Dom | Trykrust.com

I make pixels sell | 100k MRR
Partnerships |

[Dom | Trykrust.com @heardofvinu]

my hook testing loop:

1. write 20 hooks
2. generate the first 3 seconds only
3. kill the weak ones
4. build full ads from the survivors

never render a full ad for a hook you haven't tested.

[Dom | Trykrust.com @heardofvinu]

before you kill an ad, read it like this:

low hook rate: the first second is the problem. new hooks.
high hook, low hold: they stop, then leave. fix the middle.
high hold, no buys: the ad works. your offer or page doesn't.

most people kill the third one.

[Dom | Trykrust.com @heardofvinu]

@AnovaCEO conversions over ads is the right unit, I was counting spend like an idiot. so I'm guessing a floor now: ~30 conversions per variant before I call it, and anything under that gets a retest. do you hold a hard floor or go by confidence?

[Dom | Trykrust.com @heardofvinu]

AI ad lifecycle:

monday: "we're going all in on AI creative"
tuesday: 30 ads that look identical
friday: "AI doesn't work for us"

it worked fine. nobody wrote a hook.

[Dom | Trykrust.com @heardofvinu]

hot take: the best AI ad makers right now aren't the ones with the best tools.

they're the ones who watched the most ads before they wrote a single prompt.

taste is the moat.

[Dom | Trykrust.com @heardofvinu]

"this ad would've cost $50k" posts are getting silly.

nobody was ever paying $50k for that ad.

the real math is boring: a creator video is $100-300. a 5 sec seedance 2.5 clip is about $2 on my stack.

the win was never one cheap ad. it's 40 tests instead of 4.

[Dom | Trykrust.com @heardofvinu]

AI UGC bingo.

if your ad hits 3 in a row, the viewer already scrolled.

how many is yours on?

[Dom | Trykrust.com @heardofvinu]

@AnovaCEO fair, one at a time is the clean read. the catch is it burns budget fast, so i test hook first since it moves the thumb-stop rate most, then lock the winner and move to proof.

how many ads per variant before you call it? i've been burned calling it too early.

[Dom | Trykrust.com @heardofvinu]

@scaledgabo yep. easiest way to check: pull 20 reviews and read the exact phrases people use before they bought, not after. if your hook doesn't sound like one of those lines, it's competitor phrasing in disguise. where do you source yours, reviews or calls?

[Dom | Trykrust.com @heardofvinu]

screw it. giving the whole thing away.

our entire AI UGC ad playbook. every step from a product link to an ad that actually sells. it's how we make ads at krust (100k MRR, yes i built it)

the first 4 parts happen before a single ad gets generated.

here's what's inside:
1) the product truth. what it really fixes, in your buyers' words, not yours
2) the research. a review language bank and a teardown of every competitor's ads
3) the angles. angle bank, awareness levels, saturation check
4) the avatar. casting the AI face your buyer trusts
5) the format. which UGC format fits which product, and when statics win
6) the hooks. a hook library and the first-3-seconds check
7) the script. shapes that sell, plus an on-label check so claims don't get you flagged
8) production. product link to finished ad, remaking a winner, killing the AI sheen
9) the cost. AI ad vs creator video, real numbers
10) testing. how many variants, what to change first
11) the numbers. what's a winner, what to iterate, a learnings file per brand

every part comes with a claude prompt you can run on your own brand. krust (my tool) runs inside claude too, so the whole thing can live in one chat.

ps. a better hook can't save a bad angle. that's why hooks only show up in part 6.

it's free. reply "playbook" and i'll send it 👇

[Dom | Trykrust.com @heardofvinu]

this is what happens when an agency discovers a new AI video tool:

week 1: "this changes everything"
week 2: 40 identical ads
week 3: "why isn't it converting"

it's the hook. it's always the hook.

[Dom | Trykrust.com @heardofvinu]

for anyone tired of waiting on creators to send footage.

I work with Krust. we make AI UGC ads you can actually test fast.

https://t.co/ao2JXI90OE

[Dom | Trykrust.com @heardofvinu]

@SulcasJuras This is insane

[Dom | Trykrust.com @heardofvinu]

most AI UGC looks like it was made by someone who has never sold anything.

perfect skin, perfect lighting, zero reason to keep watching.

ugly and specific beats pretty and generic.

[Dom | Trykrust.com @heardofvinu]

the prompt change that fixes most uncanny AI faces:

stop asking for "realistic". describe the flaws. pores, uneven skin, a bit of shine, light from one window.

perfect is what makes it creepy.

[Dom | Trykrust.com @heardofvinu]

if your AI UGC still has the AI sheen and floating hands, you're not iterating hard enough.

the tool isn't the problem. v1 never is.

[Dom | Trykrust.com @heardofvinu]

most underrated AI ad format right now: the history explainer.

old fix everyone loved → the problem nobody mentioned → the new way → your product → proof

people learn something before they get sold anything. AI makes the b-roll cheap, so the real work is research.

[Dom | Trykrust.com @heardofvinu]

honest numbers, no "$0 ad" nonsense:

creator UGC video: $100-300, days to weeks
AI UGC ad in krust (my tool): about $6-12, minutes

40 tests a month: $4,000-12,000 vs about $240-480.

creators still win the hero ads. AI wins the volume.

[Dom | Trykrust.com @heardofvinu]

6. most AI ads die before the first generation. if you can't say who it's for and why they'd stop scrolling, no model saves it.

which one do you disagree with?

[Dom | Trykrust.com @heardofvinu]

a few thoughts on AI ads after a month of watching way too many of them:

1. the model isn't your edge anymore. everyone has seedance and kling. the edge is the script and the first 2 seconds.

[Dom | Trykrust.com @heardofvinu]

5. editors who learn to direct agents get more valuable, not less. taste doesn't automate. timelines do.