, he timestamps concrete complaints (a tag that doesn't make sense, a word cut off, colors clashing) and lets the AI apply them in the next pass, capped at three rounds.
Steal forany AI-generated content review workflow — video, copy, or design — where iterative, timestamped feedback beats a single mega-prompt
How they asked for the click.
VERBAL ASK
03:11product
“make sure to check the link in the description”
Mid-video pitch for his own product, Loop Studio, framed as low-risk with a 24-hour refund guarantee, then repeated as a direct-to-camera pointing gesture at the very end.
MENTIONED ON CAMERA
FROM THE DESCRIPTION
PRIMARY CTAWhere the creator wants you to go next.
Visual structure at a glance.
