Preston Chin turns one coaching-call transcript into a week of carousels, a newsletter, and short-form outlines inside a local Codex project, and shows the exact feedback loop that teaches the system to write in his voice.
Posted
1 months ago
Duration
Format
Demo
educational
Views
1.7K
32 likes
57 · 43
Big Idea
The argument in one line.
A local AI project built around reusable 'skill' files can turn one transcript into a week of carousels, newsletters, and short-form outlines, as long as a human reviews every draft and converts corrections into permanent rules.
Who This Is For
Read if. Skip if.
READ IF YOU ARE…
A creator or coach who already runs calls, podcasts, or long-form videos and wants to turn that raw material into a week of social content without starting from a blank page each time.
Someone comfortable setting up a local AI coding tool like Codex and willing to spend time teaching it their voice through repeated feedback rather than expecting a perfect first draft.
A solo creator looking for a concrete folder structure and workflow for organizing brand voice, visual preferences, and reusable prompts instead of re-explaining them every session.
SKIP IF…
You want a fully automated, zero-touch content pipeline -- this system still requires you to review, correct, and approve every draft before it ships.
You don't already have long-form source material like calls, podcasts, or videos to feed it -- the system repurposes existing content, it doesn't generate ideas from nothing.
You're not willing to install and configure a local AI coding tool -- this isn't a no-code SaaS product.
TL;DR
The full version, fast.
Preston Chin demonstrates the Codex-based system he uses to turn one coaching-call transcript into a week of content: an Instagram carousel, a newsletter draft, and six short-form video outlines. The workflow runs five steps -- enter a long-form source, extract ideas and proof points, draft outlines, review as a human, and log the finished piece to a database. The real mechanic is the feedback loop: every correction he gives, a weak hook, an unreadable cover slide, gets written into a reusable skill file as a standing rule, so the fix carries into every future piece instead of being a one-off edit. He walks through the project folder structure, a live carousel revision, and a newsletter pitch rewrite.
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.
Cold open stating the promise: one coaching-call transcript, repurposed with Codex, becomes a week's worth of content across formats.
00:33 – 03:22
02 · See the five-step content workflow
Lays out the five-step pipeline -- source, extract, draft, review, log -- and stresses that a human must review AI drafts before they ship.
03:22 – 03:48
03 · Download and open the Content System Demo
Points to the free starter-kit folder in the description and shows opening it on the desktop.
03:48 – 07:02
04 · Understand the five key project folders
Walks through the project's input, output, skills, config, and screenshots folders and what belongs in each.
07:02 – 09:56
05 · Build an Instagram carousel from a transcript
Creates a local Codex project, runs the transcript-to-carousel skill, and reviews the first-draft carousel copy and hook options.
09:56 – 13:41
06 · Teach Codex your writing preferences
Gives detailed feedback on the weak hook options and shallow slide-two structure, and has Codex update the skill file's rules before regenerating.
13:41 – 15:14
07 · Add your images and generate the carousel
Drops in personal cover and CTA photos, reviews the visual plan, and lets Codex render the final branded carousel.
15:14 – 16:18
08 · Fix visual mistakes and improve future outputs
Shows a past bad cover with low contrast and too much text, and converts the fix into a permanent rule for cover and CTA slides.
16:18 – 17:52
09 · Log finished content in your database
Logs the approved carousel into a Google Sheet system of record, complete with script doc, asset links, and auto-generated captions.
17:52 – 20:11
10 · Draft a newsletter and email from the same transcript
Reuses the same transcript with a transcript-to-blog skill to draft a newsletter, then adds a standing rule to close every issue with a program pitch.
20:11 – 21:25
11 · Move approved drafts to your website and email tool
Shows the separate, more advanced system that publishes an approved draft to his website and drafts the matching broadcast in his email CRM automatically.
21:25 – 25:07
12 · Generate short-form video outlines
Runs a transcript-to-clips skill that applies proven hook templates to produce six short-form video outlines, then records five videos from them in 45 minutes.
25:07 – 25:26
13 · Turn social media viewers into email leads
Closing plug for a related video on converting social leads to email subscribers, plus a like-and-subscribe ask.
Atomic Insights
Lines worth screenshotting.
