STORYBOARD — V2 My Daughter Weighed It for Thirty Days — VO-only × D09 Prop-Demo (measured-artifact object lesson) [KD_C0090]

The locked script is an Echo-chain deliverable and does not carry v2 Fields 19/20/21/25. Mapping used:

The no-presenter ruling (standing order, and the single easiest thing for a production team to break). This ad has no face, no talking head, no piece-to-camera and no credentialed persona at any point. The donor's [R] §09 records raw simulated-UGC home footage plus stock plus 3D medical animation under a first-person VO, with no presenter appearing anywhere, and the batch's standing order (research_brief line 23, and again in the V5 conflicts resolution) forbids inventing one: "V5 is the one slot in this batch where an on-camera speaker is donor-sanctioned. V2 and V4 must NOT invent one." Disposition, boarded: the mother exists as a voice and as hands, a wrist, a forearm and one over-the-shoulder POV. The daughter exists as hands, handwriting and one back-of-head silhouette at the fridge (the script's own last shot). The pharmacist exists as hands on a retail counter and a lanyard, nothing more. The doctor exists as a hand writing on a pad. Casting a presenter, adding a face, or opening on a person addressing the lens are all BANNED on this board. product_facts §5 is held by construction: nobody on screen carries a credential.

Timecode basis, stated once. The per-beat headers below are the word-derived cumulative boundaries, because that is the basis the downstream parser and the editor's cut plan share. The three [pause] holds add ≈0.8s each and push the delivered boundaries later from the Reveal beat onward — the delivered column is in the pause note below, and the headers are never adjusted for it (a decimal or a per-variant header would break the parse).

Review-flag dispositions (ship order, gates delegated). Copy is LOCKED — this storyboard changes zero words. The review predates the v1.1 and v1.2 fix passes; where it conflicts with the current script text, the SCRIPT wins, and both HARDs are already resolved in the copy this board is built on. ① HARD-1 (the problem→solution bridge) — resolved in copy at v1.1. Boarded so the bridge is also visible: Beat 7 cuts out of the animation for one beat to a creased parking stub with a stranger's handwriting across the back, which is the third handwriting artifact in the ad (green marker → fridge dates → the stub) and the visual proof the recommendation was scribbled, not sold. ② HARD-2 (orphaned pronoun in the close) — resolved in copy at v1.1. Boarded so the referent is unambiguous sound-off as well: the real Kindled bottle is on screen continuously through the entire close, beside the scale, so "it" has a picture and not only an antecedent. ③ S2 (the AS-IS permanence line carried slightly cooled) — a copy call the human owns; the board does not touch it and does not compensate for it visually. The Inquisition beat is boarded at full aggression: the permanence line lands on a hard cut with the music out. ④ S6 (no quotation marks on reported speech) — the review calls this a VO-session flag, not a copy defect. Boarded: the pharmacist's four sentences are given a distinct visual world (retail counter, end cap, script pull) that does the voice-separation work the punctuation does not, so the viewer never confuses who is speaking. ⑤ S8 (close smallness) — resolved at v1.1; the close is now 22 words at 4.20% against the donor's 4.23%. Boarded as the smallest beat, held on two objects, with no urgency stack, no countdown, no discount badge and no burned guarantee badge. See the end-card ruling in the OST section. ⑥ S10 (hook B and C converge on the provenance sentence) — the provenance is non-negotiable per the concept. Boarded around: the three hooks are given three genuinely different visual worlds (forensic close-up / behavioural inventory / time-lapse), so the interchange test is carried by the picture even where the last sentence converges.

Hook-variant handling (Echo 1×3 contract). Beat 1 exists as THREE storyboard variants — 1A / 1B / 1C, interchangeable opening sections. Beats 2 to 9 are the shared locked-body storyboard. All three hooks are boarded at 0:00 to 0:17. The hook beat's duration is the donor's own duration-share (16.65s), not a word-derived figure, so hook C's one-word deficit (41w vs 42w, 0.4s at 170 wpm) sits inside the hook beat as breathing room and the body boundaries are IDENTICAL on all three cuts. There is no per-variant shift anywhere on this board. All three exit on the IDENTICAL seam frame — a locked daylight macro of the taped mesh screen sitting flat on the bathroom counter beside the scale, both empty, tare light on — so any variant cuts straight onto Beat 2's opening line ("She'd tape the screen down before my shower…") with no re-introduction and no visual discontinuity.

Delivery markup. [pause] ×3 and [quieter] ×1 are stripped from all on-screen text and carried into per-beat DIRECTION notes. The three pauses are holds and are boarded as picture events: [pause] opening the Reveal beat (hold on the poster board), [pause] mid-Inquisition before the permanence line (the one cut in that beat that is allowed to breathe), [pause] after "It stops" in Root Cause (the follicle goes quiet on screen). [quieter] on the final Results sentence is voice only, boarded as a picture that does not change. CAPS emphasis (WEIGHED, DIE, ON PURPOSE) is voice, never a burn.

Pause note and headroom. 3 delivery tokens ≈ +2.4s over word-derived runtime → delivered A 202.2s · B 202.2s · C 201.8s against the donor band top of 207s. Headroom vs the band top: A 4.8s · B 4.8s · C 5.2s. Headroom vs the 240s ceiling: 37.8s on the worst variant — the ceiling never binds. This is the tightest band margin in the batch and it is stated plainly rather than rounded away. Editors must NOT pad, must keep all three holds tight at ≤0.8s, and must add NO fourth pause. Per the no-word-shaving rule nothing is trimmed elsewhere to buy margin; any future change is a beat REWRITE, never an append. Budget before the band top is breached: ≈13 spoken words.

Nine-beat discipline (donor budget preserved). 9/9 donor beat order, no beat added, dropped or re-ordered. Heaviest beats by duration are Reveal 30.0s and Root Cause 28.9s — the donor's own two longest slots — while the heaviest by word-share is Inquisition at 22.14% against the donor's 22.62%, which is the density spike the whole ad is built on. The board reproduces both facts as picture: the two longest beats are the two slowest-cut beats on the board, and the densest beat is the fastest-cut beat on the board. The close is the smallest beat and is boarded on two objects and a stop.

delivery_catalog §D09 and the donor card are unambiguous: in a D09 the object carries the ad. In the donor, Cup 5 versus CLR in a toilet bowl is the ad. In ours, the artifact is a hand-drawn chart produced by a twelve-year-old, and the instrument is a kitchen scale. If the chart is decorative, the format is not being executed. Everything below exists so that an editor cannot accidentally demote it to B-roll.

