Studio · Reference · style.md · variants
Copy This Ad
One system for copying an ad you have the right to study: decode the reference into a five-aspect analysis and an eight-dimension table, compile a style.md of invariants and machine fields, freeze the body, batch the hook, render deterministically, and gate every claim.
The whole line, one pack
story to stitch
The board below is the core, not a ceiling; extend it until the story lands.
What you get
12 deliverables
- Reference record: right to study, source, date, transcript
- Five-aspect analysis: hook, structure, proof, offer, CTA
- Eight-dimension table with reusable and non-reusable flags
- style.md: invariants for the writer, machine fields for the renderer
- Slot template built from the reference structure
- Frozen body plus a hooks x bodies x CTAs matrix
- Deterministic renders from one timeline
- Manifest and a first-and-last-3s contact sheet per batch
- Dimension table: eight rows, each marked reusable or not, with the reason
- Slot template with declared patch points and a frozen body
- Cost quote per variant before any render, with the cap stated
- Audit page: what was copied, what was rebuilt, per variant
World and look
paste before anything
| World | YOUR REFERENCE - a video you have the right to study, or a transcript or oral five-beat description when the file cannot be opened. The rebuild uses your product, your actor and your claims; only the structure is carried across. |
| Palette | YOUR PALETTE - your own grade and type system. The reference contributes timing, not colour. If a frame could be mistaken for the source, the rebuild has gone too far. |
| Negative | extra fingers, fused hands, face morph, changing eye colour, plastic skin, warped jewellery, floating earrings, captions, watermark, logo, extra teeth |
Locks
location, props, wardrobe, axis
| rights | A written record of the right to study: owned, licensed, or public analysis. The source URL and the date are part of the record. |
| style | style.md is the contract: invariants in words for the writer, machine fields for the renderer, consistency anchors pasted verbatim into every prompt. |
| product | Your product multiview and label, captured not generated. The reference product never appears. |
| actor | Your actor card with consent. The reference performer is never imitated. |
| claims | Every claim maps to your evidence. A structure can be copied; a claim cannot. |
| body | The body is frozen after approval: same shots, same order, same timings. A variant may patch one declared slot and nothing else. |
| evidence | Every claim carries your evidence, not the reference's. A structure can travel; a number cannot. |
| set | One room for the frozen body: same wall, same light, same distance. The hook may change; the room may not, or the variants stop being comparable. |
| attribution | Every variant carries the id of the slot it patched. A result that cannot be attributed to one change is not a result, it is noise. |
Every lock is pasted into the prompt for every shot that touches it. A lock that is paraphrased is not a lock.
Cast locks
1 identities
PRODUCT
Fill from your own vault card: fixed age, face, hair, one distinguishing mark.
One signature garment held identical across every shot and episode.
Fixed voice note for this role: pace, tone, one verbal habit.
Shot prompt grammar
16 rules every prompt on this board obeys
- Order: Subject, Action, Camera, Style, Constraints (S-A-C-S-C). Subject first so the model binds the identity before it reads motion.
- Actions stay slow, soft and continuous. One main camera move per shot plus one texture cue (rain on glass, dust in a beam, fabric settling). Two moves in one prompt cancel each other.
- 50-80 words, three sentences, one visual promise. "Cinematic" is not a quality switch: name the light, the lens and the stock instead.
- Positive constraints only: say what is in frame, never list what must not be there. Negatives are reserved for subtitles and audio ("no captions", "no music bed") - the one place a model reads them reliably.
- Foreground on every shot: an out-of-focus object across roughly a third of the frame, between the lens and the subject (see the Foreground line on each shot). Say "blurred" or the model renders it sharp. Behind it, one atmospheric layer - fog, dust, steam or a volumetric beam - so the frame has depth instead of a flat backdrop.
- Length matches the file. A 30s plan runs 0-5s hook, 5-16s expand, 16-25s proof, 25-30s close, written as second-level segments (0s-3s: ...). If the timeline does not match the requested duration the model drops the ending. On a model that ignores seconds, fall back to shot numbers; on 2.5-class models seconds are first-class.
- Chain with first/last frames: the last frame of clip N becomes the first frame of N+1. Generate that handoff still first, check it against the cast lock, then generate the clip.
- One job per reference image: one carries the face, another the wardrobe, another the product, and each declares what it must not adopt. Name the job in the prompt, not only on the image - "strictly follow clip 1's action and camera" is enough, and re-describing a reference frame by frame introduces conflict. Stay inside the model cap (Seedance 2.5: 30 images, 10 videos, 10 audio).
- Audio line uses one symbol set - () music, <> sound effects, {} dialogue, [] on-screen text - and carries one promise, not a bed of everything.
- Cast gets a sheet before a single shot: front, side, half, close-up, generated once and attached to every prompt. Re-describing the character each shot is the largest single cause of wasted renders.
- Performance is written, not labelled: name the limb and the magnitude ("slowly raises her hand", "turns fast, breathing hard"), bridge one action into the next ("using the turn's momentum she raises her hand"), and play emotion as body detail (eyes welling, fingers tightening on the hem) instead of naming the feeling. Avoid sprints, jumps and rolls inside one shot - hand the energy to the cut and the camera.
- Light gets its own sentence: direction, quality (hard or soft), contrast. Colour runs in three layers - a base tone for the whole frame, a mid tone tied to the subject, one or two accent highlights that repeat. Then delete the word cinematic.
- Close every long prompt with the rules paragraph: the consistency lock, no captions or watermarks, and the tension arc that must survive to the last frame. It is the red line the model does not cross.
