Big Idea

The argument in one line.

The unglamorous parts of a YouTube upload, title testing, clean captions, and researched tags, can be fully automated by a Claude Code skill that reads a video's transcript and cross-references VidIQ data, replacing a creator's single guess with a data-backed shortlist.

Who This Is For

Read if. Skip if.

READ IF YOU ARE…

You upload to YouTube regularly and currently write one title, skip proper SRT captions, and reuse old tags out of habit.
You already use Claude Code or Codex for other tasks and are comfortable installing a skill file and running it against a local video.
You run (or want to run) YouTube's built-in A/B title testing but don't have a system for generating real alternatives to test.
You care about mobile viewers who watch with sound off and rely on captions to decide whether to keep watching.

SKIP IF…

You don't use Claude Code, Codex, or any tool capable of running local file-reading skills, since this only works inside those environments.
You upload to platforms other than YouTube where tags and A/B title testing don't apply.

TL;DR

The full version, fast.

Vic built a Claude Code skill, YouTube Upload Pack, that takes only a video file path and produces the three assets most creators skip: A/B-testable titles, broadcast-clean SRT captions, and researched tags. It transcribes the video, cleans the transcript for grammar and proper nouns, then generates fifteen titles from three sources, similar competitor videos, the creator's own top performers, and built-in title rules, each scored by character count for YouTube's 55-character display limit. Using the VidIQ MCP, it pulls a broad set of niche tags plus a narrower set of specific-problem tags, all under YouTube's 500-character limit. Everything lands in one HTML dashboard for copy-pasting straight into YouTube Studio. Vic shows a real prior upload where the skill's suggested title won the live A/B test with a 7.8-9% peak CTR.

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 – 00:44

01 · The upload problem

Framing: exporting and uploading is easy, but title variations, clean SRT captions, and researched tags are the overlooked parts of a serious upload.

00:44 – 01:31

02 · Meet the skill

Introduces the Claude skill (works with Claude Code or Codex) that generates all three assets from one video file path. Channel intro.

01:31 – 02:51

03 · Live demo setup + proof it worked

Vic is mid-upload on his own channel. Cuts to a 'future Vic' segment showing the AI-suggested title won the real A/B test, peaking at 8-9% CTR.

02:51 – 03:37

04 · Installing the skill + VidIQ requirement

Copy the video file path, drop the skill into Claude, tell it to install. Skill depends on the VidIQ MCP, which it installs if missing (requires a free vidiq.com account).

03:37 – 04:34

05 · How the title engine works

Skill transcribes the video, cleans the transcript for grammar, then generates 15 titles across three strategies: competitor-title pull, channel top-performer pull, and built-in title rules.

04:34 – 05:46

06 · Tag research: broad then niche

Once titles and SRT are done, the skill scans VidIQ for two tiers of tags: broad industry/niche tags, then a narrower sweep for the video's specific problems and search terms. Mid-video pitch for Vic's content-strategy calls.

05:46 – 07:04

07 · Reviewing the generated files

Output folder appears beside the video file with separate docs for tags, titles, and the SRT. Proper nouns (like 'Alex Hormozi') are spelled correctly and capitalized properly.

07:04 – 07:49

08 · The dashboard: comparing title buckets

One HTML dashboard bundles everything. Vic compares his own title against the three influenced-by-competitor, influenced-by-channel, and rule-based title buckets, each with character counts.

07:49 – 09:52

09 · Publishing in YouTube Studio

Sets up the A/B title test with three options, writes the description, adds the thumbnail and playlist, and pastes the researched tags straight from the dashboard.

09:52 – 10:59

10 · Captions upload + monetization

Uploads the generated SRT as timed captions, sets English as the caption language, and places ads roughly every three minutes since the video is monetized.

10:59 – 11:43

11 · Free download + outro

Skill is free inside Vic's Skool community under the AI Automations classroom section. Recap that the AI titles beat his own, then standard subscribe CTA.

Atomic Insights

Lines worth screenshotting.