A five-step content system -- source, extract, draft, review, log -- turns one transcript into a week of carousels, newsletters, and video outlines.
Feeding an AI content system someone else's video instead of your own breaks it two ways: it's unethical, and it won't sound like you because it isn't built from your own words.
Every piece of feedback gets written into a reusable skill file as a standing rule, so a correction fixes not just the current draft but every future one.
A hook that names the framework up front kills curiosity; a hook that teases a pain point or outcome without revealing the method performs better.
The first body slide of a carousel should deepen the reader's sense of the problem before introducing the tactic, not jump straight to the solution.
A full first-draft Instagram carousel -- three angle options, eight slides, multiple hooks, a call to action -- took about two minutes to generate from one transcript.
A finished, branded carousel with final images, colors, and fonts applied took three to five minutes to render after the copy was approved.
Short-form outlines are deliberately not full scripts, because AI rarely nails word-for-word copy -- hitting the right beats in your own words performs better than reading a script verbatim.
Re-recording a highlighted moment from long-form content outperforms clipping it, because re-recording lets you front-load the hook instead of inheriting the original's pacing.
A transcript-to-clips skill applied six proven hook templates to one transcript and returned six short-form video concepts with outlines in about three minutes.
Five short-form videos got recorded from AI-generated outlines in about 45 minutes total, at roughly nine to ten minutes of talking-head recording per video.
The same transcript was reused across three different outputs -- a carousel, a newsletter, and short-form outlines -- without re-entering any new source material.
A system of record, a database of every finished asset, is what makes future AI repurposing possible, because an agent can't reuse content it can't find.
One correction, stop summarizing the article in the newsletter's conclusion and pitch the paid program instead, got applied as a standing rule for every future issue, not just that one.
Takeaway
One Transcript Becomes A Week Of Content
SYSTEM & FEEDBACK LOOP
A local AI project turns a single transcript into carousels, a newsletter, and short-form outlines, and every piece of feedback you give it becomes a permanent rule for next time.
02See the five-step content workflow
The system has five repeatable steps: enter a long-form source, extract ideas and proof points, draft short-form and carousel copy, review with a human in the loop, then log the finished piece to a library.
Feeding the system someone else's content instead of your own breaks it two ways: it's unethical, and the output won't sound like you because it isn't built from your own words.
Every correction gets converted into a standing rule attached to a reusable skill file, so the fix compounds instead of being a one-time edit.
03Download and open the Content System Demo
The free starter kit is a zip file containing a pre-built folder structure plus skills and config files, meant to be downloaded and dropped straight onto the desktop.
You don't need to learn the folder layout up front -- the AI agent intuitively knows which subfolder a new file belongs in once it's dropped into the project.
Opening the folder in a local Codex project is the only setup step before the first real run -- there's no separate install or account configuration shown.
04Understand the five key project folders
The project folder has five key subfolders: input for raw transcripts and images, output for finished assets, skills for reusable instruction sets, config for brand voice and visual preferences, and reference screenshots.
The config folder is where brand voice, audience definition, writing preferences, and visual preferences live -- stable material that stays constant across every piece of content the system produces.
Dropping reference screenshots of carousels you already like into the project gives the AI a visual starting point to build from instead of generating blind.
05Build an Instagram carousel from a transcript
A 'local' Codex project type is recommended over cloud because it lets the agent modify files directly on your computer, which is faster to iterate on.
Typing a skill name with a dollar sign, like '$transcript-to-carousel', triggers that specific reusable workflow instead of a generic prompt.
A full first-draft carousel, three angle options, up to eight slides, multiple hooks, and a call to action, took about two minutes to generate from one transcript.
06Teach Codex your writing preferences
When giving feedback, the creator doesn't just fix the one draft -- he tells the system to update the underlying skill file so every future carousel inherits the same rule.
The specific fix: hooks that name the technique up front kill curiosity; hooks that tee up a pain point or outcome without revealing the method work better.
A second locked-in rule: the first body slide should deepen the reader's sense of the problem before introducing the tactic, instead of jumping straight to the solution.
07Add your images and generate the carousel
Once the copy is approved, the creator drops in a couple of personal photos for the cover and closing call-to-action slide, then tells Codex to continue.
