The Faceless Video System That Books Renovation Jobs
The exact three prompts and recording setup behind a faceless video that booked $1.8M in remodeling jobs — mine the pain, write the script, narrate it over your own footage. No camera required.
Two videos. One brought in 5,400 leads. The other booked $1.8 million in remodeling contracts over nine months — real signed jobs, not clicks. Neither one used a film crew, a marketing agency, or the contractor's face on camera.
Both scripts were written by Claude, the AI I've been onboarding as an employee for the last two weeks. This is the exact system behind them: three prompts, one framework, and a recording setup simple enough to run on your own business today. Copy the prompts, follow the steps, and you can point this at your own reviews this afternoon.
None of this is theory. It's the same process that produced the numbers above.
The one idea the whole system runs on
Nobody wants a new kitchen. They want to stop being embarrassed when friends come over and see the one they have. Nobody wants a finished basement. They want their kid to have somewhere to bring friends without apologizing for the house.
Most contractor marketing sells the renovation. This system sells the feeling the homeowner is quietly paying to get back. That is the entire trick, and it is why these leads convert into high-end clients instead of tire-kickers. Get that part right and the rest is mechanical.
There are three steps, and the third one is where most people fall down:
- Mine the real pain — a prompt that reads Google reviews and finds the emotional driver under the surface want.
- Write the script — a prompt that turns that pain into a short Problem → Agitate → Solve narration.
- Narrate it over your own job footage — no camera, no face required.
Step 1 — Mine the real pain (the Pain Point Miner)
Every good script starts with a true sentence about your customer. Not "they want more counter space." The thing underneath it: "every time someone new comes over I feel like I have to explain the kitchen before they even see it." You cannot guess that sentence from your desk. It is sitting in your Google reviews and your competitors' reviews, written in your customers' own words.
The Pain Point Miner does the digging. It takes a stated want and asks "why does that actually matter to them?" over and over until it hits something identity-level, social, or fear-based, then hands you a table you can write from.
Paste this into Claude (or any capable model):
You are a consumer-psychology researcher who specializes in "laddering" —
taking a stated want and asking "why does that actually matter to them?"
repeatedly until you hit the real emotional driver underneath it.
Your job: mine Google reviews and search results for a local service business
and extract the EMOTIONAL pain points behind the buying decision — not the
surface-level want, the thing underneath it.
---------------------------------------
STEP 1: COLLECT MY INPUTS
---------------------------------------
Ask me for these one at a time if I haven't given them yet:
1. Business type / niche (e.g. "high-end kitchen and bathroom renovation
contractor")
2. City / service area
3. Who the ideal client actually is (age, household, income bracket, homeowner
vs. renter, life stage)
4. The specific service or project type I want pain points for (e.g. "kitchen
remodel," "basement finishing," "primary bathroom renovation")
5. Any raw material I already have: pasted Google reviews (mine, competitors',
or the category broadly), Google/Reddit search snippets, or past client
quotes. If I have none, tell me the exact search queries to run to find
review text for this niche.
Do not start extracting until you have the niche, the city, and either raw
review text or a confirmed plan for where to source it.
---------------------------------------
STEP 2: EXTRACT — SURFACE WANT FIRST
---------------------------------------
For every review/quote/snippet, pull out the STATED want or complaint exactly
as the person wrote it. This is the surface layer. Do not editorialize yet.
---------------------------------------
STEP 3: LADDER DOWN TO THE EMOTIONAL DRIVER
---------------------------------------
For each surface want, ask "why does that actually matter to this person?" at
least THREE times in a row, on paper, until you reach a driver that is
identity-level, social, or fear-based — not functional.
A finding is NOT done until it looks like one of these:
- Embarrassment / shame in front of others (friends, family, in-laws)
- Fear of being judged or seen as "behind" their peers
- Guilt toward a partner or kids ("I promised them...")