A YouTube upload has three assets creators routinely skip: alternate titles for A/B testing, a properly punctuated SRT caption file, and tags researched for that specific video rather than reused from an old upload.
Most mobile viewers scroll past a video's audio entirely and judge it by the captions alone, so caption quality functions as a first impression, not an accessibility afterthought.
A title generator that pulls from three separate sources, competitor titles, your own top performers, and general title-writing rules, produces five testable options per source instead of one guess.
YouTube's title field has a 55-character practical limit for full display, which is why every generated title option is scored by character count before you pick one.
Tag research works better as two passes: broad industry/niche tags first, then a narrower sweep for the specific problems and search terms unique to that one video.
Dictation and auto-captioning tools routinely misspell proper nouns (like a guest's name); running the transcript through an LLM before generating the SRT catches and fixes those errors.
In Vic's real A/B test, the AI-suggested title beat two human-written alternatives, largely because it was generated from the same title-writing principles the creator already follows, applied more exhaustively.
Leaving YouTube's native tag field blank until the tag-research step is done avoids the common mistake of tagging a new video with leftover tags from a video made months earlier.
A skill built for Claude Code or Codex only works in those environments because it needs local file read/write access, so it can't run inside the regular Claude.ai chat interface.
Placing an ad roughly every three minutes is the monetization pattern shown for a video slightly under twelve minutes long.

Takeaway

A single video file path can generate every upload asset you skip

WHAT TO LEARN

The parts of a YouTube upload creators routinely skip, title variants, clean captions, and fresh tags, can be fully generated from a video's own transcript plus a tag-research data source.

02Meet the skill

A skill built for Claude Code or Codex only works inside those environments because it needs local file read/write access.

03Live demo setup + proof it worked

In a real A/B test, the AI-suggested title beat two human-written alternatives, peaking at 8-9% click-through rate.
Watching results from a real prior upload before explaining the mechanism is a stronger opener than explaining the mechanism first.

04Installing the skill + VidIQ requirement

The skill's only required input is a video file path; it does its own transcription rather than requiring a pre-made script.

05How the title engine works

Generating titles from three separate sources, competitor titles, your own top performers, and general rules, produces testable variety instead of one guess.
Every generated title is scored by character count because YouTube's title field has a practical 55-character display limit.

06Tag research: broad then niche

Tag research works in two passes: broad industry tags first, then a narrower sweep for the specific problems and search terms unique to that video.
Leaving the native tag field blank until research is finished avoids reusing stale tags from an old upload.

07Reviewing the generated files

Dictation and captioning tools routinely misspell proper nouns; running the transcript through an LLM before generating the SRT file catches and fixes those errors.

08The dashboard: comparing title buckets

Bundling every generated asset into one HTML dashboard, rather than separate files, makes it faster to copy-paste directly into YouTube Studio.

10Captions upload + monetization

Most mobile viewers judge a video by its captions rather than its audio, so caption quality functions as a first impression.
Placing an ad roughly every three minutes is the monetization pattern used for a video under twelve minutes long.

11Free download + outro

A soft gate behind a free community join can distribute a tool as effectively as an email-gated download, without the extra friction of a form.

Glossary

Terms worth knowing.

Claude Skill: A packaged set of instructions that runs inside Claude Code or Codex, giving the AI a defined multi-step task (like reading a video file and generating upload assets) instead of a one-off chat reply.
MCP (used here as VidIQ MCP): A connector that lets an AI tool call an external service's data directly, in this case pulling YouTube analytics and tag-research data from VidIQ into the skill's workflow.
SRT file: A plain-text subtitle format with timestamps, used to upload closed captions to a video platform so the exact wording appears synced to the audio.
A/B title test: A native YouTube Studio feature that rotates two or three different titles on the same video to different viewers and reports which one drives a better click-through rate.
Click-through rate (CTR): The percentage of people who saw a video's thumbnail and title in feed or search and actually clicked to watch it.

Resources

Things they pointed at.

