← Both skills · wildhorn-static-ads · 1-understand/template-reading.md
A template is a proven ad (usually a competitor's or another niche's winner) that we will clone onto our product. Cloning fails when the template is not UNDERSTOOD first. This file is the reading method. It was built from nine founder review rounds on a real brand (Wildhorn bison organ capsules vs ~79 + 80 templates): run 1 validated 0 / 79; after the decode rules below, the next batch validated 6 / 10 on the first try, then 26 / 34. Every rule here exists because a specific ad failed without it.
Sources: /root/workspace/wh-remix/LESSONS.md (A = decode, B = copy, C = visual, D = production,
E = what worked), /srv/blueprint/protocols.html (Protocol 0–4 + every review round), the blueprint
rules R1–R14 (/srv/blueprint/index.html).
The reading has five layers. Do them in order and write each down; the generator brief is assembled from these notes, never from a glance.
1. WHY IT WINS the scroll-stop trick, in one sentence
2. WHAT IT SAYS angle carried, copy formula, text inventory, hidden reading
3. WHERE / WHAT scene logic, props, people, body part, gesture, action (the verb)
4. HOW IT LOOKS visual trick, shape equivalents, render style, shock level, text treatment
5. HOW IT WAS MADE production method (photo + paste, illustration, screenshot, 3D, painted text)
└──► KEEP / SWAP inventory, zone by zone, each with a reason
Write the scroll-stop trick in ONE sentence before anything else. If you can't, you don't understand the ad yet.
Then ask: is the trick visual (the image alone stops), verbal (the line stops), or relational (the gap between image and text — the joke, the innuendo, the contradiction)? Relational tricks are the easiest to break in a clone, because swapping either side kills the gap.
Before changing any line, read it as our buyer would. If it already works for him, copy it word for word. Evidence: S14 "Lost my waistline — last seen 3 weeks ago" was read as negative (bloat) and flipped; for our customer it already meant "I got slim". Flipping a line that already works = fail.
Identify the slots of their sentence and keep the slots, refill the values:
Count every text element: headline, sub-line, bubbles, labels, badges, stat lines, signature, footnote. Ours has the same count, the same hierarchy, the same spacing. Nothing added — no context line, no tagline, no disclaimer, no "tie-in" sentence, no deadline the template doesn't have (B6: "in 21 days" added to #65 was wrong). If the template is a signed review, keep a first name + initial; never add a signature it doesn't have (B12).
Decode innuendo, double meaning and social subtext on EVERY template:
Ask why the scene is where it is:
Every person shown as the customer gets the avatar's age and look (45–60, greying for Wildhorn), sexual scenes included (#79: young man = fail). Remove the template's person and every leftover of their props (cowboy holding THEIR bottle; sequin skirt on our Ken).
Their body part = their problem. Draw OURS (#2: an arm for a GLP-1 muscle-loss template while we sell the belly = off-angle). The image must SHOW the avatar's problem or result (#6: keep the format, rewrite the scene until the belly is visible — "lean man beats beer-belly man").
Copy exactly how the product is held and where it sits: fingers wrapped around it (not floating on a palm), on the lounger, in front of the face. Check hands at full resolution (S33: phone badly held).
The ACTION is part of the winner. If the template POURS the product into the drink he already has, ours pours the organ powder into his beer (S18 — capsules placed next to the beer lost the verb). Other verbs to look for: hiding, measuring, comparing, opening, falling into, melting, guarding.
If the woman's face IS the product, our jar IS her face, in the same painted style (S11 failed three times as "photoreal with a jar on her head"). If the billboard is MADE OF the product, ours is made of our product AND made clear with labels (#35: "keep what made the template a winner AND make it clear — never trade one for the other"). If the product pours into a container, ours pours with our literal content in recognizable form (#39 organs falling into the jar with tags: "exactly what I was thinking").
Find an object with the same shape or function as OUR problem so the template's visual pun survives intact:
Match it. If the template is gross or shocking (worms coming out of a belly), ours is equally shocking for our problem (belly melting like wax, bursting) — never a tame substitute (S7: a popping button was not as shocking as worms). Aggression = precision on the deepest pain (#47 "Leave the lights on": the bedroom shame, "the greatest").
