We didn’t “crack” the account with a new targeting trick.

We cracked it with one winning hook.

Simple story. Clear promise. Stupidly high CTR.

Then we ran that hook through our AI stack.

Suddenly, it wasn’t “one ad” anymore. It was:

• 15+ creative variations

• 4 aspect ratios (1:1, 9:16, 4:5, 16:9)

• Multiple tones, creators, and openings

• All built on the same core idea

That’s when it clicked:

The Old Playbook Is Broken

Here’s how most teams still work:

Creative team makes an ad. It runs for two weeks. Performance dips. Everyone scrambles to make “the next ad.”

Rinse. Repeat. Burnout.

The problem? On saturated channels like Meta, ads are a volume game.

The algorithm doesn’t reward your one beautiful ad. It rewards whoever can feed it enough fresh creative to find cheap pockets of attention.

One great hook isn’t the finish line. It’s the starting point.

One hook = a whole creative tree. 🌳

Your job isn’t to find a miracle ad. Your job is to find a repeatable hook and teach AI to spin it into a small army of creatives.

Here my (agile) workflow

1 Start With the Hook That Deserves 15 Lives

In crowded feeds, you don’t have room for “clever.”

You need a hook that is brutally obvious:

Problem → “This is annoying, right?” Shift → “Here’s the new way.” Payoff → “Here’s what you get.”

Think things like:

• “Turn any bad photo into a perfect profile pic.”

• “Fix your lighting in one tap—no studio, no gear.”

• “Erase anything from your photos in 3 seconds.”

Once one of these hooks proves it can pull attention and clicks, that’s the seed.

You don’t just duplicate the ad and tweak colors. You multiply the hook.

100 Hooks for Ecom

I’m going to show you how to take one video asset and multiply it into a library of 15+ high-converting variations using the modern AI stack. We aren’t just changing captions here; we are changing the video reality itself.

2 🧠 The Framework: Models vs. Tools vs. Platforms

Before we get to the tactic, you need to stop confusing the terminology. Most marketers use these words interchangeably. That’s a mistake.

To pull this off, you need to understand the AI Video Stack:

• The Engine (Models): These are the brains. Think Kling or VEO. They are the raw intelligence generating the pixels.

• The Workshop (Tools): This is where you tinker. ComfyUI is the heavyweight here. It gives you infinite control to chain models together, but it requires a “builder” mindset.

• The Factory (Platforms): This is for scale. MakeUGC and ArcadS. These platforms wrap the technology into a user-friendly interface, specifically for things like talking avatars and consistent character generation.

The “Moonshot” Idea: Stop testing text variations. Start testing reality variations.

🧪 The “Hook Library” Strategy

Here is the exact workflow to turn one video into a scalable PPC asset for E-com.

Step 1: Isolate the “Source Code”

Take your winning video. Identify the first 3 seconds (the Hook).

Maybe it’s a UGC creator saying, “Stop using regular soap.”

Step 2: Choose Your Weapon (ComfyUI vs. Platforms)

Option A: The Control Freak (ComfyUI)

If you want granular control over lighting, texture, and scene composition, use ComfyUI with the Kling model.

• The Play: Use “Video-to-Video” transformations. Keep the motion of your winning ad, but swap the environment.

• Result: The same winning hook, but now the creator is on a beach, in a luxury apartment, or in a neon-lit studio.

Option B: The Scale Master (MakeUGC / ArcadS)

If you need talking heads that look human (but aren’t), this is where the money is.

• The Play: Upload your script to MakeUGC or ArcadS. Their tech uses advanced lip-sync and facial modeling (often superior for direct-to-camera speech compared to raw models).

• Result: You generate the same winning script delivered by 5 different “Avatars” (different ages, ethnicities, and genders) in minutes.

Step 3: Build the Matrix

Now, you don’t just have one ad. You have a Hook Library.

💡 Why It Works (Psychology)

This taps into the “Mere Exposure Effect” but bypasses the “Habituation” filter.

1. Familiarity: The core message (the script/hook) remains the same, building brand memory.

2. Novelty: The visual “wrapper” changes completely.

To the human brain (and the Meta algorithm), these look like completely new ads. But they are built on the skeleton of a proven winner.

The Result: You extend the life of a winning concept from 2 weeks to 6 months.

3 Build a Media Library That Remembers What Worked

When you’re pumping out variants, chaos is a feature… until it kills you.

So you need a simple structure that answers:

• Which hooks are proven?

• Which styles won for each hook?

• What died from fatigue and shouldn’t be revived?

Even a basic spreadsheet with these columns works:

This turns your ad account into a library of proven patterns, not random uploads.

When a hook starts fatiguing, you don’t panic. You check your library, see which styles haven’t been tested yet, and spin up the next batch.

The New Question

On Meta and other saturated platforms, the algorithm rewards volume and resilience, not perfection.