02:52toolVidIQ ↗

10:59linkFree Skool community (skill download) ↗

01:30linkContent Solutions call booking ↗

Quotables

Lines you could clip.

02:59

“I am from the future. I posted this video a week ago now... the title this skill wrote actually ended up winning the AB test, and it gave me a pretty solid click through rate.”

proof-of-results moment, breaks the tutorial format with a time jump→ TikTok hook↗ Tweet quote

06:33

“Like, even Alex Hormozi has become the GOAT of online business. Like, it has GOAT in all caps. Alex Harmosi is spelled properly.”

concrete, funny proof of the caption-quality claim→ IG reel cold open↗ Tweet quote

09:05

“The titles that it created were genuinely, in my opinion, better than the titles that I created because they were just sourced off of more data.”

clean summary line, works as a pull-quote→ 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.

metaphor

So you just finished editing your YouTube video, you've exported it, and now it's time to upload the thing. So you come over to YouTube Studio, you drag your file right here, and starts the whole upload process. If you're taking YouTube seriously, you should already have your title and thumbnail mapped out, but what about all of the other little components?

I'm specifically talking about title variations to test an extremely accurate and grammatically correct SRT file that does these closed captions that happen at the bottom, and research backed tags that are specific to this video, not like the last video or a video you made six months ago where you just have those tags still.

This is one of those YouTuber nuances that often get overlooked, but they're still important. So I created a Claude skill that also works with Codecs.

That gives us those three important upload assets for every single video, and all we have to give Claude is our actual export file path. In this video, I'm gonna show you exactly how it works, how it exports this beautiful dashboard with all the assets baked right into it, and I'll even give you this AI skill file completely for free and show you where to download it.

My name is Vic. Content is my business. I've gotten over 200,000,000 views across my network of channels, and I've generated millions of dollars as a content creator.

I use this channel to focus on making AI systems for human content. So if that's something you're looking for, make sure to subscribe and hit the notification bell. I am literally right now uploading my last video, Like, the last video on my channel.

So I thought what a perfect time than to show you how this skill works. So right here I have the actual video file that I started uploading to YouTube. I'm lit I literally am uploading it.

Like, is happening in real time. It has 53% left.

Okay. Time out. Time out.

Time out. This is future Vic. I am from the future.

I posted this video a week ago now, and I'm sorry. I'm I'm a little sick. I'm funny and cold.

That's why I got a nose drip, but I wanted to show you something. The adjustments I made from the skill that I used in this video actually gave me some really good results. In fact, you will see that the title this skill wrote actually ended up winning the AB test, and it gave me a pretty solid click through rate.

Right now, it's at 7.8, but we peaked around eight and nine percent. So before I show you how it works, I just wanted to show you that it does work. Anyway, back to the video.

And all you need to do is come over to the video file and right click on it, copy as file path. On Mac, it'll be like near the bottom. There'll be your file path.

You just right click copy there. And just like all the other Claude or Codex skills that I create, these only work with either Claude code or Codex because they need to write files and read files on your PC. So you just download the zip file for the skill that I'm gonna give you at the end of the video, drop it into Claude, tell it to install it, and you should be good to go.

So the skill is called YouTube upload pack, and it does rely on the vidIQ MCP, which if you don't have that installed yet, the YouTube upload pack will install it for you.

You just might need to log in and make an account on vidiq.com, which gives us super detailed YouTube analytics that the skill uses to help you create some title variations and also finds all the tag data. So this is super important, very easy to plug in.

So YouTube upload pack is the skill, and then I just drop the file path, and that's it. What the skill is going to do now is scan your video, create a transcript file. It will then clean up that transcript file and make it grammatically perfect, which is crucial.

It's then going to understand understand the the context of your video, and it's going to create 15 different title variations using three different title strategies.

The first strategy is that it's literally going to go to YouTube, find similar videos, and pull from their title structures. The second strategy is it's actually going to go to your YouTube channel if it knows what your channel link is, and it's going to look at your top performing videos and what your title structures are.