Codex returns a visual plan showing which image is slotted for which slide before committing to the final render, giving a checkpoint to approve or redirect.
Generating the finished, branded carousel, colors, fonts, copy, and images combined, took roughly three to five minutes end to end.
08Fix visual mistakes and improve future outputs
A past failure case: a carousel cover with text that blended into a dark, non-transparent background, with too much copy crammed onto one slide.
The fix gets logged as a standing rule for the skill, not a one-off edit: no dark non-transparent background on cover or CTA slides, and the text needs more contrast to pop.
Because the correction updates the skill file itself, later carousels need fewer visual fixes -- the creator expects convergence to a good result within one or two rounds.
09Log finished content in your database
The finished carousel gets logged into a Google Sheet that functions as the content database, minting a new row with the script document and the Google Drive asset link.
Captions ready to paste into Instagram are generated automatically alongside the carousel, not as a separate task.
The point of a system of record is that an AI agent can't repurpose work it can't find -- organizing everything into one database is what makes future automation possible.
10Draft a newsletter and email from the same transcript
The exact same coaching-call transcript gets reused a second time, run through a transcript-to-blog skill to produce a complete newsletter draft in under two minutes.
The one correction given: stop the newsletter's closing paragraph from summarizing the article and instead pitch the paid coaching program with a link.
That correction gets applied as a standing rule for every future newsletter issue, not just the one being reviewed -- the same compounding-feedback pattern used on the carousel.
11Move approved drafts to your website and email tool
A separate, more involved system, built with Codex and Claude Code, takes an approved draft and publishes it live on the creator's own website with formatted images and layout.
The same command also auto-drafts the matching broadcast in his email CRM, which goes out to roughly 12,000 subscribers every week.
The high-level point: you should have some process for moving an approved draft onto your own property, even if the exact publishing pipeline shown here is custom-built.
12Generate short-form video outlines
Re-recording a highlighted moment from long-form content outperforms clipping it verbatim, because re-recording lets you front-load the hook instead of inheriting the original's pacing.
The system deliberately generates beat outlines, not word-for-word scripts, because AI rarely nails a full script -- hitting the right beats in your own words works better than reading one verbatim.
A transcript-to-clips skill applied roughly six proven hook templates to the source material and returned six short-form concepts with outlines in about three minutes, which became five recorded videos in 45 minutes.
Glossary
Terms worth knowing.
Codex
OpenAI's agentic coding tool, used here as a local AI assistant that reads project files and generates content drafts rather than writing software.
Skill (in a Codex project)
A reusable instruction file that tells the AI how to perform a specific repeatable task, such as turning a transcript into a carousel; feedback updates the skill so future runs improve.
System of record
A single organized database, here a Google Sheet, that logs every finished content asset so an AI agent can find and reuse past work instead of starting from scratch each time.
CQE
A three-step framework for asking better follow-up questions: context, question, example. Used in the video as a worked example of content pulled from a coaching call transcript.
Local project (Codex)
A Codex project type that lets the AI read and modify files directly on your own computer rather than in the cloud, making iteration faster.
“This is how I automate 90% of my social media with Codex. I take one transcript and turn it into a week's worth of content.”
cold open hook with a specific, credible stat→ TikTok hook↗ Tweet quote
01:38
“AI is not going to get your content at 100% every single time on first passes.”
clean thesis statement for the whole system→ newsletter pull-quote↗ Tweet quote
16:08
“It may take a little bit of time initially with the first few carousels, but over time you're going to get it closer to your ideal output.”
honest, non-hypey framing of how the feedback loop actually works→ IG reel cold open↗ Tweet quote
22:05
“It's better to actually re-record those moments... when you are more intentional about the hook and the structure, they're more likely to perform.”
contrarian, specific production advice→ TikTok hook↗ 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.
17px
This is how I automate 90 % of my social media with Codex. I take one transcript and turn it into a week's worth of content. And I've grown to over 750 ,000 followers across Instagram, TikTok, and YouTube with this exact system.
