The input · 2 min 20 sec
Two minutes
of
rambling.
No script. No prep. He hit record and talked about compliance.
The output · 1 min 43 sec
A finished
video.
What came back
No camera.
No editor.
No video software.
One minute forty-three of finished, branded, narrated video — built from a voice memo and a folder of text files.
One voice memo in. A finished video out. Here's every piece of how that happens.
The raw material
It didn't rewrite him.
It
cleaned him up.
What Amir actually said
"…if you are not following
compliancy
rules you have to make sure that all the leads are being opted in. All the leads need to opt in…
and that's all I got for you. So if you have any questions, just reach out.
"
What went into the video
"…if you're not following
compliance
rules, you can get yourself into trouble. The most important rule is that
every lead has to opt in
."
Grammar tightened, the doubled-back sentence untangled, the sign-off cut. Every warning he gave stayed exactly as strong as he made it.
The brand kit
This is why it looks like
your company
and not a template.
#011528
#033362
#057EEF
#4CA6FC
#7B102C
Anton
Inter
JetBrains Mono
One file. Written once. Every video after this one inherits it for free.
The voice
Two script files.
One word
different.
02-compliance.txt · for humans
…dropping aged
leads
into a dialer…
02-compliance.tts.txt · for the robot
…dropping aged
leeds
into a dialer…
Text-to-speech reads "lead" like the metal. Every time. So the robot gets its own spelling — and the captions on screen still read the real word.
The timing
It listens to
its own voice.
The AI voice gets transcribed back — so the machine knows the exact millisecond every word is spoken.
The
8.64s
most
9.20s
important
9.50s
rule
9.72s
is
10.02s
that
10.20s
every
10.44s
lead
10.74s
Slides aren't pinned to a clock. They're pinned to
words
. Change the script, re-run it, and the whole video re-times itself.
The render
It can't drift
out of sync.
It isn't recording a screen. It's computing what the screen looks like at every single moment, then gluing those moments to the voice.
Slow computer, fast computer — doesn't matter. Output time always equals source time.
page
The video is a web page
index.html
seek
Jump it to an exact moment
no playback, no recording
shot
Screenshot that moment
one jpeg per frame
mix
Glue frames to the audio
voice + music bed + sfx
mp4
Finished file
1:43
The part that actually matters
The thing you build
is
the next video.
Compliance · 1:43
Intro · 1:05
Scripts & Cadence · 2:07
The first one was hours of real work. The second and third were new scripts into the same machine.
Where it doesn't work
This doesn't replace
you.
Not a talking head.
If people need to trust a person, put a person on camera.
Not a software demo.
Sometimes they need to watch the actual clicks happen.
The first build is real work.
The leverage shows up on video two.
What it's great at is the explainer in the middle — the one nobody ever gets around to making.
Go take it.
The prompts, the folder structure, the tools, the order to do it in.
Free build pack · link in the description