The three physical layers

⚑ THE WITHHELD-NUMBER LAW (the board's hardest rule, and it is doing two jobs at once)

No legible measured value ever appears on screen. Not on the scale, not on the chart's Y axis, not on a card.

Two independent reasons, both binding, which is why this is a law and not a preference:

1. It is the ad's engine. The hook is "by day nine she stopped telling me the number." The withheld number is the open loop that carries a 200-second ad. A legible readout in the first ten seconds closes the loop and kills the ad.
2. It is the no-invented-numbers rule. The script contains zero invented numbers and the self-audit says so line by line. A legible "4.7 g" burned onto a scale is a fabricated measurement claim on screen, which is exactly the class of thing product_facts §6 and the house rule forbid. The picture may not assert what the copy refused to assert.

How the law is executed, every time the scale or the chart is on screen:

• The scale's LCD settles and we cut, or the clump occludes it, or it is shot at an oblique angle where the segments are legible as digits but not as a value, or it sits soft in a shallow-depth foreground. Digits reading as motion, never as a number.
• The chart is legible as shape, handwriting and dates. The Y-axis numbers are either absent (the child ruled the grid but labelled it in her own shorthand) or rendered below reading size. The script's own Results-beat direction already says it: "the two slopes side by side, no numbers legible." That rule is extended backwards over the whole board.
• The one comparison the ad does make out loud — "the drain weighed more than four times the hundred" — is boarded as a physical ratio, not a printed one: the counted hundred hairs laid in a thin row on the counter beside the drain clump in the same frame, same lens, same light. The viewer does the arithmetic with her eyes. Nothing is printed and nothing is invented.
• The only counting number allowed on screen is Hook C's running day-counter (1 to 30), which is a count of days, not a measurement, and which the script explicitly specifies.

The chart, state by state (the co-star's continuity spec)

The chart must be one physical object that ages, built once by the art department in a genuine twelve-year-old's hand, photographed in five states. Same paper, same pen weights, same ruled grid, same slight buckle where the tape has pulled. A re-drawn or CGI chart will read instantly and will cost the ad its only proof.

Chart law, additional: CHART-4 is never shown alone and never shown before beat 8. The two charts are only ever compared in one frame, on the same fridge, in the same daylight — no split screen, no wipe, no before-and-after title. The script's brand_bible §9.2 constraint is absolute and boarded: no before-and-after of a head appears anywhere in this ad.

Hard prop rules

• Real household, real daylight, real mess. One kitchen, one bathroom, one fridge. Bathroom light is the actual window at roughly 6am; kitchen is late afternoon. No seamless, no styled tabletop, no product-photography lighting anywhere except the single dropper macro in beat 7.
• Imagery is never self-nerfed. A clump of shed hair on a kitchen scale pan is the entire premise. It is shot in daylight, in focus, at scale. It is not prettified, softened, shot out of focus, colour-graded warm, cropped down, or replaced with a symbolic stand-in. This is a standing rule and it is the first thing a nervous edit will break.
• Only one branded object exists in the ad: the Kindled bottle, and it does not appear until beat 7. Every competing category item (the biotin end cap, the minoxidil box, the amber prescription bottle) is anonymized by blur, crop or turn. The picture disqualifies a category, never a named competitor.
• Hands only, always. Any human presence is a hand, a wrist, a forearm, a back of a head or a shoulder. If a face enters a frame, the frame is wrong.
• Cuts every 1.5 to 2.5s across the whole ad, with exactly two sanctioned slow zones: the Authority beat (the ad's first hold, [R] "slowing on the pharmacist's dialogue") and the Root Cause animation block. The Inquisition beat is the fastest cutting in the ad.
• Audio spine: cinematic driving documentary bed under the whole run, swelling on the Reveal. Scale-tare clicks and paper SFX carry the demo beats. Heartbeat and low whoosh under the animation. **Music drops out entirely under "they don't grow back," then returns loud** — the script names this and it is the single loudest editorial gesture in the ad.

This is the batch's recurring defect surface (8 cards failed qualify-first across 4 of 5 scripts), and the v1.2 pass re-authored this script's cards to pass. They are carried into this board VERBATIM. Not one character is changed, softened, shortened, re-punctuated or "improved" into a teaser. The board adds ZERO new burned-in attention cards.

Two-line rendering of card 1 is a layout decision, not a copy edit. The verbatim string is two sentences; it is set on two lines so it reads in two beats at phone size. Every character, every word and both full stops are the script's. VS2 flag logged (12 words is at the ceiling, not under it) — see SOFT flags.

The end-card ruling (why nothing is burned on the close). The doctrine is that every burned-in card must qualify sound-off via SYMPTOM / IDENTITY / SITUATION / FAILED_SOLUTION / BELIEF. A guarantee badge, a discount flash, a countdown or a "30-DAY MONEY-BACK" stamp is none of those species — it is an offer element, and it cannot pass the test. So none is boarded. This is also exactly what the script asks for: "No urgency stack, no disqualification block, no countdown, no discount badge. The ad simply stops." The guarantee is spoken, once, in the close, and it stays there. The end card is the wordmark on the product's own label plus one static plate, which is a naming element under the same rule as card 5.

The caption-track ruling (so it cannot smuggle unqualified copy in through the side door). A native platform-style verbatim VO caption track is permitted as a retention and accessibility layer, and it is NOT an OST surface: it is a transcript of words already spoken, never authored copy, never a headline, never an emphasis burn. Two hard conditions: (a) it carries only the exact spoken words, and (b) it is suppressed for the entire first 5 seconds of every hook variant, so the qualifying card owns the frame during the window that decides whether a muted viewer stays. If production cannot honour condition (b), it ships without a caption track.

U12 no-dash scan — RUN, and clean. Every viewer-visible on-screen string above was scanned character by character for — (em dash) and – (en dash). Zero occurrences. The only hyphens present are compound hyphens in SHOWER-DRAIN and MONEY-BACK, which U12 explicitly permits and which stay. Re-scanned after the two-line layout of card 1 and after the end-card ruling. Nothing was recast, because nothing needed it.