So in this video, I'll give you the step -by -step instructions to turn one coaching call transcript into a week's worth of short form video outlines, an Instagram image carousel, and a newsletter and email draft. By the way, if you're new, my name is Preston and I share on this channel everything I building with ai to grow my business so you can too let's start with a bird's eye view of how this system works there are five key steps starting with the first step entering your source into the system this is long form content ideally your own like youtube videos podcasts or coaching call transcripts and then dropping them into codex where codex will extract the raw material from this long form content to repurpose into different formats ideally ideas hooks proof examples one call out i recommend is to focus on long form content that you've created not taking another creators because one that's just unethical to just steal someone else's content and two it won't sound good because it won't be in your own voice so it's much better to take what you've said in your own words and have ai just structure that now
Codex is going to then draft the short form video outlines, not scripts, which I'll explain why in a second, image carousels that you typically see go viral on Instagram and blog posts and email drafts that you can send to your audience once a week. This is where the magic happens and it's where the system does a lot of its work.
But then the part of the system that not enough people emphasize is having a human in the loop because AI is not going to get your content at 100 % every single time on first passes. So you need to actually have a system for how you give feedback to this content engine such that the engine self -improves. And to explain this, I'm going to walk you through four simple steps within a concrete example.
Let's say that you get a script outline and the hook is written in passive. voice. So you read it and you say, oh, I really think this feels off.
This should sound more like active voice. Well, then you're going to name that error and you're going to say, I want to turn this into a rule that's going to get applied to what's called the improve skill. This skill is just a set of reusable instructions that Codex can reference to.
for example create a script outline so you're just going to say i want you to apply this rule to this draft and i want you to modify the improve skill accordingly you can say this at the very end after you've given all of your modifications by the way but then what's going to happen is you're going to make those changes you're going to review it you're going to apply the skill and it'll just keep improving that way.
Then take that refined copy and get it logged into a system of record, your library. It can include captions, scripts. images, all the material, even video files that you need organized in a single database.
It's so important to have a system of record as a creator because without one, you're going to be really scattershot. AI is not going to know your full context to be able to repurpose things for you. So if you keep things really organized, it allows you to create content at scale.
Now let me walk you through this setup. You're going to see a link in the description which contains a zip file to a folder called Content Assistant Demo, and you can open it. print it in your desktop, and this is what Codex is going to reference.
When you click into it, you'll see a bunch of folders. Now, you don't need to understand how all these folders work because, frankly, Codex is going to do a lot of the work for you. Whenever you want to drop certain files into certain folders, you don't even have to tell Codex where to drop it.
It will intuitively know where to put things. But there are five key folders that I think are worth highlighting. The first one is input.
Input is just where you drop those long -form source materials like, let's say, coaching call transcripts like I did. did here, or even images. For example, if you have images of yourself sitting, doing certain things, acting, moving around, those are great images to include as covers for your image carousels, and you can drop them in there.
Then you have output. That's pretty straightforward. It's just where the final output from the system goes.
You have skills. So these are reusable skills. Like I told you, there's the improve skill right here.
You also have call to clip. So this is the skill that translates your transcript into short form video outlines you have transcript to blog so you're going to literally enter these skills into codex that's going to turn it into a blog draft for you to review and then transcript to carousel as well and if you look into some of these skill files you're going to see a lot of detail in here about how we actually want to structure these the copy as well as the images but you don't need to understand what goes under the hood you're going to get these templated skills out of the box and you can eventually customize them to make them your own.
Now, the next piece of this system is the config. So the config is what are all of the materials that will stay fairly constant throughout your system. They're usually in a few buckets.
The first one I have is brand. So brand represents my audience. You can say, who am I trying to help?
What's my voice? What's my point of view? Writing preferences, visual preferences.
You'll notice that this is tailored to my brand. But if you want to create a new brand profile for yourself, you can simply tell Codex to interview you based off of my brand profile. You can tell it all the things that are custom to your brand.
You can even give it in. images or websites you've created or content you've created, the transcripts, and it can extract those things to craft your own brand profile. You can also have the carousel preferences baked in here.