And then the last set of titles are going to be created based on title rules that I actually programmed into the skill. So five titles for each set. And then it's going to pick the best option and basically suggest that you run an AB title test from what you planned in advance to what it just gave you.

So you're getting a free AB test out of it. Once all of the title work is out of the way and it has that transcript grammatically perfect and baked into a time stamped SRT file for you to upload as your captions file, it's then going to scan VidIQ for the best performing tags that have to do with your specific video.

It does this in two different sets. The first set are more broad tags, like the general industry, the general niche, some similar topics. And then in the second set, it does a much more niche sweep.

So the little nuances, the specific problems and solutions that have to do with your video that people are actually searching for. And you can see right now, it's actually scanning my channel for my title structures, and now it's scanning all of vidIQ.

You can see this video is about Alex Hormozi creating an AI with Alex Hormozi's transcript. So you can see it's searching for Alex Hormozi tags, Claude code, Claude skills. This is all based on what I talk about in the transcript.

Now as Claude is cooking, I wanna let you know that my team just opened up our calendar, and we're now taking calls specifically with experts or professionals who wanna grow an online presence and authority, create value based content, and turn that into a powerful online business. Now I don't know how long our calendars will be open because we can only work with so many people at a time, but if the link is still in the description to book a call with us, that means that the calendar is still open, but slots are limited.

Okay. So Claude is all done. It did the research.

It put together everything, and it actually saved right in a folder beside the video. So if I open this up, I have the tags in a doc, I have the titles, all of the titles in a doc, and the actual SRT file that I'm gonna use.

And you can see how clean this is. Like, even Alex Hormozi has become the GOAT of online business. Like, it has GOAT in all caps.

Alex Harmosi is spelled properly. Like, this SRT is so dialed, and and a lot of people don't really consider that the closed captions should be super professional. But what happens is that most people, especially on mobile, will scroll past your video, and they're not gonna hear your video in preview.

They're just gonna see the captions at the bottom. So making sure your captions are really, really proper gives a really good first impression and also properly represents your video. Like, Alex Hormozi often gets spelled wrong by dictation tools.

But since we're using Claude to edit the transcript and build this SRT, it's very formal when it comes to these proper nouns. Then lastly, we have the dashboard itself, which I'm gonna open up. So this is all of those files baked into one dashboard.

And the title that it thinks we should use is I trained Claude on 2,000 Alex Hormozi videos, which is very similar to what I'm using. The title that I came up with was I trained AI on 2,039 Alex Hormozi videos, and it's insane.

I honestly think that this is a little cleaner for sure, and this is genuinely happy in real time. Like, it cooks.

And then if I come down, I can compare all of the three buckets of titles that it created. So these ones were influenced by me. It tells me how many characters, which is super important because you want 55 characters or less with the title.

And then we have titles that are influenced by competitors. This one's pretty good. I trained an AI on 2,000 Alex from Rosie coaching videos.

That's not bad. And you can see, like, there's little copy buttons, I can copy that and boom, it's copied to my clipboard. But I actually want this one.

I wanna copy this one because I like it. And then this section are titles that are influenced by just the principles of building titles properly.

I trained an AI to coach me like Alex Hormozi. My free AI coach is trained on 2,000 Hormozi videos. That's pretty dope, actually.

I kinda like that. So I'm gonna come back over to YouTube Studio. I'm gonna click on a b testing and go to title only.

So right here, I have my main one. I really like the one that it gave me. I trained Claude on 2,000 Alex Hermosa videos.

Like, that's super clean. I have some space to work with, so I could put, like, a plus sign here, 2,000 plus Alex Hermosa videos, or I could say on over 2,000. Let's just keep it 2,000.

I think that's clean. And then coming back to the dashboard, the third one that I liked, I built my own Alex Hormozi AI.

You can steal it. That's pretty good. That is a 99 from vidIQ.

That's pretty crazy. I'm gonna go with these three. Two of them came from the actual skill.

One of them is mine, and I'm feeling like mine is the worst one, to be completely honest. But we will see the test is gonna tell us the truth. So I can just click set test, and now I have my three titles testing.