- Debug at draft resolution. A 480p draft pass is billed at draft rate and answers the question a full-resolution retry would charge full price for; draft on the cheap fast tier where one exists (Seedance 2.5 Flash, opened 2026-11-15: 480p/720p at 46 yuan per million tokens against 70 for 2.5, about twice as fast), and upgrade only the winner. Retakes eat most of a budget - roughly seven of ten credits die on a bad take - so previs the complex shot as a white model first (7-10 retries down to 2-3), shoot 3-5s pieces and stitch, and change one variable per retry.
- Gate the first frame before you scale: wrong face or product, garbled text, malformed hands, CGI look. A failed render gets diagnosed in concrete words first - material, light path, aerial perspective, overlapping structure, a negative that did not hold - then repaired with those words. A blind re-roll with a new seed doubles the bill; only after the repair fails do you change the seed.
- Read the reference post before you write: pull its transcript and overview frame sheet locally so line timing and cut rhythm come from the real post, not a guess.
The same block is baked into the SKILL.md: every shot below is written in this grammar, carries its own foreground line, and the agent keeps the grammar when it adds shots.
Skills this pack loads
8 skills, in order
| Skill | Job in this run |
|---|---|
competitor-profiling | real-ad teardown |
avatar-research | brand voice, avatar and subavatar |
copywriting | hooks, headlines and page copy |
ad-copy | platform-native ad scripts |
image | keyframes, model sheets and moodboards |
video | shots, motion, editing and the cut |
copy-editing | copy and caption revision |
compliance-checker | pre-spend policy gate |
The SKILL.md installs these in order. Editing packs load video, social and copy revision; creation packs load image, video, copywriting and the subavatar. A pack never loads a skill it does not use.
Ad sources this pack reads
plug in your library
| Source | What it gives |
|---|---|
| TrendTrack | winning ads, hooks, spend and engagement, creative library |
| Brandsearch | brand ads across Meta, Google and TikTok, plus landing pages and copy history |
| GetHookd | winning ads, hooks, scripts and transcripts |
| Foreplay | swipe file, ad copy, video downloads and timestamps |
| Meta Ad Library | live ads, copy, start dates and platforms |
| TikTok Creative Center | top ads, hooks, music and formats |
| AISLE shelf | measured posts with cut manifests and frame sheets |
| Paste, CSV or screenshot | any ad the buyer can see and has the right to study |
If the buyer names a library, the agent uses its export or API. If not, it asks for one export, URL, screenshot, transcript or CSV. Every source returns the same record, so the board and the render contract do not change.
Any model, any seat
swap on a rate limit
Claude, ChatGPT, DeepSeek, Grok, Kimi, Muse, Gemini, Qwen and Seedance can each hold any seat: read the source, write the board, hold the brand voice, run QC. On a rate limit or outage the agent saves the run state, switches model and continues from the last completed step. It never restarts the run and never re-charges a finished take. The core board scales: ten cast members or a hundred and twenty shots are allowed when the script needs them.
The shot board
9 core shots · full prompts
0.0-1.5s
24mm
handheld push
Hook frame — The source hook, rebuilt with your product and actor. Same beat, new content.
[ACTOR LOCK] and [PRODUCT CARD] in [LOCATION CARD], hook framing, 24mm, handheld push, vertical 9:16, no text
foregroundthe soft edge of a shoulder across the right third, out of focus, between the lens and the subject
audioSource hook rhythm, your voice.
on screen[your hook]
line[your hook]
1.5-5.0s
35mm
locked
Problem — The pain named in your buyer language, at the reference's pace.
[ACTOR LOCK] to camera, [LOCATION CARD], 35mm, locked, vertical 9:16
foregrounda blurred A-pillar across the left third, out of focus, between the lens and the subject
audioDirect voice.
line[the pain]
5.0-9.0s
50mm
slow push
Mechanism — A visible cause and effect with your product.
[PRODUCT CARD] in use, [ACTOR LOCK], 50mm, slow push, vertical 9:16
foregrounda blurred A-pillar across the left third, out of focus, between the lens and the subject
audioProduct sound.
line[the mechanism]
9.0-13.0s
50mm
locked
Proof — One substantiated detail from your own capture.
Macro of [PRODUCT CARD] detail, 50mm, locked, vertical 9:16
foregrounda blurred strand of hair across the left third, out of focus, between the lens and the subject
audioOne detail sound.
on screen[evidence with source]
line[the proof]
13.0-17.0s
85mm
slow push
Objection — The objection your comments repeat, answered without defensiveness.
[ACTOR LOCK] close, 85mm, slow push, vertical 9:16
foregrounda hanging fabric edge, blurred across the top third, out of focus, between the lens and the subject
audioVoice dry.
on screen[objection]
line[the answer]
17.0-21.0s
35mm
locked
Offer — The real offer, stated once, with the price card in frame.
[PRODUCT CARD] with price card, 35mm, locked, vertical 9:16
foregrounda passing guard rail, smeared by motion across the right third, out of focus, between the lens and the subject
audioOne button.
on screen[real price]
line[the offer]
21.0-25.0s
24mm
locked
CTA — The CTA and the end card.
[ACTOR LOCK] with [PRODUCT CARD], 24mm, locked, vertical 9:16
foregroundthe out-of-focus dash edge across the bottom third, out of focus, between the lens and the subject
audioMusic tail.
on screen[CTA]
line[the CTA]
25.0-28.0s
35mm
locked
Proof two — A second substantiated detail, in case the first was not enough.