So you'll see that I've baked in preferences that are tied to my particular brand colors that I prefer fonts as well. You'll notice that I have image folders, video folders, cover safe images, certain things that are custom to me, but you can also tell. codex to interview you if you want your carousels to have a different look and feel than mine but you'll notice here i have certain rules like i want to avoid dense paragraphs multi -competing accent colors decorative labels that don't teach anything polished looking sly with no source evidence i want to avoid these things typically a lot of my clients like this look and feel but if you don't like it you can have codex interview you to modify those things so aside from the config the last piece are the screenshots so these are screenshots that you could include from other carousels that you like online i just drop them in here i've included a few from my own carousels in case you want to build off of these this is an example of a cover style this is an example of body images that go in the carousel but either way you have a starting point to work from
All right, now here's the end -to -end workflow to build a carousel. You're going to go into the Codex app. So once you've downloaded it from ChatGPT, you'll be able to click here to Codex.
Then you're going to go to the project section and you're going to click on the plus sign next to projects. From there, I strongly recommend using the local project type because this allows you to actually modify files directly on your computer, which means that you can move much faster. then i like to put in the project name in this case i'll just call it content system demo and then adding the folders is pretty straightforward just click here and i've already got it loaded onto that same project folder i walked you through in the previous section so i'm just going to go here and click on open then i'll click on create project and then from here you're going to see that the project gets created right here so i'm going to click on this new icon and now i have a new project within the content system demo next i'll drop in my transcript from my coaching call just like this and i will type in the skill with the dollar sign and then transcript to carousel and from here i'll make sure that the model is selected to sole which is codex's strongest model i think high effort is good but you could also do medium effort if you're willing to wait a little bit longer and then i will click go
now from here codex is going to parse through the entire transcript looking for those key ideas that it can repurpose into possible carousels and then it will give me drafts for me to review so it took about two minutes to create the carousel copy so the document will contain three different angles for carousels these are three possible script options for the text on your image carousels so you'll see that it's got up to eight slides it's got a clear call to action and it's even got here's the viewer problem the promised results some hook options i'm going to now walk you through how i would actually interact with ai when i get this first draft because it's not often perfect so let's look at angle three ask better follow -up questions with cqe cqe just happens to be one of my frameworks that i taught my students in this group coaching call so i'll notice here the viewer wants to learn what another person needs but a string of questions can feel like an interview or a sales pitch this angle teaches cqe a three -step way to make follow -up feel heard useful and natural okay interesting so the hook options this is what i'm really looking at because this is going to be the first text that shows up on the image carousel three steps to
and ask better follow -up questions. That's like, okay, it doesn't reveal the technique, but it's not like as sexy for me. It doesn't really create enough curiosity.
Use CQE to make your questions feel like a conversation. Not great because it gives away the framework. Do your follow -up questions make people feel heard or interviewed?
That's like an okay question, a tee up, but I think we can do better. So what I'm gonna do is I'm going to think of a better hook that's kind of inspired by what I'm seeing here, but it's slightly different. And then I'm going to ask, codex to actually adjust the way it writes hooks for carousels accordingly.
So watch what I do. So I want to go with angle three, but I want to give you feedback on each of the hook options you've given me so that we can update the rules for how you craft hooks accordingly for carousels. So I think that the second one, use CQE to make your questions feel like a conversation.
I don't like that we're giving away or teeing up a framework. I'd much rather tee up a pain point, a desired outcome, create more curiosity without actually revealing a possible technique because it's not giving the user enough context. And then the third one, do your follow -up questions make people feel heard or interviewed?
I think that we're assuming here that follow -up questions is what people naturally think of, but I think it's better to tee up something that's even more dumbed down, relatable, simple. I think the hook I want to do is something like this. The secret to sounding more conversational and casual?
as the big main text then as i'm reviewing the rest of the copy i generally like it it's piecing together all the context the examples the the different parts of the framework and i think it's pretty straightforward the only thing i would add is right before we get into the framework i like to tee up more context around the problem like most people tend to do this instead of that and so that's the feedback i'm going to give to codex now I would like to insert a slide in between the first and second slide that gives more context related to the problem.
Like what's a more common mistake that people tend to do? And I think the most common mistake that is not mentioned here is people like to ask a bunch of questions in succession because they're afraid of the silence. But the key is to only ask one question and provide people with more context.
Here's how. So that's what I want to be. articulated in the second slide.
And in general, as a rule, I would like the second slide to not jump right into the tactic, but instead to give some more context around the problem that the user can relate to. I'm going to fire it away and I'll just, let's modify the angle three because that's what I want to go with. Okay, so now I'm just going to fire it off and wait for the engine codex to give me a better angle three.