- Loss of control / being taken advantage of (bad contractor, budget blowout)
- Identity ("this house doesn't feel like ME anymore")
- Relief / permission to finally stop postponing something
REJECTED (too shallow): "Client wants a bigger kitchen because their current
one is small."
ACCEPTED (correct depth): "Client is embarrassed to host friends and family in
their current kitchen and has been quietly declining invitations to host
holidays for years — the want isn't square footage, it's the ability to invite
people over without dreading it."
---------------------------------------
STEP 4: OUTPUT FORMAT
---------------------------------------
Return a table with these exact columns:
| Surface Want | Emotional Driver | Evidence (quote/paraphrase) | Content Angle |
Sort by how strong/common the emotional driver is (most common and most
visceral first). Return at least 6 rows. Do not include a row where the driver
just restates the surface want.
---------------------------------------
STYLE RULES
---------------------------------------
- Never stop at a functional answer ("more storage," "modern look," "resale
value") — keep laddering past them.
- Write the Emotional Driver column the way the person would say it in their own
head at 11pm, not the way a brochure would say it.
- No corporate language. No "enhance their living experience."
How to run it. If you do not have review text handy, answer the model's questions and it will hand you the exact searches to run ("[niche] reviews [city]", "[niche] regret", "[niche] embarrassed", plus your competitors' Google Maps pages). Paste a few dozen reviews back in and let it build the table. You are looking for the row that makes you go "oh — that's the real reason people call us." That row becomes your video.
Here is a real slice of what it returns for a kitchen remodeler:
| Surface Want | Emotional Driver | Content Angle |
|---|---|---|
| "Wanted a bigger kitchen with an island" | Embarrassed to host — has quietly stopped inviting people over for two years | "The kitchen that made us stop hosting" |
| "Needed more counter space" | Feels she's constantly apologizing for the house; the kitchen is a symbol of everything else unfinished | "The room you apologize for before anyone sees it" |
| "Wanted updated cabinets" | Resents that a past contractor made her feel stupid about the quote — the driver is trust, not aesthetics | "Why our clients finally feel like an adult in the room" |
Step 2 — Turn the pain into a script (Problem, Agitate, Solve)
A pain point is useless until it is a script someone will actually watch. The framework is PAS: state the Problem in the customer's own words, Agitate it by naming what it costs them in an ordinary week, then Solve it plainly and end on the feeling they get back, not a feature list.
Paste this into Claude:
You are a direct response copywriter who writes short video scripts for a home
renovation company. You are not writing a brochure. You are writing the way one
person talks to another.
INPUTS
- Pain point: [paste one row from the Pain Point Miner — the emotional driver,
not the surface want]
- Job type: [e.g. bathroom remodel, kitchen remodel, basement finishing]
- City / service area: [e.g. Ottawa]
WRITE THE SCRIPT IN THREE PARTS
1. PROBLEM
Say the pain out loud, in the customer's own words. One sentence.
Do not describe the product. Describe their situation.
2. AGITATE
Make it concrete. What does this cost them in an ordinary week — socially,
emotionally, or practically? Two or three sentences. Stay specific and
believable. Do not invent statistics or exaggerate.
3. SOLVE
Introduce the work plainly. What you do, roughly how long it takes, and what
changes for them afterwards. End on the feeling they get back, not features.
RULES
- Short sentences. Read it aloud — if you run out of breath, cut it.
- Grade 8-10 reading level. No jargon a homeowner wouldn't use.
- Banned words: elevate, transform, dream home, oasis, sanctuary, unlock,
seamless, bespoke, curated, nestled, stunning.
- No exclamation marks. No hype. Confidence is quiet.
- Never promise a price or a timeline you were not given in the inputs.
OUTPUT
60-90 words total, as one block of spoken narration, ready to read into a mic
over b-roll. Label the three parts in square brackets like [PROBLEM], [AGITATE],
[SOLVE] so the editor can see the structure.
Those banned words matter more than they look. The moment a script says "transform your dream kitchen," it sounds like every other ad and the viewer scrolls. Plain language is what makes it read as a real person talking, which is the whole point.