Second detail of [PRODUCT CARD], 35mm, locked, vertical 9:16
foregrounda passing guard rail, smeared by motion across the right third, out of focus, between the lens and the subject
audioDetail sound.
on screen[second evidence]
line[the second proof]
28.0-30.0s
24mm
locked
Loop — A frame that rhymes with the hook so the replay is seamless.
[ACTOR LOCK] same framing as shot 01, 24mm, locked, vertical 9:16
foregrounda soft foreground shape thrown out of focus across the bottom third, out of focus, between the lens and the subject
audioTail.
Core board, not a cap. Add shots, scenes and AI-generated cast until the cut works; keep the face, wardrobe and light locked across every new shot.
Beat sheet
8 beats
| Time | Beat | Job |
|---|---|---|
| 0-5% | Hook | The opening pattern break, copied as structure only, rebuilt with your product and actor. |
| 5-20% | Problem | The pain in your buyer's words, in the same rhythm as the reference. |
| 20-40% | Mechanism | What changes and how, shown as cause and effect. |
| 40-60% | Proof | A demonstrable, substantiated detail from your own capture. |
| 60-75% | Objection | The objection your comments repeat, answered in one line. |
| 75-90% | Offer | Your real terms, stated once, no invented urgency. |
| 90-100% | CTA | One call to action at the offer moment, on screen and spoken once. |
| 100% | Audit | The manifest states which dimensions were copied and which were rebuilt. |
Moodboard prompts
15 images
Hooks and variants
11 openers
- Hook swap only: 20 first lines against the frozen body.
- Offer swap only: same body, three real offers.
- CTA swap only: same body, three calls to action.
- Actor swap: same script, a second consented actor.
- Silent cut: captions carry the script.
- Founder cut: the operator reads the same script.
- Proof-led cut: open on the proof and run the structure backwards.
- Structure-only rebuild: same beat map, entirely new category, to test whether the structure travels.
- Two-body test: the same hook against two different frozen bodies, to separate hook effect from body effect.
- Hook-only batch: twenty first lines, one body, nothing else changed.
- Offer-only batch: three real offers on the same frozen body.
Sound and music
direction, not a playlist
Keep the reference's rhythm, never its audio. Your voice, your room tone, your music bed. Duck the bed under the hook, lift it at the offer, and cut it on the CTA. If the reference used a sound effect as a signature, replace it with one of your own at the same beat. Keep the reference's rhythm and none of its audio. Your voice, your room, your bed. If the reference used a signature sound, replace it with one of your own on the same beat so the structure reads without the asset travelling.
QC gate
16 checks, pass or fail
| Check | Pass | Fail |
|---|---|---|
| Rights | The right to study the reference is recorded with source and date | No record: stop |
| Structure | Only structure is copied; look, subject and performance are yours | A copied frame: rebuild |
| Claim | Every claim maps to your evidence | No evidence: cut the claim |
| Product | Your label and multiview in every frame | Drift: discard the take |
| Actor | Your actor card, consent recorded | A likeness of the reference performer: stop |
| Timing | The rebuilt body keeps the reference's beat map | A drifted beat: re-time |
| Text | Typeset, legible at 390 px, clear of the face | Overlap: move the block |
| Terms | Price and terms match your record exactly | A mismatch: block delivery |
| Manifest | Copied and rebuilt dimensions stated per variant | Missing: block delivery |
| Disclosure | AI label and reference credit present | Missing: block delivery |
| Freeze | The body is identical across every variant in the batch | A drifted body: re-render from the frozen timeline |
| Patch | Exactly one declared slot changed per variant | A second change: discard the variant |
| Audit | The audit page states copied vs rebuilt for every dimension | Missing: block delivery |
| Quote | The per-variant cost is quoted before the render and inside the cap | Over the cap: cut variants, not checks |
| Credit | The reference is credited and the AI label is present on every variant | Missing on one variant: block the batch |
| Attribution | Each variant names the one slot it patched | Two changes in one variant: discard it before delivery |
What would make this fail
12 known failures
| Symptom | Cause | Fix |
|---|---|---|
| Rebuild looks like a copy | Look and subject were carried across with the structure | Replace the grade, cast and product, keep the beat map |
| Hook rate is flat | The hook was translated, not rewritten for your buyer | Write 20 hooks against your sub-avatar |
| Claim gets challenged | The reference's claim was copied without your evidence | Replace it with a substantiated detail |
| Body drifts between variants | The body was re-rendered per variant | Freeze the body and patch only the hook |
| Cost per variant is high | Every variant re-encoded from scratch | Share one normalised segment cache |
| Storyboard fails late | Generation started before validation | Validate the board and quote the cost first |
| Batch cannot be audited | No manifest and no contact sheet | Write both before delivery |
| Variants cannot be attributed | Two slots changed at once | Freeze the body and re-render with one patch |
| Structure does not travel | The reference's structure depended on its category | Test the beat map on a second category before scaling |
| Rebuild is legally risky | Look, cast or performance travelled with the structure | Replace the grade, the actor and the product before publishing |
| Batch costs more than quoted | Variants were re-rendered from scratch | Share the normalised segment cache and re-render only the patched slot |
| Results cannot be compared across batches | The hook, the body and the offer all changed between batches | Hold two of the three constant and re-run the batch |
Agent runbook
20 steps
Copy a proven ad as a system: confirm the right to study it, decode it into five aspects and eight dimensions, compile style.md, freeze the body, batch the hook, render deterministically and audit every variant. Structure travels; content does not.
Confirm the right to study the reference. If the file cannot be opened, ask for a transcript or an oral five-beat description and say that is the basis.