So let's see what comes back. So now you can see that Codex is using the same transcript to carousel to revise angle three, and it's modifying the agent feedback. So every time I give it feedback, it's going to modify the skill accordingly so that the next time it can get a little bit closer to my ideal output.
And let's see what it generated. So if I look at the updated hook rules, it's going to walk me through. some of these rules lead with a familiar pain point or desired outcome awesome do not assume the reader uses expert labels describe simpler relatable situations until the source proves that label fits audience that's exactly what i was looking for and then cover may pair a broad outcome headline with plain language subhead that promises the lesson so that was also implied based off my feedback with a specific example so it seems to have captured a lot of my feedback including here slide two should deepen the problem before teaching the tactic so Let's go ahead and look at the revised carousel copy.
I'm going to go right down to angle three. And it looks like this. The secret to sounding more conversation and casual.
You need to ask questions the right way. Here's how. The most common mistake.
You ask one question, then silence gets uncomfortable. So you ask another and another. Now it feels like an interview.
Ask one question, then give context. Here's a three -part structure. Context, question, example.
It helps one question feel like a conversation. awesome and then it goes through the rest and it sounds great so now i'm at the point where i think the copy is pretty good all i'm going to do is drop some images of myself that i want to be repurposed as the title screen or the call to action screen at the very end like these two and i'm just going to drop them in here like this and i'm going to say continue forward and then enter.
So now what it's going to do is take my images that I've provided it, and it's going to apply that copy to those slides and create the body copy as well. So let's give it a few minutes and then I'll come back with the results. All right, now that the copy is approved, it's going to give me a visual plan.
And so if I click into the visual plan, it'll actually show me the images that it's planning to use. So this was going to be... 1925.
This is going to be the first image. And then the final image, which will be the call to action, will be the image that looks something like this. I think that looks pretty good.
So I'm going to say proceed. And from there, it's going to take those images along with my general instructions for how I like carousels done, the colors, the fonts, and the copy. And it will combine it in about three minutes or so to create a new looking carousel.
So let's see what it looks like. okay after almost five minutes here's what the carousel looks like you'll see it looks really nice it pops out really nicely it's got clean copy so it's very easy for me to read and i like how it's formatting different parts of the framework and it's using my colors and my fonts so i'm really enjoying this and then yeah a nice cta where the text pops up and a clear call to action so i'd say that's pretty good now if you don't like how it turns out the first time that's very normal so what i've done is i wanted to share with you some poor examples it's created in the past and so how I would modify it.
So if it created something like this, which it did at one point. I personally think that this font just blends too easily in with the background. And then as I reviewed this, I was like, there's too much text here on the cover.
So I'd like to simplify it and I'd like to change it as a rule. So then you'll notice here, I said, I want to modify the cover in CTA slides, not just for this one, but for carousels moving forward. So that's a cue.
I'm not just giving it feedback for this round. I want it to modify the skill. The biggest thing is I don't like having a dark background that's not transparent.
I want the text. to pop more. I want there to be more contrast.
I don't want there to be this faded effect. And so as you give your system more feedback, it may take a little bit of time initially with the first few carousels but over time you're going to get it closer to your ideal output after maybe just one or two rounds of iterations now that the carousel is finished i want to make sure that it's logged into my content database so that i always have a record of it if you're a content creator who publishes regularly you should have all of your files all of your scripts all of your image all your videos organized in a database because then it's a lot easier for agents like codex to repurpose them automatically for you what i did is i asked codex to now put my content carousel in my content engine in about three minutes it was able to automatically add it to my google sheet which basically comprises of my database you can see here and i have a carousels tab and here all the carousels that i've ever created but but it just minted a new row, C3297, and you can see here it's clearly the same.
carousel that we built together the secret to sounding more conversational and casual and then within this particular row you'll see the script document which literally contains the copy that we included in that slot in that slideshow and then the google drive link which also includes images here and you can see each individual carousel slide which is pretty awesome and on top of that it automatically created captions for me that i can include in my instagram post.