Pixar stays Pixar, sketch stays sketch, watercolour stays watercolour (S62 validated as-is with the angle swapped). Follow the template's visual code on every panel, result included (#16 lost the orange glow on the result). Special text treatments (sunscreen bubbles, clay letters, handwriting, text painted on skin) are reproduced, never replaced by a flat font (#65). Product rendering quality = the template's: real 3D object, contact shadow, matching light.
The format itself should reveal what the product is made of in one second (a butcher-chart animal with its organs; organs visible inside a transparent capsule). Never a generic capsule. Our content must look premium and unambiguous: raw chunks look disgusting, a powder pile reads as dirt, brown dried organs looked like dung (S17). If the template shows its ingredient in a certain rendering, copy that rendering.
Before re-prompting a failed ad, work out HOW the template was produced and use the same method. S11 was a real photo with the product pasted on as the woman's head (a simple composite); three rounds were lost regenerating the whole scene. The fix was: edit the template itself minimally (age the man, clear the text), swap only their jar for ours with a masked edit, add the text by script, keep the 9:16 aspect ratio.
Classify every template:
| Method | Tell-tale signs | Clone with |
|---|---|---|
| Real photo + pasted product | Product lighting/edges differ from the scene | Minimal edit of the template + masked product swap + scripted text |
| Full illustration / cartoon / 3D | Consistent drawn style | Full regeneration in the same style |
| Screenshot / native UI (iPhone note, tweet, DM, Reddit) | System fonts, UI chrome | Script-rendered text on a clean UI recreation (zero text errors) |
| Text painted into the scene (sunscreen, clay, skin) | Letters follow surface/lighting | An engine that paints text well, or composite with care |
| Pure text / letter | No image | Script-rendered text (S4 "text-only via script = zero text errors") |
| Collage / comparison table | Grid, ticks/crosses | Recreate the grid; product photos composited |
| Sexual / skin scene | Lingerie, bodies | A permissive engine for the scene only, product + text composited |
Note it in the brief as production_method. 2-make/ picks engines (2-make/engines.md); the analysis only names
the method and flags whether the image model is likely to refuse (then: switch engine + composite,
never water the ad down — D1).
Before any prompt, inventory EVERY element of the template, zone by zone, and mark each KEEP or SWAP → our equivalent, with the reason. The brief is written from this inventory. "You could have done this from the first try."
Template:
Template: [id] · family: [concept family] · aspect: [1:1 / 4:5 / 9:16] · method: [production method]
Scroll-stop trick: [one sentence]
Angle carried → our angle: [A_] · mechanism (from map): [M_]
Hidden reading: [innuendo / joke / social mechanism / none] → ours: [...]
Shock level: [calm / cheeky / gross / shocking] → match
Text slots: [n]
| Zone | Element in template | KEEP / SWAP | Our version | Reason |
|------|---------------------|-------------|-------------|--------|
| Setting | kid's bedroom | SWAP | living room, game night | scene logic: their product = kids' sleep |
| Prop | bunk beds | SWAP | beer cans, pizza box | props of OUR problem |
| Person | tired mom, 35 | SWAP | tired dad, 50, greying | avatar age everywhere |
| Gesture | holds jar at chest, fingers wrapped | KEEP | same, our jar | placement + gesture |
| Light / camera | warm lamp, eye level | KEEP | same | render style |
| Headline | "Finally sleeping through the night" | SWAP | "[bank line]" | angle swap, same length |
| Sub-line | "…" | KEEP | word for word | already works for our buyer |
| Badge | "Doctor approved" | SWAP | "60-day guarantee" | no fake authority; our real proof |
| Product spot | bottom right, 20 % width | KEEP | our jar, same spot/size | placement |
| Logo slot | small, top left | SWAP | our icon only | brand slot |
Rules for the inventory:
| Case | What happened | Lesson |
|---|---|---|
| #6 Abs vs beer belly | Kept the format, rhythm and bottom line; changed the scene so the image shows our problem | Keep the format, rewrite the scene until the problem is visible |
| #21 "The beer belly is not the problem" | Kept the exact sentence structure, swapped symptom → beer belly and cause → estrogen | Same structure + noun swap |