Rotation: N/A this run — the donor is PINNED per the run order (SEED, binding). No rotation re-query and no fresh §2b/§2c semantic draw were executed, per the task directive. Supporting anchors are the concept's own cited formula and evidence rows plus the batch's two other measured D09 donors — all declared choices per §2f, which the rotation rule never blocks. Log append SKIPPED (Echo run — handled at package, per the E3 / S2 / S4 convention). Pool note. Same working resolution the E3, S2 and S4 boards reached: a SEED board inherits its anchor set from the pinned donor plus the mini-brief's and script's own citations, and re-queries only when a beat type is left uncovered. Nothing is uncovered here. The donor is itself a D09 with a nine-beat labelled skeleton that matches ours 9/9, so it supplies grammar for every beat; the four supporting anchors cover the three alternate hook shapes, the animation block and the in-niche mechanism beat. All five anchors are video — zero static demotions were needed on this board, which is a first for the chain. Within-batch and cross-session diversity. Three of the five anchors are this batch's own donors (V2, V4, V5), which is a deliberate consequence of a pinned all-SEED run and is declared rather than hidden. Zero overlap with the pilot-1, Run-2 or Run-3 anchor sets except 26839214942445844, which appeared on the E3 board and is re-used as a declared choice here because it is the single best in-niche follicle-starvation evidence in the library and it was formally adopted into this script at the /score-echo angle audit.

[Anchor A — CORNERSTONE] 1238484691057182 — elare — winner_score 73.61 — video (THE DONOR, SEED — pinned CORNERSTONE) Hook pattern: H07 Lecture Bait in its DEMONSTRATED form — a 15s hook beat, 3× the H07 median of 5s. "My daughter dissolved calcium faster than a bathroom cleaner. She's 11. It was a science fair project." hook_seconds.md §H07 cites this exact line as one of its three named archetype exemplars. Grammar: [an innocent] + [a physical thing done to a familiar object] + [an absurd-sounding measured result] + [the disarming reason it happened at all] On-screen text style: sparse. The object carries the frame; text is a caption layer, never a headline stack. No continuous verdict burns B-roll cuing: there is no B-roll, there is a PROP. Raw simulated-UGC home footage of the demo itself, plus stock, plus one 3D medical animation block. The prop is on screen for most of the ad Transition style: cuts every 1.5 to 2.5s, slowing on the pharmacist's dialogue — one deliberate hold in an otherwise high-dynamism ad Mute-test asset: a child's mason jar cut against green cleaner in a toilet bowl. [R]: passes "exceptionally" with no sound

Best-fit beat type in our script: ALL (SEED full-pattern borrow, 9/9 beat order). Primary: the demonstrated hook (1A), the protocol-and-control demo (beat 2), the insider-reads-the-artifact hold (beat 3), the innocent's-motive reveal (beat 4), the reported-dialogue annihilation and agitation rush (beat 5), the sustained animated mechanism block (beat 6), the late product entry (beat 7), the soft-comparative after-state (beat 8), the smallest-beat stop (beat 9)

Tooling stack inferred: real household shoot + stock + one 3D medical animation block under a first-person VO, no presenter — maps 1:1 to our production target. No adaptation strain anywhere: this is the donor whose production we are literally reproducing

[Anchor B] 814438051152496 — lirae — winner_score 81.87 — video (the pool's HIGHEST ws; the batch's V4 donor; also D09) Hook pattern: H01 Symptom Mirror — a second-person conditional inventory, four items deep, that qualifies the viewer against her own daily routine before any claim is made. This is the declared shape source for our Hook B (H01 ← 814438051152496, script formula lineage) On-screen text style: native-styled captions over a rapid compilation; mechanism specificity 5/5, the highest in the batch B-roll cuing: a rapid-fire compilation of low-production UGC clips and stock, unified by a hero tactile demo — the macro softgel slice — used as the pattern interrupt. Cuts every 1.5 to 2s Transition style: hard contrast structure, struggle (dark, indoor) against liberation (bright, outdoor) Mute-test asset: the hero macro demo. The object interrupts, not a face

Best-fit beat type in our script: primary — Hook 1B (the four-item conditional inventory, which our hook reproduces as DOWN / DRAIN / PILLOW / CAR SEAT). Secondary — beat 7's product macro (the hero-tactile-demo grammar: one object, one macro, shot as the pattern interrupt rather than as packshot)

Tooling stack inferred: compilation shoot plus one hero macro demo. FLAG (VS6 cross-adaptation logged): its ugc.compilation-voiceover sub-mode is NOT imported wholesale — our donor is a sustained single-household demo, not a compilation, and its struggle/liberation contrast grammar is rejected outright because a "liberation" montage would import the before-and-after promise brand_bible §9.2 bans. Borrowed at hook-shape and macro-grammar level only. Its single-source lower-confidence caveat (no workbook exists for it) is noted and it is therefore never used to settle a judgment call alone

[Anchor C] 26907486858952165 — resilia — winner_score 65.38 — video (the batch's V3 donor) Hook pattern: H03 Time-Lapse Trial, a 5s hook beat, exactly the H03 median. Formula: [subject] + [protocol] + [explicit timeframe] + [withheld bodily result]. This is the declared shape source for our Hook C (H03 ← 26907486858952165) On-screen text style: a full caption over a cross-sectional 3D render at t=0 B-roll cuing: three alternating pillars — high-end 3D medical animation, raw UGC lifestyle B-roll, and sourcing footage. Pacing is variable by design: fast on the hook, deliberately slow on the animation "so the brain can process the science" Transition style: fast montage with deliberate holds on every animation beat Mute-test asset: the 3D render plus the caption

Best-fit beat type in our script: primary — Hook 1C (the time-lapse trial grammar and the withheld bodily result, which our board executes as the 30-frame same-pan time-lapse). Secondary — beat 6, Root Cause (the animation-hold rule: the mechanism block is the one place the cutting is allowed to slow to a crawl, and this anchor is the corpus's clearest statement of why)

Tooling stack inferred: 3D cellular animation as the spine of the shoot, no on-camera presenter at all — which matches our board exactly on the thing that matters most. FLAG (VS6 cross-adaptation logged): its week-numbered benefit ladder and its manufacturing-scarcity close are NOT imported — our donor's close is the smallest beat in the ad and carries no scarcity, and inventing a count is banned. Borrowed at hook-shape and animation-pacing level only

[Anchor D] 26839214942445844 — trysoluma — winner_score 62.07 — video (the in-niche EVIDENCE anchor, adopted at the /score-echo angle audit) Hook pattern: BETRAY, Authority FAQ variant. Solution-Aware, sophistication 5 On-screen text style: question-headed segment cards B-roll cuing: a dedicated ~55-second beat explicitly labelled "Question 4: Hair Shedding & Follicle Starvation" (transcript window 3:15 to 4:10) — hair follicles as the most T3-dependent tissue, follicles exiting the growth phase early when conversion is blocked, "the clumps in the drain every morning, the part line getting wider," labs reading fine because TSH says nothing about the T3 reaching the follicle, and an explicit biotin disqualification Transition style: segmented FAQ chaptering Mute-test asset: the question card