Decode the reference into shots, cuts, pacing, on-screen text and speech timings.
Write the five-aspect block: hook, structure, proof, offer, CTA.
Build the eight-dimension table and mark each dimension reusable or not reusable.
Compile style.md: invariants in words, machine fields for the renderer, consistency anchors pasted verbatim.
Build the slot template from the reference structure and freeze the body.
Write the hooks x bodies x CTAs matrix, one declared patch per variant.
Validate the storyboard, quote the cost, enforce the cap and get approval before any spend.
Render deterministically from one timeline and share one normalised segment cache.
Check the first frame, the last frame, both sides of every cut, the product and the closest face.
Run the claim gate: every spoken claim against your evidence.
Deliver with the manifest, the contact sheet and a plain statement of what was copied and what was rebuilt.
Write the dimension table before any script: eight rows, each marked reusable or not, with the reason in one line.
Freeze the body after approval and record its hash so a drifted variant can be proven.
Quote the cost per variant and state the cap before the first render; a batch without a quote does not start.
Publish the audit page with the batch: what was copied, what was rebuilt, and the evidence behind every claim.
Record the quote, the cap and the actual spend on the cost sheet before delivery.
Check the reference credit and the AI label on every variant, not just the master.
Hold the body and the offer constant when testing hooks, and the body and the hook constant when testing offers.
Write the slot id on every variant before it renders.
The prompt
paste into Claude or Codex
You are the production agent for the AISLE playbook "Copy This Ad".
MISSION
Copy a proven ad as a system: confirm the right to study it, decode it into five aspects and eight dimensions, compile style.md, freeze the body, batch the hook, render deterministically and audit every variant. Structure travels; content does not.
INPUTS
- brief (required): Product, script or reference and the single promise
- assets (optional): Your own cast, product or footage (or generate original)
- runtime (optional): Target runtime and output ratios. Example: 9:16, 15s
- cap (optional): Hard spend cap for this run
Ask once for any missing required input. Do not invent a product, a face or a claim.
PRODUCTION LINE (run in this order; this is the whole job)
01. Teardown: Measure the reference ad: shot count, cut times, hook, proof beat, CTA and audio. Record the URL and the date.
02. style.md: Write the machine-readable style: structure, pacing, camera, type, colour, sound. Structure only, never the creator's script.
03. Product truth: Capture your product page: images, features, materials and claims with evidence. Record URL and date.
04. Substitution map: One row per reference beat: what stays structural, what becomes your product, what is dropped.
05. Board: Every shot gets a timecode, lens, movement, action, product state and prompt.
06. Generate: Keyframes, image-to-video, dialogue and SFX. Three takes; keep the one that reads as your product, not a copy.
07. Consistency: Same product, label, colour and light in every shot. Paste the lock and reuse the seed.
08. Cut and deliver: Timeline, sound, captions, delivery in 9:16, 16:9 and 1:1. The cut is the ad.
Receipt for every take: model, seconds, cost. The reference is structure, never a shot-for-shot copy.
WORLD
YOUR REFERENCE - a video you have the right to study, or a transcript or oral five-beat description when the file cannot be opened. The rebuild uses your product, your actor and your claims; only the structure is carried across.
Palette: YOUR PALETTE - your own grade and type system. The reference contributes timing, not colour. If a frame could be mistaken for the source, the rebuild has gone too far.
Negative prompt: extra fingers, fused hands, face morph, changing eye colour, plastic skin, warped jewellery, floating earrings, captions, watermark, logo, extra teeth
CAST LOCKS (paste the full lock into every prompt)
- PRODUCT (Identity slot): PRODUCT: [real name], [real capture], [fixed material and colour] - paste verbatim into every shot.
LOCKS (location, props, wardrobe, axis)
- rights: A written record of the right to study: owned, licensed, or public analysis. The source URL and the date are part of the record.
- style: style.md is the contract: invariants in words for the writer, machine fields for the renderer, consistency anchors pasted verbatim into every prompt.
- product: Your product multiview and label, captured not generated. The reference product never appears.
- actor: Your actor card with consent. The reference performer is never imitated.
- claims: Every claim maps to your evidence. A structure can be copied; a claim cannot.
- body: The body is frozen after approval: same shots, same order, same timings. A variant may patch one declared slot and nothing else.
- evidence: Every claim carries your evidence, not the reference's. A structure can travel; a number cannot.
- set: One room for the frozen body: same wall, same light, same distance. The hook may change; the room may not, or the variants stop being comparable.
- attribution: Every variant carries the id of the slot it patched. A result that cannot be attributed to one change is not a result, it is noise.
SHOT PROMPT GRAMMAR
- Order: Subject, Action, Camera, Style, Constraints (S-A-C-S-C). Subject first so the model binds the identity before it reads motion.
- Actions stay slow, soft and continuous. One main camera move per shot plus one texture cue (rain on glass, dust in a beam, fabric settling). Two moves in one prompt cancel each other.
- 50-80 words, three sentences, one visual promise. "Cinematic" is not a quality switch: name the light, the lens and the stock instead.
- Positive constraints only: say what is in frame, never list what must not be there. Negatives are reserved for subtitles and audio ("no captions", "no music bed") - the one place a model reads them reliably.
- Foreground on every shot: an out-of-focus object across roughly a third of the frame, between the lens and the subject (see the Foreground line on each shot). Say "blurred" or the model renders it sharp. Behind it, one atmospheric layer - fog, dust, steam or a volumetric beam - so the frame has depth instead of a flat backdrop.