So all of this is now ready for you to go. And if you want this particular engine built for you so that you have a database of all of your content that you can easily manipulate, easily repurpose, just make sure to click the link in the description where you can work with me, set up a one -on -one, and I can talk about ways we can set this up.
But now that you've seen this, let's actually turn your coaching call into a newsletter issue on your website and an email draft you can send. so the first thing we're going to do is now click on the new button within the same content system demo project and all we're going to do is just drop in that same coaching call transcript that i added and then we're going to type in the skill transcript to blog and from here it's going to extract a core idea from this coaching call transcript and using the skill that's already in the project create a blog post that's in my ideal style It is easy to read.
It's not too long, but it's great to add to my website. So let's see what it comes back with. Okay, in just less than two minutes, I already have a complete draft for my newsletter.
And I spent a lot of time refining the newsletter prompt to sound like me using direct language, short sentences, making it easy to read. And as I read this, it's definitely a reflection of my conversation with my client. And it's got a great intro.
It's backed by the story, the real experience. that I had with him. And then it gives some tactical sentences.
You can see here it's formatted. Here's the main points. Do not ramble in sentence one.
Use this particular tactic. Here's some examples. Give the rest of your answer a shape.
Here's the framework I like to follow. I think this is great. And so if I were to only give it like one particular piece of feedback, I would say that I would prefer to have a call to action to subscribe to my coaching program, the Impromptu Speakers Academy, at the end and to not just regurgitate or summarize the article.
So I'm going to say that here. I would prefer that the conclusion section not just regurgitate or rephrase the same part of the article. I would rather have a call to action for people, if they found this helpful, actually enroll in my Impromptu Speakers Academy and include a link to that and make sure that that call to action, that soft sell is included in every one of these newsletter issues.
Okay, so now that I have added that context, I'm going to adjust the skill accordingly, and it's going to make that change. And in about a minute, I'll get a revision.
And once I like it, there's one last simple step I'm going to share with you. Now, even though this is not part of the setup I've given you, there's a lot more that is involved in creating this. I want to show you the power of having a system that your agent deeply understands.
For my entire website, this was something I built with Codex and Cloud Code. It has every single article that I've ever published. So a simple command that I give it to take a draft and turn it into a newsletter makes it look something like this, a clean website images, as well as cover blog posts, and then clear formatting like this exactly as I had it.
And I even get it to automatically draft a broadcast like this in my. kit, which is my email CRM. This is the thing that goes out to my subscribers.
I've got about 12 ,000 of them every single week. So if you want to know how to figure all of this up so that it's just a simple one line command to get all of this published on your own properties and definitely recommend you reaching out to me so we can work one -on -one and getting you set up. But the high level here is that you should at least have a process for taking your drafts and bringing them to your own property so that they can be easily sent.
The third automation is to turn your long -form content into short -form outlines or scripts. Instead of trying to clip various moments from, let's say, an interview with a podcast or a long -form YouTube video, it's better to actually re -record those moments. And the reason why is because when you are more intentional about the hook and the structure, they're more likely to perform.
You can see a few examples where I got over 3 ,600 likes on this video alone just by reformatting it, re -editing it, making sure that the key value is up front. And what I've even done is created a few options for you to do this as well.
So you could go to melda .co. This is a website that I actually built to solve for this very problem. You can actually do this for free where you can get onto the free trial and just put in a YouTube transcript or a YouTube link or an Instagram link and turn that into your own ready to record scripts.
Just add your idea. We'll do a lot of research for you. You can get pretty far.
But the other way you can do this is you can actually take this philosophy of turning those viral moments, getting the full transcript, getting those moments, getting a short list of hooks you can use, and then getting the read aloud outline and doing it within the system yourself. So the way you do this is you're going to drop in inside of Codex in a new task, just drop in that particular long form script, and then you can type in transcript to clips.
like this. And from here, we're going to extract outlines based off of some winning hook templates. So we're going to take the idea.
We're then going to apply the hook template to the idea, about six of them. So you can see different hook options to consider and outlines that cover the core beats of your script. Now we're not going to give you word for word script because in a lot of instances, AI can't get it a hundred percent right.
But if you understand the beats that you need to hit, then when you record, you can reference those and make each beat sound like your own. So it took about three minutes to generate this review board.