Best-fit beat type in our script: primary — beat 5, the Inquisition (the biotin disqualification framed as the follicles aren't broken, they're starved — our beat performs the same move in kind, in reported dialogue). Secondary — beat 6, Root Cause (follicle starvation as T3-dependence). This is the highest-value in-niche evidence in the library for this angle and it is why the script's evidence line was corrected upward

Tooling stack inferred: credentialed authority talking head. FLAG (VS6 cross-adaptation logged): the entire presenter apparatus is REJECTED — this board has no face and no credential by standing order, so nothing visual transfers from this anchor at all. It is carried at mechanism-and-symptom-specificity level only. The competitor-copy screen was already run line by line at write time and no phrase from this ad appears in our script; the same discipline applies to the picture — no frame, no card design and no chapter device is borrowed

[Anchor E] 1508278883991064 — lirae — winner_score 60.81 — video (the batch's V5 donor, and delivery_catalog §D09's FIRST LISTED reference ad) Hook pattern: H07 Lecture Bait, 3s / 11 words. [R] mute test 5/5 On-screen text style: a jarring text overlay over a shock visual B-roll cuing: the phase that matters to us is phase 2, 0:03 to 0:09, "empirical proof": rapid cuts between a ticking vintage clock and a digital glucose meter. That is the corpus's cleanest example of a measuring instrument used as a co-star — which is precisely what our kitchen scale has to be Transition style: rapid cuts on the proof phase, ticking-clock and whoosh SFX Mute-test asset: the physical stunt plus the instrument

Best-fit beat type in our script: primary — beat 2, Experiment & Shocking Result (the instrument-as-co-star grammar: the tare button, the settling readout and the tick of the process cut as empirical proof, not as kitchen B-roll). Secondary — the hook mute test (the object arrests before any voice does)

Tooling stack inferred: shot practical stunt + a real measuring device + 3D animation + an on-camera presenter for the mechanism block. FLAG (VS6 cross-adaptation logged): the presenter is NOT imported. The research brief's own resolution is explicit — "V5 is the one slot in this batch where an on-camera speaker is donor-sanctioned. V2 and V4 must NOT invent one." Borrowed at instrument-grammar level only. Its transgression-stunt voltage device is also not imported: our voltage is carried by demonstration and by a converted insider, per the donor

Global visual spine (applies to every beat unless a beat overrides it). 9:16 vertical, shot 4K wide for a 9:16 crop so post can reframe without amputating a hand or a chart edge. NO PRESENTER, NO FACE, NO PIECE TO CAMERA, anywhere, at any point. Every human is hands, a wrist, a forearm, a shoulder or a back of a head. CAMERA: handheld with real micro-drift on the household spine — the ad must feel like it was shot by someone who lives there, never gimbal-smooth; locked on a small tripod for the scale macros only, because the instrument shots need to read as measurement rather than as vibe. LIGHT: practical daylight only. Bathroom is the actual window at roughly 6am, cool and slightly under. Kitchen is late-afternoon warm through a window. The school gym is fluorescent and unflattering. No fill, no bounce, no beauty light, no colour grade toward warmth on the hair. LENS: one lens family throughout the household spine; macros go tighter but never switch aesthetic. CUTTING: 1.5 to 2.5s baseline; slow only on beat 3 and beat 6; fastest on beat 5. AUDIO: cinematic driving documentary bed under the whole run; scale-tare clicks and paper SFX on the demo beats; heartbeat and low whoosh under the animation; music out under "they don't grow back," back loud immediately after. OST: the six declared surfaces above and nothing else. PROP LAW: the chart is on screen in six of nine beats and it ages; the Kindled bottle is the only branded object and does not appear before beat 7; every competing item is anonymized.

PRIMARY_LAYER: real shoot, household, no performer visible above the wrist
SHOT 1 (0.0-2.2s): TOP-DOWN MACRO, locked, 6am bathroom window light.
        Child's hands peel a taped mesh screen off a shower drain. A dark clump sits in the mesh.
        IN FOCUS, AT SCALE, DAYLIGHT. Do NOT soften, blur, warm-grade or crop this shot.
SHOT 2 (2.2-4.6s): 45 DEG MACRO, locked, tripod. Clump transfers onto the kitchen scale pan.
        LCD digits settle. CUT BEFORE ANY VALUE RESOLVES. Occlude with the hand if needed.
        NO legible number may exist in this shot (Withheld-Number Law).
SHOT 3 (4.6-16.6s, intercut): SLOW HANDHELD PUSH on the fridge. CHART-2 taped up,
        one corner curling. Column of dates in a 12-year-old's ballpoint.
        Y-axis values illegible by build. The HANDWRITING is the subject.
CUTTING: 3 to 5 cuts across the beat, 1.5-2.5s, holding longest on the chart
ON_SCREEN_TEXT (post): 2-line card, upper third, IN t=0.6s, HELD to beat end:
        "SHE WEIGHED MY SHOWER-DRAIN HAIR." / "DAY 9 SHE STOPPED SAYING THE NUMBER."
        VERBATIM. Do not re-word, shorten or add a third line.
SFX: single scale-tare click on shot 2; paper/tape texture; documentary bed from t=0
VO DIRECTION: calm, first-person, slightly awe-struck. She is not upset and she is not selling.
        A mother reporting something her child did that turned out to matter.
        "A mesh screen, a kitchen scale, same time each morning." reads as spoken inventory, flat.
        Land "By day nine she stopped telling me the number" with dread, NOT with a reveal lift.
DURATION: 16.6s (donor duration-share basis; 42 spoken words)