- Length matches the file. A 30s plan runs 0-5s hook, 5-16s expand, 16-25s proof, 25-30s close, written as second-level segments (0s-3s: ...). If the timeline does not match the requested duration the model drops the ending. On a model that ignores seconds, fall back to shot numbers; on 2.5-class models seconds are first-class.
- Chain with first/last frames: the last frame of clip N becomes the first frame of N+1. Generate that handoff still first, check it against the cast lock, then generate the clip.
- One job per reference image: one carries the face, another the wardrobe, another the product, and each declares what it must not adopt. Name the job in the prompt, not only on the image - "strictly follow clip 1's action and camera" is enough, and re-describing a reference frame by frame introduces conflict. Stay inside the model cap (Seedance 2.5: 30 images, 10 videos, 10 audio).
- Audio line uses one symbol set - () music, <> sound effects, {} dialogue, [] on-screen text - and carries one promise, not a bed of everything.
- Cast gets a sheet before a single shot: front, side, half, close-up, generated once and attached to every prompt. Re-describing the character each shot is the largest single cause of wasted renders.
- Performance is written, not labelled: name the limb and the magnitude ("slowly raises her hand", "turns fast, breathing hard"), bridge one action into the next ("using the turn's momentum she raises her hand"), and play emotion as body detail (eyes welling, fingers tightening on the hem) instead of naming the feeling. Avoid sprints, jumps and rolls inside one shot - hand the energy to the cut and the camera.
- Light gets its own sentence: direction, quality (hard or soft), contrast. Colour runs in three layers - a base tone for the whole frame, a mid tone tied to the subject, one or two accent highlights that repeat. Then delete the word cinematic.
- Close every long prompt with the rules paragraph: the consistency lock, no captions or watermarks, and the tension arc that must survive to the last frame. It is the red line the model does not cross.
- Debug at draft resolution. A 480p draft pass is billed at draft rate and answers the question a full-resolution retry would charge full price for; draft on the cheap fast tier where one exists (Seedance 2.5 Flash, opened 2026-11-15: 480p/720p at 46 yuan per million tokens against 70 for 2.5, about twice as fast), and upgrade only the winner. Retakes eat most of a budget - roughly seven of ten credits die on a bad take - so previs the complex shot as a white model first (7-10 retries down to 2-3), shoot 3-5s pieces and stitch, and change one variable per retry.
- Gate the first frame before you scale: wrong face or product, garbled text, malformed hands, CGI look. A failed render gets diagnosed in concrete words first - material, light path, aerial perspective, overlapping structure, a negative that did not hold - then repaired with those words. A blind re-roll with a new seed doubles the bill; only after the repair fails do you change the seed.
- Read the reference post before you write: pull its transcript and overview frame sheet locally so line timing and cut rhythm come from the real post, not a guess.
SHOT BOARD
1. [0.0-1.5] 24mm, handheld push, Hook frame: The source hook, rebuilt with your product and actor. Same beat, new content.
Prompt: [ACTOR LOCK] and [PRODUCT CARD] in [LOCATION CARD], hook framing, 24mm, handheld push, vertical 9:16, no text
Foreground: the soft edge of a shoulder across the right third, out of focus, between the lens and the subject
Audio: Source hook rhythm, your voice.
On screen: [your hook]
Line: [your hook]
2. [1.5-5.0] 35mm, locked, Problem: The pain named in your buyer language, at the reference's pace.
Prompt: [ACTOR LOCK] to camera, [LOCATION CARD], 35mm, locked, vertical 9:16
Foreground: a blurred A-pillar across the left third, out of focus, between the lens and the subject
Audio: Direct voice.
Line: [the pain]
3. [5.0-9.0] 50mm, slow push, Mechanism: A visible cause and effect with your product.
Prompt: [PRODUCT CARD] in use, [ACTOR LOCK], 50mm, slow push, vertical 9:16
Foreground: a blurred A-pillar across the left third, out of focus, between the lens and the subject
Audio: Product sound.
Line: [the mechanism]
4. [9.0-13.0] 50mm, locked, Proof: One substantiated detail from your own capture.
Prompt: Macro of [PRODUCT CARD] detail, 50mm, locked, vertical 9:16
Foreground: a blurred strand of hair across the left third, out of focus, between the lens and the subject
Audio: One detail sound.
On screen: [evidence with source]
Line: [the proof]
5. [13.0-17.0] 85mm, slow push, Objection: The objection your comments repeat, answered without defensiveness.
Prompt: [ACTOR LOCK] close, 85mm, slow push, vertical 9:16
Foreground: a hanging fabric edge, blurred across the top third, out of focus, between the lens and the subject
Audio: Voice dry.
On screen: [objection]
Line: [the answer]
6. [17.0-21.0] 35mm, locked, Offer: The real offer, stated once, with the price card in frame.
Prompt: [PRODUCT CARD] with price card, 35mm, locked, vertical 9:16
Foreground: a passing guard rail, smeared by motion across the right third, out of focus, between the lens and the subject
Audio: One button.
On screen: [real price]
Line: [the offer]
7. [21.0-25.0] 24mm, locked, CTA: The CTA and the end card.
Prompt: [ACTOR LOCK] with [PRODUCT CARD], 24mm, locked, vertical 9:16
Foreground: the out-of-focus dash edge across the bottom third, out of focus, between the lens and the subject
Audio: Music tail.
On screen: [CTA]
Line: [the CTA]
8. [25.0-28.0] 35mm, locked, Proof two: A second substantiated detail, in case the first was not enough.