I'm gonna scroll down here. I already have my template for my description, but I just like to write a quick sentence or two here. So I just gave a brief description of the video.

I'm just gonna scroll down. I do have my thumbnail already created, so I will upload that there. Come down to my playlist, add it to the only playlist that I have, not made for kids.

I did not use AI generated video or photo. And then we have the tags section, which I keep completely blank for this reason.

Because if I come back over to my dashboard and I go to the tags page right here, I can literally just go copy the entire tags that it put together and have all the vidIQ scores on every single tag that it picked. Also, this fits within the 500 tag range, which you need for YouTube, and all of them are only one to three words with commas in between, so they just pick up immediately when I drop them right in, and you can see it works super well.

I'm also gonna add maybe my name and my channel name so it knows that I made it. Now in order to upload the SRT file, I do have to come down to language and captions. Go to language and choose English.

I do have monetization, so I will plug in my ads. And you wanna place ads if you are monetized, like, every three minutes.

So I have three ads placed. I'll hit continue. Now on the video elements page, this is where we actually add the subtitles SRT files.

So I'll go to add. I'll go to upload file with timing, hit continue. And then I'm gonna go to where everything's saved, is just in this folder.

I open it, and now I have all of the captions with time stamps, with perfect grammar, perfect spelling, context of what the video is about, just absolutely dialed. Hit done, add my end screen and my cards and all that stuff, and my video is ready to be uploaded super dialed, very professional, and AI just did all of the heavy lifting for the annoying things while also potentially boosting my CTR.

Like, the titles that it created were genuinely, in my opinion, better than the titles that I created because they were just sourced off of more data, and they also follow the same rules that I'm gonna follow anyway. So like I promised at the start of this video, if you want this skill completely for free, it's yours. All you have to do is jump over to my free school community at school.com/vic.

Come over to my classroom section here. Go to AI automations. Find the video that you're watching right here out of all the other videos, and then go to the skill file right at the bottom under resources.

If you are a creator, I also recommend you watching this build a content empire using AI course. It's completely free.

I talk about why you probably shouldn't make AI avatar content, and I also give you the full formula of how creators actually make millions online. I hope you got value from today's video. If you did, make sure you subscribe and hit the notification bell.

My name is Vic, and as always, stay here now.

The Hook

The bait, then the rug-pull.

Vic opens on the moment every creator skips past: the export is done, the file is dragged into YouTube Studio, and now come the parts nobody enjoys, title variations, clean captions, and fresh tags. He built a Claude skill to do all three from nothing but a file path, and before explaining how it works, he shows the real result: the AI-written title beat his own in a live A/B test.

Frameworks

Named ideas worth stealing.

03:37list

Three-source title engine

Competitor titles (similar top-performing videos)
Channel top performers (your own historical top titles)
Rule-based title principles (programmed heuristics)

Five title options are generated from each of three sources, giving fifteen total testable titles instead of one guess, with the skill flagging its top pick for an A/B test.

Steal forAny headline/title generator workflow where you want testable variants instead of a single guess

04:34list

Two-tier tag sweep

Broad tags: general industry, niche, adjacent topics
Niche tags: specific problems/solutions and exact search terms from the transcript

Tags are researched in two passes through VidIQ so the video is tagged for both its general category and its specific, searchable angle.

Steal forSEO tagging on any video platform that supports tags

CTA Breakdown

How they asked for the click.

VERBAL ASK

10:59link

“if you want this skill completely for free, it's yours... jump over to my free school community at school.com/vic”

Soft-gates a genuinely free download behind a free community join rather than an email form, framed as generosity ('completely for free') rather than a hard sell.

MENTIONED ON CAMERA

02:52toolVidIQ ↗

10:59linkFree Skool community (skill download) ↗

01:30linkContent Solutions call booking ↗

FROM THE DESCRIPTION

PRIMARY CTAWhere the creator wants you to go next.

OTHER LINKSAlso linked in the description.

Storyboard