PRIMARY_LAYER: real shoot, household + car, hands and objects only
SHOT 1 (0.0-1.6s): MACRO handheld. Hair tie on a wrist, hair being pulled up. Bathroom.
SHOT 2 (1.6-3.4s): TOP-DOWN, low light, pre-dawn. A hand wipes the shower drain clean.
        The gesture must read as SECRETIVE and ROUTINE, not as cleaning.
SHOT 3 (3.4-5.2s): Pillow lifted off a bed, raking morning light so loose strands catch.
SHOT 4 (5.2-7.0s): Car headrest in flat daylight through the window. Strands visible.
SHOT 5 (7.0-16.6s, intercut): The taped mesh screen going down over the drain.
CUTTING: HARD cuts, one per conditional, metronomic. No dissolves anywhere in this beat.
FOUR DIFFERENT LIGHT STATES across shots 1-4 — four moments in one life, not one set-up.
ON_SCREEN_TEXT (post): single-word cards, centre, one per cut, in on the cut, out on the next:
        "DOWN" / "DRAIN" / "PILLOW" / "CAR SEAT"
        VERBATIM. One word per card. Do not combine, expand or add a fifth.
SFX: a soft cloth/water hit per cut; documentary bed from t=0
VO DIRECTION: flat, even, second person, NO rising inflection and NO accusation.
        She is reading a list she already knows. The provenance sentence
        ("My twelve-year-old put a mesh screen over my drain...") arrives as a plain fact.
DURATION: 16.6s (donor duration-share basis; 42 spoken words)

PRIMARY_LAYER: real shoot, time-lapse, locked tripod with a floor mark
SHOT 1 (0.0-13.0s): 30-FRAME TIME-LAPSE. Same kitchen scale pan, same angle,
        same morning window light, 30 consecutive days. The clump changes size day to day.
        REGISTRATION IS CRITICAL — mark the tripod feet and the pan position.
        NO legible value on the LCD in ANY frame (Withheld-Number Law) — shoot with the
        readout occluded by the pan lip or defocused in the foreground.
SHOT 2 (13.0-16.6s): SLOW HANDHELD PUSH onto CHART-3, the science-fair poster board,
        the green-marker hypothesis line visible but not yet read.
ON_SCREEN_TEXT (post): card, upper third, IN t=0.5s, held through the lapse:
        "WHAT YOUR SHED WEIGHS IN THIRTY DAYS."
        VERBATIM. Do not re-word. "SHED" is the cohort's own noun and is the qualifier.
        PLUS: running day-counter, corner, "DAY 1" to "DAY 30", advancing with the frames.
        The counter FREEZES AT 9 on the exit and holds 4 frames before the cut.
SFX: one tare click per lapse frame, accelerating into a rhythm; documentary bed from t=0
VO DIRECTION: second person on sentence one, then first person from sentence two.
        Even, curious, unhurried. "She stopped reading the number on day nine" is the
        dread beat and is delivered plainly, not landed.
DURATION: 16.6s (donor duration-share basis; 41 spoken words — the 0.4s deficit vs A/B
        sits INSIDE this beat as breathing room; the body boundaries do not shift)

PRIMARY_LAYER: real shoot, procedural macro, hands only
SHOT 1: MACRO locked. Tape pressed down over a mesh screen seating on a shower drain.
SHOT 2: MACRO locked. A thumb hits TARE. The display goes to ZERO.
        (ZERO is the ONE readout allowed to be legible on this board — it is not a measurement.)
SHOT 3: MACRO locked. The screen lifted after the shower, clump intact.
SHOT 4 (give it ~3s): HANDHELD. A child's fingers laying a hundred hairs in a thin row
        on the bathroom counter. Patient, obsessive, slightly funny. Do not rush it.
SHOT 5 (THE RATIO SHOT — the beat's payoff): ONE FRAME, ONE LENS, ONE LIGHT.
        The thin counted row on the left, the drain clump on the right.
        The viewer does the arithmetic with her eyes. NOTHING IS PRINTED.
        NO number, no caption, no overlay, no ratio graphic. This is not negotiable.
SHOT 6: CHART-0 (blank hand-ruled grid, dates down the left) -> CHART-1 (nine rows,
        trace climbing, a ballpoint smudge where a hand rested).
CUTTING: fast, 1.5-2.5s, procedural rhythm
SFX: scale-tare CLICKS and paper texture CARRY THIS BEAT (script direction).
        Pull the music down under the clicks and let the process be audible.
VO DIRECTION: ~184 wpm, brisk and procedural. She is describing method, not selling it.
        Read "four times the hundred" WITH WEIGHT BUT NO THEATRE (script emphasis list).
DURATION: 16.6s (51 spoken words, ~184 wpm)

PRIMARY_LAYER: real shoot, hands + objects, NO FACES
SHOT 1: School gym, fluorescent, chest height so heads are OUT OF FRAME.
        Ribbons, other kids' tri-fold boards, folding tables.
SHOT 2: MACRO. A division ribbon pinned to the corner of CHART-3.
SHOT 3 (THE AD'S FIRST HOLD — give it 4-5s): LOCKED MEDIUM on CHART-3.
        An adult woman's hands at the edge of frame, STILL, NOT TOUCHING IT.
        Hold past comfortable. This is where the ad changes register.
        THE CUTTING SLOWS HERE AND ONLY HERE IN THE FIRST HALF (donor [R]: "slowing on
        the pharmacist's dialogue").
SHOT 4: Retail pharmacy counter, HANDS ONLY. A pill bottle set down. A lanyard.
        NO face, NO white coat framing that implies a face above it, NO credential legible.
SHOT 5: Empty parking lot, flat afternoon, waist height, handheld.
CUTTING: SLOW. 3-4s per shot. This is the ad's held beat.
SFX: gym room tone, then counter ambience. Music HOLDS (does not swell, does not drop).
VO DIRECTION: the register DROPS the moment the judge turns out to be a pharmacist
        (script emotional shift #1). ~189 wpm but the delivery is heavier, not faster.
        "Thirty-one years behind a retail counter." is spoken inventory, flat.
        Read "taking" with weight but no theatre (script emphasis list).
        The final two fragments are the twist — no vocal underline, let them sit.
DURATION: 20.0s (63 spoken words, ~189 wpm)

PRIMARY_LAYER: real shoot, artifact-led, one POV insert
SHOT 1 (the [pause], ~1.5s of picture BEFORE the first word):
        STATIC HOLD on CHART-3 in a domestic room. NO camera movement. NO cut.
        Let the board sit there in silence. This is the hold the script bought.
SHOT 2: SLOW DRIFT across the poster board — the mounted chart, the ribbon, the grid.
SHOT 3: MOTHER'S POV, bathroom, 6am. Over-the-shoulder, NO FACE, no mirror reflection
        of a face. Doorway, shower, drain. SAME LIGHT AS HOOK A (closes the loop).
SHOT 4 (THE AD'S SIGNATURE FRAME, ~7-8s, THE LONGEST HOLD ON THE BOARD):
        The green-marker hypothesis line FILLS FRAME, in the child's handwriting,
        read on screen as it is spoken:
        "My mom says a hundred hairs a day is normal. I want to know if that's true."
        DIEGETIC — photographed handwriting, NOT a post burn. Do not typeset it.
        Do not add a highlight, a box, an underline or an animation to it.
CUTTING: SLOWEST IN THE FIRST HALF. 4 shots across 30 seconds.
SFX: room tone. Music SWELLS here (donor [R]: the bed swells on the reveal).
VO DIRECTION: the guilt arrives (script emotional shift #2). ~140 wpm, the ad's held pace.
        The hypothesis line is the ONLY thing in the ad the narrator READS rather than TELLS
        — it should sound like she is reading her own indictment.
DURATION: 30.0s (70 spoken words, ~140 wpm) — the ad's LONGEST beat.
        Delivered boundary shifts +0.8s from the opening [pause].