| #8 Hide-the-chest routine | Turned a generic routine into the one he has BECAUSE of his problem | Routine and angle are the same thing |
| #23 "You're not fat" | Every block of the template kept, each filled with one sharp research idea in his words | Block-by-block fill |
| #26 Measuring tape | Their product symbol (hair ribbon) → our result symbol (tape) | Replace every symbol of their product |
| #31 Keep your beer | Their GLP-1 (what he won't give up) → beer | Find the avatar's equivalent of their reference |
| #36 Capsules flanking the jar | Their content (brisket) shown the same way, with ours | Same display, our literal content |
| #37 Labelled organs + live bison | Mechanism of the image changed (no milking) but the idea stayed: from the source into the jar | Same idea, our product's own mechanism |
| #40 Organs inside the capsule | "Way bigger improvement" | Make the invisible ingredient visible |
| #44 Old jeans fit again | "Even better than the example" | Proof object he's proud to see |
| #47 Leave the lights on | "Most aggressive — the greatest" | Aggression = precision on the deepest pain |
| #58 Fat bottle | Missed their visual mechanism | Same angle → copy the visual mechanism too |
| #63 Step 1 / Step 2 | Kept both lines word for word except the subject + the social mechanism | Same angle → keep their words |
| #65 Sunscreen letters | Invented a deadline, flat font, jar off the lounger | No added deadline; copy text treatment; copy placement |
| #70 Shower | Shower copied for no reason | Scene logic |
| #73 Eggplant | Text not centred on the eggplant's axis | Text centred on the template's anchor; same spacing |
| #75 Kids' beds | Template furniture kept | Inventory every pixel; props of our problem |
| #77 Split body | Two different bodies | One body cut exactly in half |
| #79 Review | Young man, jar not on the face, no name | Avatar age everywhere; copy gesture; keep first name on review formats |
| S1 Fewer inches | Growing bananas → shrinking belts | Flip their object for our result; the joke stays |
| S3 Puzzle jar | Product opened into ingredient pieces, filled with recognizable organs + one job each | Keep their mechanism, our recognizable content |
| S7 Worms | Button popping not as shocking | Match the shock level |
| S9 "You can't farm this" | Unclear to a stranger | Stranger test |
| S11 Face = product | Photoreal head swap ×3 failed; then a minimal composite worked | Copy the visual trick literally + the production method |
| S14 Waistline | Flipped a line that already worked | Read the line as our customer |
| S15 Shoelaces | Indirect consequence | Move only to where the problem is directly visible |
| S17 Brown organs | Over-correction looked like dung | Fresh, premium, butcher-counter organs |
| S18 Pour into beer | Capsules beside the beer | Copy the verb |
| S23 Moon | Round moon ↔ round belly | Shape equivalents first |
| S28 Arcade | Whole UI kept + one stat row added | Fuse without breaking the format |
| S50 Pear | Sexual fruit instead of belly metaphor | Objects map to the belly |
| S74 Before/after | Two lean bellies | Before/after stays before/after |
| Round 2 bulk (S12, S16, S20–S22, S27, S32, S36–S38) | 26/34 validated first try | Exact structure + one-word angle swap + our proof props |
The 79-template Wildhorn batch sorted into 9 families; reuse the list for any brand:
Assignment rules (R1–R4, R8, R11): accept every template; sort into families first; one ad per template (volume = number of templates); spread the top angles inside each family so every family covers every angle — but each template still gets the angle it naturally carries; never a group outside the top angles.
TEMPLATE [id] family [n] aspect [..] source [brand · days live · copies]
1 Trick (1 sentence):
2 Angle carried → ours (+ mechanism from map):
3 Read-as-our-customer lines kept verbatim:
4 Copy formula (slots → our values):
5 Text slots (count, hierarchy, treatment):
6 Hidden reading → ours (same intensity):
7 Scene logic → our scene (directly visible problem):
8 People → our avatar; body part → ours:
9 Gesture / placement / verb of the scene:
10 Visual trick + shape/function equivalents:
11 Structure type (before/after, split, progression, table, UI):
12 Shock level / render style / special text treatment:
13 Production method (+ refusal risk):
14 KEEP/SWAP inventory (table):
15 Compliance flags:
16 Missing facts ([MISSING]):