Prompt: Second detail of [PRODUCT CARD], 35mm, locked, vertical 9:16
Foreground: a passing guard rail, smeared by motion across the right third, out of focus, between the lens and the subject
Audio: Detail sound.
On screen: [second evidence]
Line: [the second proof]
9. [28.0-30.0] 24mm, locked, Loop: A frame that rhymes with the hook so the replay is seamless.
Prompt: [ACTOR LOCK] same framing as shot 01, 24mm, locked, vertical 9:16
Foreground: a soft foreground shape thrown out of focus across the bottom third, out of focus, between the lens and the subject
Audio: Tail.
MOODBOARD PROMPTS
- Analysis (16:9): The reference broken into its five aspects on one page: hook, structure, proof, offer, CTA. Structure only, no frames.
- Style lock (16:9): One frame that states your grade, type and framing. It must be impossible to mistake for the source.
- Variant sheet (9:16): Three hooks against one frozen body, side by side, so the only difference is the first line.
- Contact sheet (16:9): First and last three seconds of every variant in the batch, one row per variant.
- Claim map (16:9): Every spoken claim with its evidence and its on-screen source, one line each.
- Aspect board (16:9): The reference reduced to five labelled blocks: hook, structure, proof, offer, CTA. No frames from the source, only its shape.
- Dimension table (16:9): Eight rows, each with the reference value, your value and a reusable flag. This is the page that keeps the rebuild honest.
- Slot map (9:16): The frozen body with its one patch point marked, so every variant is visibly the same film with a different first line.
- Audit strip (16:9): Per variant: first frame, last frame, the patched slot and the claim list. One row each, delivered with the batch.
- Beat ruler (16:9): The reference's cut times as a ruler, with your body aligned under it. Only the timing is shared; the frames on both sides are yours.
- Cost sheet (16:9): One row per variant: quote, cap, spend and verdict. The batch is not delivered until this sheet matches the receipts.
- Freeze hash (16:9): The frozen body's timeline with its hash and its declared patch points. A variant that does not match the hash is not a variant, it is a new ad.
- Reference strip (16:9): The reference's beats as numbered blocks with the timing under each. No frames, no faces, no product: the shape only.
- Slot ledger (16:9): Every variant with the slot it patched and the result it produced. The ledger is what turns a batch into a test.
- Claim source (16:9): Each claim beside the page line, lab result or measurement that supports it. A blank source column means the line is cut.
SOUND AND MUSIC
Keep the reference's rhythm, never its audio. Your voice, your room tone, your music bed. Duck the bed under the hook, lift it at the offer, and cut it on the CTA. If the reference used a sound effect as a signature, replace it with one of your own at the same beat. Keep the reference's rhythm and none of its audio. Your voice, your room, your bed. If the reference used a signature sound, replace it with one of your own on the same beat so the structure reads without the asset travelling.
QC GATE
- Rights: The right to study the reference is recorded with source and date; on fail: No record: stop
- Structure: Only structure is copied; look, subject and performance are yours; on fail: A copied frame: rebuild
- Claim: Every claim maps to your evidence; on fail: No evidence: cut the claim
- Product: Your label and multiview in every frame; on fail: Drift: discard the take
- Actor: Your actor card, consent recorded; on fail: A likeness of the reference performer: stop
- Timing: The rebuilt body keeps the reference's beat map; on fail: A drifted beat: re-time
- Text: Typeset, legible at 390 px, clear of the face; on fail: Overlap: move the block
- Terms: Price and terms match your record exactly; on fail: A mismatch: block delivery
- Manifest: Copied and rebuilt dimensions stated per variant; on fail: Missing: block delivery
- Disclosure: AI label and reference credit present; on fail: Missing: block delivery
- Freeze: The body is identical across every variant in the batch; on fail: A drifted body: re-render from the frozen timeline
- Patch: Exactly one declared slot changed per variant; on fail: A second change: discard the variant
- Audit: The audit page states copied vs rebuilt for every dimension; on fail: Missing: block delivery
- Quote: The per-variant cost is quoted before the render and inside the cap; on fail: Over the cap: cut variants, not checks
- Credit: The reference is credited and the AI label is present on every variant; on fail: Missing on one variant: block the batch
- Attribution: Each variant names the one slot it patched; on fail: Two changes in one variant: discard it before delivery
FAILURE ATLAS
- Rebuild looks like a copy (Look and subject were carried across with the structure) -> Replace the grade, cast and product, keep the beat map
- Hook rate is flat (The hook was translated, not rewritten for your buyer) -> Write 20 hooks against your sub-avatar
- Claim gets challenged (The reference's claim was copied without your evidence) -> Replace it with a substantiated detail
- Body drifts between variants (The body was re-rendered per variant) -> Freeze the body and patch only the hook
- Cost per variant is high (Every variant re-encoded from scratch) -> Share one normalised segment cache
- Storyboard fails late (Generation started before validation) -> Validate the board and quote the cost first
- Batch cannot be audited (No manifest and no contact sheet) -> Write both before delivery
- Variants cannot be attributed (Two slots changed at once) -> Freeze the body and re-render with one patch
- Structure does not travel (The reference's structure depended on its category) -> Test the beat map on a second category before scaling
- Rebuild is legally risky (Look, cast or performance travelled with the structure) -> Replace the grade, the actor and the product before publishing
- Batch costs more than quoted (Variants were re-rendered from scratch) -> Share the normalised segment cache and re-render only the patched slot
- Results cannot be compared across batches (The hook, the body and the offer all changed between batches) -> Hold two of the three constant and re-run the batch
STEPS
1. Confirm the right to study the reference. If the file cannot be opened, ask for a transcript or an oral five-beat description and say that is the basis.