PRIMARY_LAYER: real shoot (retail + household) + 2 x 3D wireframe stabs
SHOT 1: Pharmacy END CAP of biotin bottles. Handheld, moving. Brands ANONYMIZED
        (blur, crop or turn). We disqualify a CATEGORY, never a named competitor.
SHOT 2: A minoxidil box, label anonymized.
SHOT 3: A hand pulling a script from a rack.
SHOT 4: CHART-2, FOUR FRAMES ONLY, on "what your daughter just WEIGHED."
SHOT 5-7: HARD CUTS to the PILLOW, the CAR SEAT, the DRAIN — the SAME three objects
        as Hook B, same framing, now indicting instead of qualifying.
SHOT 8: 3D CHEST WIREFRAME, heart lit. HALF A SECOND. CUT.
SHOT 9: 3D SKULL WIREFRAME, brain lit. HALF A SECOND. CUT.
        Do NOT linger on 8 or 9. A held anatomy shot turns the fear bridge into a lecture.
CUTTING: FASTEST IN THE AD. 10+ shots in 24s. Roughly 1 to 1.5s average.
THE ONE HOLD: the scripted [pause] before "Once you lose forty percent..." is the ONLY
        breath in this beat. Keep it TIGHT (<=0.8s) and let the picture go momentarily still.
AUDIO — THE AD'S LOUDEST EDITORIAL GESTURE:
        MUSIC DROPS OUT COMPLETELY under "they don't grow back."
        MUSIC RETURNS LOUD immediately after. This is scripted and it is not optional.
VO DIRECTION: ~285 wpm. A RUSH. Spoken almost without breath, as reported speech
        tumbling out. The voice tightens by SPEEDING UP, never by getting louder
        (script: "it tightens by speeding up, not by getting louder").
        WEIGHED is the emphasis word. The two closing questions are asked, not shouted.
        ANTI-SELF-NERF: this beat is the ad's whole aggression payload. Nothing here is
        softened, shortened, hedged or trimmed in the edit. The permanence line and the
        heart/brain bridge are load-bearing and stay exactly as written.
DURATION: 24.4s (116 spoken words, ~285 wpm — the ad's dominant peak).
        Delivered boundary shifts +1.6s cumulative.

PRIMARY_LAYER: higgsfield-generate 3D animation (GPT Image 2 keyframes -> Seedance 2.0 i2v)
SHOT 1: The follicle BULB in cross-section, warm, alive, on a near-black plate.
SHOT 2: The BLOOD SUPPLY feeding it. Vessel geometry, unhurried camera.
SHOT 3: ACTIVE T3 ARRIVING as bright pulses travelling the vessel INTO the bulb.
        The bulb BUILDS — visible fibre extending.
SHOT 4: T3 STOPS ARRIVING. The pulses thin, then cease.
SHOT 5 (the scripted [pause] lands HERE, right after "It stops."):
        The bulb STANDS DOWN — it DIMS and STILLS. It does NOT die, collapse, blacken,
        shrivel or explode. The script says "It doesn't DIE. It stops." That is the brief,
        literally. Hold the quiet bulb for the length of the pause.
SHOT 6: CALENDAR WIPE — pages wipe BACK three months. CHART-2 reappears OVER the calendar.
CUTTING: SLOWEST IN THE AD alongside beat 3. Long sustained moves, no montage.
        This is the "deliberately slow on the animation so the brain can process it" beat.
SFX: HEARTBEAT and low WHOOSH under the whole block. Music thins to let it through.
VO DIRECTION: ~127 wpm — the ad's slowest beat except the close. Over the animation,
        unhurried, explanatory. DIE is the emphasis word (script emphasis list).
        "It doesn't DIE. It stops." is the hinge and the [pause] follows it.
        SUSTAINED, NOT METERED: one continuous explanation, no benefit interruptions.
DURATION: 28.9s (61 spoken words, ~127 wpm). Delivered boundary shifts +2.4s cumulative.

PRIMARY_LAYER: higgsfield-generate 3D animation -> real shoot bridge -> product macro
SHOT 1: T4 STACKED AND INERT. Storage. Nothing happening. Same plate as beat 6.
SHOT 2: THE SELENIUM STEP — the switch flips. Visible state change.
SHOT 3: T3 DOCKS INSIDE THE CELL and THE CELL LIGHTS.
SHOT 4 (THE BRIDGE SHOT — HARD TONAL INTERRUPT, ~2s, cut out of the animation entirely):
        A CREASED PARKING STUB held in a woman's hand. A STRANGER'S HANDWRITING
        across the back of it. Real paper, real daylight, handheld, hands only.
        This is the ad's THIRD handwriting artifact (green marker / fridge dates / stub)
        and it is the proof this was SCRIBBLED, NOT SOLD.
        Do NOT make the handwriting legible as a product name — it is a gesture, not a label.
SHOT 5-7: FIRST PRODUCT FOOTAGE IN THE AD. The dropper. The amber liquid.
        Under the tongue. Macro, practical, ONE LIGHT, dark surround.
        This is the only styled lighting on the entire board.
ON_SCREEN_TEXT (post): "Kindled" on the FINAL SENTENCE ONLY. Naming element, not a card.
        The product's own label comes into focus + one clean wordmark.
        NO strapline, NO ingredient roster, NO claim, NO badge.
CUTTING: medium. Animation moves are sustained; the bridge shot is a hard 2s interrupt.
SFX: heartbeat/whoosh carries out of beat 6 and thins. Paper crease on the stub.
VO DIRECTION: ~192 wpm. The bridge sentence is the beat's only long one and is taken
        AT THE SAME CLIP as the mechanism lines around it, NOT slowed into a reveal.
        ON PURPOSE is the emphasis (script emphasis list).
        REFERENT DISCIPLINE: "What SHE wrote on the back of my parking stub" is the
        PHARMACIST, not the daughter. Deliver it as a CALLBACK, with the small warmth
        of a thing kept. The daughter does not appear in this beat.
DURATION: 24.4s (78 spoken words, ~192 wpm). Product BEAT opens at 68.3% — donor parity.