So don’t ask, “What’s my next great ad?”

Ask instead:

“Which of your winning hooks is sitting there with only one life? And how fast can you give it 14 more?”

Build a Media Library That Actually Scales

I used to work on movies and ads. I come from production. Today I am a marketer who sits between creative and analytics and runs campaigns, which is a weird spot. At Lightricks, a campaign could go from brief to launch in one or two weeks, so documentation and asset sharing became critical. Since then I have carried this discipline into startups and freelance work. I have also seen teams hire actors and full crews on a hunch, the project collapses, the marketing group is let go, and there is no media library to reuse assets. No documentation, no tags, no tracking of which hook in the first three seconds worked, no trend notes, and no historical data on organic wins. A media library is not only storage. It is how you analyze performance and find shared traits of winning assets.

🎯 What This System Gives You

• Keep Web, Design, Organic, Paid, and PPC on the same backbone

• Never ask “Where is the final file” again

• Make new campaigns a duplicate, not a rebuild

• Backup for (.fig) Figma files, lotties, Rive, etc.
  

⚙️ Step 1: Duplicate Campaigns, Do Not Invent Them

Most teams start from scratch. New folders. New names. New chaos.
Fix it with one master campaign template. Duplicate it for every project.

Why it works

• Year folder keeps the archive browsable at a glance

• Date first naming makes matching creatives to reports trivial

• Chronological sort works in any file system

• Rule: time first, name after. Goodbye “final_final_v3”

Template starter

/MediaLibrary
  /2025
    /2025-12-04-[CampaignName]
      /01_Strategy
      /02_Scripts
      /03_Production
        /Footage
        /Audio
        /Stills
      /04_Edit
        /Projects
          /Premiere
          /AfterEffects
        /Exports_Working
      /05_Design
        /Source
        /Exports_Working
      /06_Analytics
        /Reports
        /Learnings
      /07_Delivery  <- single source of truth
        /Organic
        /Paid
        /PPC
      /08_Backups

Here you can download mine

Quick checklist

• Create the master template once

• Enforce date-first naming everywhere

• Duplicate, do not rebuild

Tip: Never Use Spaces in Folder Names on GitHub

• ❌ Spaces break command-line operations unless escaped (\ or %20)

• ❌ They can cause errors in scripts, automation, and builds

• ❌ Many programming tools assume space-free paths

  • kebab-case → my-project-folder

  • snake_case → my_project_folder

  • camelCase → myProjectFolder

⚙️ Step 2: Let Scripts Do The Boring Work

Even duplicating folders by hand gets old fast. Automate it.

Option 1: Python in VS Code

• Asks for date

• Asks for campaign name

• Generates the full folder tree in the right year

  
  Here you can download my Python file

  I advise further customization with Claude

Option 2: One click Mac app with Automator

• Click New Campaign on the dock

• Prompt for campaign name and date

• Build the tree and create clean Premiere and After Effects files

  Here you can download my Automatot file

Outcome
New campaigns become a one-click ritual, not a mini IT project.

Step 3: Make 07_Delivery The Single Source Of Truth

Team rule: If a file is not in 07_Delivery, it is not live, not approved, and not shareable. No exceptions.

What lives in 07_Delivery

• Final exports for Organic

• Final exports for Paid

• Final exports for PPC

Filename spec

YYYY_MM_DD_APP_GEO_PLATFORM_ASPECT_HookNN_vVV.mp4

Examples

2025_12_04_APP_US_TikTok_9x16_Hook01_v01.mp4
2025_12_04_APP_FR_Meta_4x5_Hook02_v02.mp4

Token guide

TokenMeaningExamplesYYYY_MM_DDProduction date2025_12_04APPProduct or app codeAPP, SHOP, EDUGEOCountry or region codeUS, FR, DE, LATAMPLATFORMPlatform nameTikTok, Meta, YouTubeASPECTAspect ratio9x16, 1x1, 4x5, 16x9HookNNCreative hook identifierHook01, Hook02vVVVersionv01, v02, v10

🧰 Tooling Stack That Actually Gets Used

Pick one cloud as the source of truth

• Google Drive

• Dropbox

• OneDrive

• S3 or a compatible bucket

🚦 Operating Rules That Make This Scale

• One template to duplicate, never rebuild

• Date first naming across folders and files

• 07_Delivery is the only place for shareable finals

• Keep hooks and versions two digits for sort stability

• Record learnings inside 06_Analytics after every campaign

🔎 Why Marketers Should Care

This is not an editor problem. It unlocks speed across the org.

• Analytics pulls last quarter winners in 10 seconds

• Campaign managers see which version actually shipped

• Localizations stay sane with dated, geo specific work

• Reporting ties any asset to a date, campaign, and channel

• AI workflows improve with clean, tagged inputs

🧭 The Point

Not storage. Navigation.