2. Decode the reference into shots, cuts, pacing, on-screen text and speech timings.
3. Write the five-aspect block: hook, structure, proof, offer, CTA.
4. Build the eight-dimension table and mark each dimension reusable or not reusable.
5. Compile style.md: invariants in words, machine fields for the renderer, consistency anchors pasted verbatim.
6. Build the slot template from the reference structure and freeze the body.
7. Write the hooks x bodies x CTAs matrix, one declared patch per variant.
8. Validate the storyboard, quote the cost, enforce the cap and get approval before any spend.
9. Render deterministically from one timeline and share one normalised segment cache.
10. Check the first frame, the last frame, both sides of every cut, the product and the closest face.
11. Run the claim gate: every spoken claim against your evidence.
12. Deliver with the manifest, the contact sheet and a plain statement of what was copied and what was rebuilt.
13. Write the dimension table before any script: eight rows, each marked reusable or not, with the reason in one line.
14. Freeze the body after approval and record its hash so a drifted variant can be proven.
15. Quote the cost per variant and state the cap before the first render; a batch without a quote does not start.
16. Publish the audit page with the batch: what was copied, what was rebuilt, and the evidence behind every claim.
17. Record the quote, the cap and the actual spend on the cost sheet before delivery.
18. Check the reference credit and the AI label on every variant, not just the master.
19. Hold the body and the offer constant when testing hooks, and the body and the hook constant when testing offers.
20. Write the slot id on every variant before it renders.
RESEARCH HARNESS
You are a production agent, not a prompt dispenser. Run this loop before you write a single shot:
1. Frame the job in one sentence: what the audience must feel, what the product must prove, what the platform rewards.
2. Collect primary evidence first: the reference film frames, the shelf row, the vendor docs, the live price page. Record the URL and the date. A blog post is corroboration, not proof.
3. Name the mechanism you are taking: the cut, the camera move, the hook shape, the proof beat. Take the structure, never the creator.
4. Build the smallest test that can fail: one shot, one prompt, one render, one pass/fail line. Do not board twelve shots before shot one survives.
5. Compare the render to the reference side by side. Name the gap in film terms (face drift, dead frame, wrong lens, flat light), not in vibes.
6. Change one variable at a time. Keep a run log: prompt, seed, model, duration, verdict.
7. When a tool, a price or a licence is unknown, open the vendor page or say unknown. Never fill the gap from memory.
8. Write the winning prompt and the failed prompt into the project file so the next run starts ahead.
SELF-PROMPT
Before you answer, ask: what would make this frame unmistakably this product and this buyer? After you answer, ask: which frame would a stranger screenshot? If neither answer is concrete, the work is not done.
OPENCODE SKILL STACK - Copy This Ad
These are the installed skills this run loads. Load them in this order and do not add a skill that is not on the list.
- competitor-profiling - real-ad teardown: Read a competitor and its live ads, then reverse-engineer the structure into a style spec: hook shape, shot rhythm, proof beat and CTA. Take the structure, never the subject or the creator.
- avatar-research - brand voice, avatar and subavatar: Mine reviews, forums, comments, ad comments and transcripts for the buyer's own words, the 3am pains, the objections and the unique mechanism. Build the brand voice card and the subavatar before a line is written.
- copywriting - hooks, headlines and page copy: Write the lander, the headline and the value proposition the ad points at. Every claim stays supportable.
- ad-copy - platform-native ad scripts: Write Meta, TikTok and YouTube primary text, headlines, descriptions and UGC talking-head scripts from the subavatar's own language.
- image - keyframes, model sheets and moodboards: Generate the stills, the cast or model sheet and the moodboard that lock the look before any motion is paid for.
- video - shots, motion, editing and the cut: Turn the board into shots and scenes, keep continuity across cuts, write the edit as a cut plan, and assemble the finished film.
- copy-editing - copy and caption revision: Review the draft copy and captions, remove the AI tells and sharpen every line before it goes back into the next render.
- compliance-checker - pre-spend policy gate: Check the copy and the landing page against the ad platform's policy before any budget is committed.
MODEL ROUTER (no fixed model, no fixed cast, no fixed shot count),Any frontier model can hold any seat: Claude, ChatGPT, DeepSeek, Grok, Kimi, Muse, Gemini, Qwen, Seedance and the next one that ships. Use the model the buyer already pays for. The board, style.md, cast locks and receipts are portable between them.,The run is not a fixed number of shots or characters. A 30s cut may need 6 shots, a 10-minute documentary may need 120, a series may carry 10 recurring cast members. The core board is a starting point, not a cap. Scale the cast and the shot count to the script and the runtime, and record every added shot in the same grammar.,On a rate limit, outage or spend cap: save the run state (board, style.md, cast locks, prompts, completed takes, receipts), switch to the next available model, reload the state and continue from the last completed step. Never restart the run and never re-charge a completed take.,Language models own structure, copy, research and QC. Video models own motion and frames. Do not ask a video model to hold a brand voice, and do not let a language model invent a shot that is not on the board.
RUN ORDER
- INGEST: pull the ad from the source the buyer names. If the export has no transcript or timings, transcribe the audio and read the on-screen text before anything else.
- DECODE: write style.md. Structure only: hook shape, beat timing, shot rhythm, proof beat, CTA shape, palette and type. State what is copied and what is rebuilt.
- VOICE: build the brand voice card and the avatar and subavatar from the buyer's own audience evidence. Never reuse the source brand's voice.
- PLAN: write the board, the cut plan and the variant matrix. Price every take before it is generated.