PRIMARY_LAYER: real shoot, household + one hands-only clinical insert
SHOT 1 (THE PAYOFF FRAME — the most important composition on this board, hold it):
        CHART-2 (ragged, HIGH) taped BESIDE CHART-4 (flat, LOW) on the SAME FRIDGE,
        SAME DAYLIGHT, ONE LENS, LOCKED.
        NO numbers legible. NO before/after of a head. NO wipe, NO split screen, NO title.
        The SHAPE DIFFERENCE is the entire claim and it asserts no value.
        CHART-4 is NEVER shown alone and NEVER shown before this beat.
SHOT 2: A doctor's hand writing on a pad. HANDS ONLY. Nothing legible on the pad.
        A third party writing our product down — a gesture, not a testimonial.
SHOT 3 (THE LAST SHOT OF THE STORY, hold it long):
        The girl at the fridge, FROM BEHIND, adding a date to the chart.
        BACK OF HEAD ONLY. Warm late light. Handheld. No turn, no face, ever.
CUTTING: medium-slow. Three shots across 22s. The picture relaxes here.
SFX: room tone. SCORE SWELLS on the two-chart frame and then THINS OUT under the
        last line (script direction).
VO DIRECTION: THE VOICE WARMS FOR THE FIRST AND ONLY TIME (script emotional shift #4).
        ~170 wpm. Read "flattest" with weight but no theatre (script emphasis list).
        The final sentence sits behind [quieter] and ON THE FAR SIDE OF A BREATH — it is
        the softest line in the script. [quieter] is VOICE ONLY; the picture does not change.
        HARD: no regrowth number, no second-person promise, no before/after language.
DURATION: 22.2s (63 spoken words, ~170 wpm)

PRIMARY_LAYER: real shoot, two objects
SHOT 1 (hold it for essentially the whole beat):
        The KINDLED BOTTLE on the bathroom counter NEXT TO THE SCALE.
        The two objects in ONE FRAME for the FIRST time in the ad.
        Locked. Real 6am light. The bottle is ON SCREEN CONTINUOUSLY from the first
        word of the close (this is what gives "it's in stock" a picture, not just a pronoun).
SHOT 2: END CARD. Wordmark on a clean plate.
        NO guarantee badge. NO countdown. NO discount flash. NO urgency stack.
        NO disqualification block. NO slogan. The ad SIMPLY STOPS.
CUTTING: TWO SHOTS. Do not add a third. Do not add a montage. Do not extend.
SFX: room tone. The score has already thinned out and does not return.
VO DIRECTION: ~80 wpm — DELIBERATELY SLOW, four or five words to a breath.
        This matches the donor's own close pace to the whole number.
        Pure sell plus reveal. NO anti-sell, NO hedge, NO softening tail, NO company "we"
        (she is a customer, not the company).
        REFERENT: "if IT's in stock" is KINDLED, named two clauses earlier in the same
        breath group. There is no orphan pronoun in this beat.
DURATION: 16.6s (22 spoken words, ~80 wpm) — THE SMALLEST BEAT IN THE AD.
        DO NOT INFLATE IT. Its smallness is a measured donor property.

• VH1 satisfied on all three variants. Fork 4 strategy = option 3, hook visual asset before any speaker — which on this board is absolute, since there is no speaker at any point and never will be.
• VS5 satisfied on all three. Every hook's qualifying card is in by t≈0.6s and survives the full 5s window (1A and 1C hold to beat end; 1B runs a continuous four-card chain). No hook depends on an asset that disappears early.
• The qualification does not lease itself to the audio on any variant. 1A names the drain clump on the card, 1B's four nouns are the arrest, 1C uses the cohort's own word. A muted viewer is qualified or dismissed by the picture and the card alone, which is the whole point of the v1.2 pass and is carried here verbatim.

production_limitations.md is a sparse skeleton (<3 non-template entries) → the formal YS6 cross-check is silently skipped per the file's own heuristic, and no false-positive flag is raised. Practical production flags applied instead, from avatar_tool_prompt_syntax.md and from the shoot itself:

1. No avatar layer exists on this board. There is no presenter, so there is no Heygen / Soul Cinema / Marketing Studio presenter route, no Soul ID to train, no lip-sync risk and no character-consistency problem for a human. This removes the largest single failure class the other Echo boards carry — and it is why the board's risk concentrates in props and one animation block instead.
2. The 28.9-second continuous 3D follicle animation (beat 6) is quality-critical. Cheap CGI kills a mechanism block. Route to a dedicated generation layer (GPT Image 2 keyframes → Seedance 2.0 i2v), stitch at the four declared shot boundaries, reference-chain the bulb geometry so it reads as one follicle, and hold a licensed medical-stock or commissioned-3D fallback if QC misses. The two half-second wireframe stabs in beat 5 must be rendered from the same build.
3. The scale LCD is the board's highest-risk frame. A careless take holds a legible value and simultaneously closes the ad's open loop and asserts a number the copy refused to assert. Shoot every scale frame three ways (occluded / oblique / cut-early). This applies to Hook A shot 2, Hook C's thirty lapse frames, and beat 2's tare sequence.
4. The chart must be one physical build in five states, made by an actual child's hand or a convincing forgery. It cannot be re-drawn per beat and it cannot be CGI. This is the single largest art-department dependency on the board and it should be built and photographed first, because six of nine beats depend on it.
5. The 30-frame time-lapse (Hook C) is the most shoot-management-heavy ask. It can be captured across one week with graduated clump masses rather than literally thirty mornings, but the tripod mark, the hour and the window must be identical. Fallback: a 12-frame lapse with the counter still running 1 to 30.
6. Anonymization is a shoot discipline. The biotin end cap, the minoxidil box and the amber prescription bottle must be framed so blur or crop is available in post (wordmarks off-centre, out of focus). Retro-fitting anonymization to a centred, sharp wordmark is expensive and usually visible.
7. Product legibility and the sublingual action (beat 7). The wordmark should be added in post rather than relying on macro focus; the hand-to-mouth dosing action stays real-shoot regardless of where the rest of the product footage is routed.
8. Runtime is the binding constraint. Delivered A 202.2s · B 202.2s · C 201.8s against a 207s band top gives 4.8s / 4.8s / 5.2s of headroom — the tightest in the batch. Editors must not pad, must keep all three holds at ≤0.8s, and must add no fourth pause. The 240s ceiling is not in play (37.8s of headroom).
9. The gym location (beat 3) is the one non-household location and the most likely budget casualty. Fallback boarded: the poster board against a folding table and a cinder-block wall with gym ambience.
10. Real shed hair, not fibre. Synthetic reads wrong at macro in daylight. This is a collection task with a lead time and it should be scheduled before the shoot, not on it.