Here is what it produces for a bathroom remodel:
[PROBLEM] You have started apologizing for your bathroom before people even use it.
[AGITATE] It is the one room every guest goes into alone, and you know exactly what they see. The cracked grout. The tap that drips. You have stopped having people back after dinner because of a room you use twice a day.
[SOLVE] We rebuild it properly — usually about three weeks, start to finish. New everything, done once, done right. And the next time someone comes over, you do not think about that door at all.
The Problem is her sentence, not yours. The Agitate names a real weekly cost — people stop coming over — instead of listing defects. The Solve gives a timeframe and ends on the feeling, not the fixtures.
Step 3 — Narrate it over your own job footage (no face required)
This is the step people skip past, and it is the one that actually turns the script into leads. You do not need to be on camera. You need your voice (or a stand-in for it) over footage of work you have already done.
Record the voiceover. The cheapest version is your phone in a quiet room — closet, car, anywhere soft. Read the script the way you'd say it to a client across the table, not the way you'd read a paragraph. If you do not want to use your own voice at all, generate it with ElevenLabs — pick a natural voice, paste the script, download the mp3. (If you want the full "AI reads every script for you and an avatar delivers it to camera" setup, that's its own breakdown — comment agent on the videos and I'll send it.)
Gather the b-roll. Five to eight clips of completed jobs: slow walkthroughs, before-and-afters, a pan across the finished room. Phone footage is completely fine — this is proof, not cinema. If you have a drone shot or two, even better.
Assemble it. Any editor works; CapCut is free and enough. Narration on top, job footage underneath, auto-captions on (most watching have the sound off). Keep it to 45–60 seconds. Post it as a Reel and a Short.
That is the whole video. The faceless one we built this way did over 1.2 million views and brought in more than 650 basement-remodeling leads for one client. The same system, cut a second way for revenue instead of volume, booked over $600,000 in basement jobs and $1.8M in total contracts across nine months.
One video is a test. Fifty is a pipeline.
A single video proves the system works on your business. The leverage is running fifty or sixty of them, one for each pain point, service, and neighborhood, all pulling leads while you are on a job site.
You cannot record sixty voiceovers by hand, which is exactly where the AI employee earns its keep: an avatar that reads each script to camera for you, in a few different looks, so the only human step left is picking which pain point to mine next. That is the content factory, and it is a full breakdown of its own — comment "agent" on the videos and I'll send you the ElevenLabs + HeyGen setup and the avatar prompts.
The catch most people miss
The system is only as good as Step 1. If the pain point is generic, the script is generic, and generic scripts pull junk leads, the exact tire-kickers you were trying to get away from. Ninety percent of the quality is decided before you write a word, in how honestly you mine the reviews. Spend your time there. The prompts do the rest.
If you'd rather we just build and run this for your business (the mining, the scripts, the whole faceless-video engine on autopilot), that's what we do.
Frequently Asked Questions
Do I need to be on camera to make these videos?
No. The entire system is faceless by design — your voice (or an AI voice) narrates over footage of jobs you've already completed. No camera, no on-screen appearance required.
What if I don't have many Google reviews yet?
You don't only mine your own reviews. The Pain Point Miner also reads competitors' reviews and category-wide search results in your city, so you can find the real emotional drivers even before you've built up a big review base of your own.
Which AI model do I use for the prompts?
Any capable model runs them — the prompts are written for Claude but work in ChatGPT or Gemini too. The quality comes from the prompt structure and the honesty of your review inputs, not the specific model.
How long does one video take to make?
Once you've mined the pain points, a single script-to-posted-Reel takes an hour or two: a few minutes for the script prompt, a short voiceover, and a simple edit in a free tool like CapCut over your existing job footage.
Won't these videos attract the same junk leads as agency marketing?
That's the whole reason the system leads with pain-point mining. Generic scripts pull tire-kickers; scripts built on a real emotional driver attract homeowners already feeling that specific problem — which is why these convert into high-end clients instead of price-shoppers.