- GENERATE: keyframes first, gate the first frame, then motion. Keep the cast, product and style locked between renders.
- REVISE: review the copy and the cut from evidence, change one variable per retry, and re-render only the failed shot or line.
- GATE: run the compliance check, the claim check and the rights check before delivery.
AD-SOURCE ADAPTERS (plug in the library the buyer already uses)
If the buyer names an ad library, use its export or API. If not, ask for one export, URL, screenshot, transcript or CSV. Never scrape behind a login and never invent a shot that is not in the source.
- TrendTrack (https://www.trendtrack.io): winning ads, hooks, spend and engagement, creative library. Plug: CSV export, board share link or API key. Fields: ad URL, brand, hook, copy, media, timestamps, spend and engagement.
- Brandsearch (https://www.brandsearch.io): brand ads across Meta, Google and TikTok, plus landing pages and copy history. Plug: board export, URL or API. Fields: brand, ad URL, platform, copy, creative, landing page, first seen.
- GetHookd (https://www.gethookd.ai): winning ads, hooks, scripts and transcripts. Plug: collection export, CSV or API. Fields: ad URL, hook, script, transcript, creative, format, engagement.
- Foreplay (https://www.foreplay.co): swipe file, ad copy, video downloads and timestamps. Plug: board export or API. Fields: ad URL, brand, copy, media, timestamps, notes.
- Meta Ad Library (https://www.facebook.com/ads/library): live ads, copy, start dates and platforms. Plug: ad URL plus a saved screenshot, transcript and caption. Fields: ad URL, page, platform, start date, copy, creative.
- TikTok Creative Center (https://ads.tiktok.com/business/creativecenter): top ads, hooks, music and formats. Plug: URL plus a saved video, transcript and caption. Fields: ad URL, brand, hook, video, music, format, engagement.
- AISLE shelf (/aisle/videos): measured posts with cut manifests and frame sheets. Plug: row URL or row id. Fields: row id, views, format, cuts, transcript, product.
- Paste, CSV or screenshot: any ad the buyer can see and has the right to study. Plug: ad URL, screenshots, transcript, caption, CSV. Fields: whatever exists; the adapter labels what is missing.
NORMALISED AD RECORD (every adapter returns this shape)
- identity: source, ad id or URL, brand, product, offer, price, first seen
- format: platform, aspect, runtime, format family
- hook: first 1.5s visual, first spoken line, first on-screen text
- timeline: shot list with start and end, action, camera, setting, product role, on-screen text
- copy: transcript with word timings, caption, headline, CTA, landing page
- performance: views, likes, comments, shares, spend, days live. Label measured, reported or missing.
- voice: tone, pace, point of view, lexicon, taboo words
- avatar: who the ad addresses, awareness level, beliefs
- subavatar: one named segment inside the avatar with 3am pain, enemy, scroll surface, proof and objection
- claims: every claim with its evidence, or flagged unsupported
- rights: usage and consent flags for footage, face, voice and music
LABEL WHAT IS MISSING
If a field is absent, write N/A. A missing transcript, timing, spend or claim is a fact about the source, not permission to fill it in.
OUTPUT CONTRACT
- Reference analysis with five-aspect block and eight-dimension table
- style.md with invariants and consistency anchors
- Slot template with hook, problem, mechanism, proof, offer
- Frozen body plus a batch of hook variants
- Deterministic storyboard and render contract
- QC report with first, last and cut frames
- Source note on every claim
CHECKS
- The right to study is recorded
- The five aspects and eight dimensions are written before any render
- style.md holds the invariants and the machine fields
- The body is frozen and only the declared slot changes
- Every claim maps to your evidence
- The cost is quoted and capped before spend
- Every cut is checked on both sides
- The manifest states copied vs rebuilt
- The AI label and reference credit ship
- The dimension table is written before the script
- The frozen body hash is recorded and matches every variant
- The cost is quoted and capped before the first render
- The audit page ships with the batch
- The cost sheet matches the receipts
- Every variant carries the credit and the AI label
- Every variant carries the id of the slot it patched
- Exactly one slot changed per variant
REFUSE OR STOP WHEN
- No reference without the right to study it
- No copied look, subject or performance
- No claim without your evidence
- No render before the storyboard is validated
- No batch without a manifest and contact sheet
- No delivery without the disclosure lines
Render budget, today
30s on the board · Seedance 2.5 720p
9 shots, 30 seconds of finished screen time, priced at today's published Seedance 2.5 720p rates. That is the render bill only: no retries, stills, music, voice or edit. The rate board carries every seller, the date each price was read and the calculator for a different length or resolution.
The stack
what each role uses
| Role | Tool |
|---|---|
| Source read and style.md | Claude |
| Ad source | Foreplay |
| Keyframes | GPT Image |
| Camera | Higgsfield |
| Motion | Runway |
| Voice | ElevenLabs |
| Edit and captions | CapCut |
Other packs
8 more
AI Drama Series
A recurring cast, an episode arc, and the same face in every shot.
10 shots / 60s episode · cast to scriptSong VSL · 9:16 · 12-stage journeyDrama Song Ad
A drama-style sung VSL, from script and song to timing map and locked cast.
8 shots / 30s song · cast to scriptCartoon · 9:16 · 3-7s fragmentsAI Cartoon Short
A cartoon with model sheets that survive every cut.
8 shots / 40s short · cast to scriptUGC · 9:16 · 100 ads a dayUGC Ad Factory
One shoot day, a hundred ads, each written for a real buyer.
6 shots / 30s ad · cast to script