So let's look into it. You can see here that we've got a bunch of different concepts that it pulled. It's got a working title, ask the room before slide one, a viewer problem, and a practical result.
So if I look at each particular clip, it's pulled out six. I want to see if the hooks sound pretty decent. So here we go.
It's got some hooked templates that it used, two questions to ask before you present to a small room, or don't open a presentation by launching into your slides. Why? Because you lose the chance to learn what the room actually needs.
Here's what I mean. Okay, cool. So it's got some pretty decent hooks, and you've got that outlines.
So I feel like this was something I could easily hit on because it's jogging my memory on the key concepts that I hit on with my client. And so I could easily go in and record this. You can even see here that I actually took a lot of these concepts and I recorded a bunch of them already.
In fact, these are just five videos just to prove it to you. Five videos that I recorded of myself that were simply me. going off the outlines and taking about nine, 10 minutes per video to just record my talking head footage.
And then I just send this off to my editor and he's off to polish it up and then we're done. So in about 45 minutes, I was able to generate about five videos based off of these outlines. And then yes, it's not a perfect script, but if you want, you can get a script and you can just tweak it a little bit, do the same thing I talked to you about refining it with your feedback.
And you can get scripts that are pretty decent. You can read off a teleprompter. I myself recommend sounding more natural by going off an outline, but it's totally your call.
So that's really the simplest way you can get a ton of outlines to record a bunch of videos in one sitting. Now, if you found this helpful, but want to actually get more leads to your business from social media, check out this video right here, where I walk you through how you can build a many chat template to take your leads from Instagram to your email list.
And as always, really appreciate you watching. Please like the video and subscribe to the channel. All right, I'll see you next time.
The Hook
The bait, then the rug-pull.
Preston Chin opens with the promise that drives the whole video: one coaching-call transcript, repurposed by a local AI system, becomes a week of carousels, newsletter copy, and short-form video outlines, and he walks through the actual folder structure and feedback loop that makes it work.
Frameworks
Named ideas worth stealing.
00:33list
The Five-Step Content Workflow
Source
Extract
Draft
Review
Log to system of record
Drop in a long-form source, have the AI extract ideas/hooks/proof, draft multiple formats from it, have a human review and correct the drafts, then log the approved piece into a database.
Steal forany repurposing pipeline built on coaching calls, podcasts, or long-form video
08:48acronym
CQE (Context, Question, Example)
Context
Question
Example
A three-step way to make a follow-up question feel heard, useful, and natural instead of like an interview: give context, ask one question, give an example.
Steal forsales/coaching follow-up questions, or any 3-part explainer carousel
CTA Breakdown
How they asked for the click.
VERBAL ASK
25:07subscribe
“like the video and subscribe to the channel”
soft end-card plug for a related video on turning social leads into email subscribers, paired with a direct like-and-subscribe ask after a 25-minute value-dense tutorial
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.
A screen-recorded walkthrough of HyperFrames, HeyGen's free open-source video editor, built entirely from prompts inside a coding agent like Codex or Claude Code.
A screen-share walkthrough of using Codex to plan and build automations, then hosting the generated code on Trigger.dev instead of burning your weekly usage limit.
A full screen-share build of a self-hosted n8n automation that scrapes viral clips, generates an AI avatar video with a cloned voice, and publishes it to eight platforms — narrated step by step by the creator who built it.
A breakdown of three automation buckets — sales, research, and content — built on Claude Code and an indexed Obsidian vault, reclaiming five to ten hours a week.
00:14
00:23
01:01
01:19
01:38
01:57
02:16
02:35
02:53
03:26
03:35
03:58
04:20
04:40
05:00
05:25
05:49
06:08
06:32
06:59
07:31
07:50
08:09
08:29
08:46
09:07
09:26
10:07
10:29
10:52
11:14
11:37
11:59
12:22
12:44
13:07
13:29
14:08
14:27
14:52
15:01
15:41
16:07
16:46
17:02
17:23
17:42
18:21
18:41
19:01
19:21
19:41
20:01
20:38
20:57
21:36
21:47
22:20
22:42
23:04
23:27
23:49
24:21
24:33
24:55
25:16
25:22
27:45
35:25
57:08
19:49
Video of the Day14:45
22:25