Art department, built first: the chart in five states (CHART-0 blank grid + dates · CHART-1 nine rows · CHART-2 complete October, fridge-taped · CHART-3 mounted on a tri-fold poster board with the green-marker hypothesis and a division ribbon · CHART-4 the April chart on newer paper) — one physical build, one hand, one grid.

Practical props: taped mesh screen ×2 (clean and loaded) · kitchen scale · genuine shed-hair clumps at graduated masses (~30 for the lapse) · a counted hundred hairs dressed in a thin row · a creased parking stub with a stranger's handwriting on the back · a division ribbon · the Kindled bottle and dropper.

Anonymized category items: a biotin end cap · a minoxidil box · an amber prescription pill bottle · a pharmacy script rack.

Locations: one bathroom (6am window) · one kitchen with the fridge (late afternoon) · one bedroom (pillow) · one car (headrest, daylight through glass) · one school gym or the cinder-block fallback · one retail counter (two feet of it) · one parking lot · one clinical desk (pad and hand).

Animation builds: the follicle bulb + blood supply + T3 pulses + stand-down (beat 6, six moves, ~29s) · T4 inert + selenium switch + T3 docking + cell lighting (beat 7, three moves) · a chest wireframe stab and a skull wireframe stab (beat 5, half a second each, same build).

Post graphics: three hook cards (verbatim strings only) · Hook C's day-counter 1 to 30 with a freeze at 9 · the calendar wipe (beat 6) · the Kindled wordmark (beat 7 and the end card). Nothing else is burned anywhere in this ad.

Per-beat specs are collated in the boards above. Layer stack, for handoff:

There is no avatar-tool prompt on this board in the literal sense — the Avatar-tool prompt fields carry real-shoot and asset specs because the standing order removes the presenter layer entirely. This is declared rather than silently substituted.

• VS1 (B-roll density) — CONFIRMED-INTENTIONAL, donor grammar. The board runs 1.5 to 2.5s cuts throughout and peaks at roughly 1s at beat 5, which is far above the 1-per-10s ceiling. This is the donor's measured rhythm ([R]: cuts every 1.5 to 2.5s, slowing on the pharmacist's dialogue) and in a D09 the "B-roll" is the ad — there is no talking head for the flag's underlying assumption to apply to. Donor-faithful, not drift.
• VS2 (on-screen text under 12 words) — 2 flags, both dispositioned, both carried verbatim. (a) Hook A's card is exactly 12 words, i.e. at the ceiling rather than under it. It is not re-authored: the v1.2 qualify-first pass wrote it deliberately at that length to carry both the SYMPTOM qualifier and the dread clause, and the standing directive is to carry the cards verbatim and not "improve" them. Mitigated by layout only (two lines, two beats of reading). (b) The Reveal beat's hypothesis line is 16 words. VS2 does not govern it: it is diegetic photographed handwriting, not a burn; it is read aloud in sync as it appears; and it is held for 7 to 8 seconds, which is triple the reading time a 16-word line needs. Logged, not changed.
• VS3 / VS3b (per-beat and per-cue citation) — satisfied. Every beat cites an anchor or logs an original-direction rationale, and every distinct B-roll cue is cited independently. The two original-direction cues on the board are declared: the half-second wireframe stabs at beat 5 (no anchor uses a sub-second anatomical stab; calibrated so the fear bridge stays a thought rather than a lecture) and the parking-stub bridge shot at beat 7 (no anchor supplies it; calibrated to make the acquisition visible and to complete the handwriting through-line).
• VS4 (transition consistency) — satisfied. Hard cuts throughout, with exactly two soft transitions, both at zone boundaries (beat 3 → 4 and beat 6 → 7). No mixed grammar within any section.
• VS5 (mute-test asset survives the full 5s) — satisfied on all three hooks. See the mute-test table.
• VS6 (anchor best-fit cross-adaptation) — 5 logged, all with rationale. (1) Anchor B at beat 7, declared best-fit Hook 1B — used secondarily for the product macro because it is the only anchor in the set showing a D09 hero macro doing persuasive work; its compilation sub-mode and struggle/liberation contrast are rejected. (2) Anchor C at beat 6, declared best-fit Hook 1C — used secondarily for the animation-hold pacing rule; its week-numbered ladder and scarcity close are not imported. (3) Anchor D at beats 3 and 5 — its entire credentialed-presenter apparatus is rejected because this board has no face; carried at mechanism-and-register level only. (4) Anchor E at beat 2 — its on-camera presenter is explicitly not imported per the research brief's own ruling; carried at instrument-grammar level only. (5) Anchor A informs every beat by design (SEED full-pattern borrow), which is not a mismatch but is recorded for completeness.
• Anchor-pool composition — declared. Three of five anchors are this batch's own donors. That is a direct consequence of a pinned, all-SEED, D09-tripled run and it is stated rather than hidden. It is also, on this particular board, an advantage: it means three of the five anchors are measured D09s, which is the format precedent this account has never had.
• Runtime headroom — the batch's tightest, restated. 4.8s on A and B against the band top. Flagged here as well as in the pause note because it is the one number an editor can accidentally spend.
• Review S2 (the cooled AS-IS permanence line) — NOT compensated for visually. The board does not add aggression to cover a copy call the human owns. Beat 5 is boarded at the copy's own voltage and no higher, which is the correct behaviour: a storyboard may never quietly re-litigate a copy decision.

Concept ID: V2 · CID: KD_C0090 Production Target: HYBRID — real-shoot household prop spine + higgsfield-generate 3D follicle animation + practical product macro + post graphics. NO avatar/presenter layer (standing order). Visual Style: donor-derived kid's lab-notebook forensic home documentary (D09, prop is the co-star) Comp Library Anchors: 1238484691057182 (DONOR/SEED, pinned CORNERSTONE), 814438051152496, 26907486858952165, 26839214942445844, 1508278883991064 — all video, zero static demotions Mute-test: PASS (all 3 hook variants; all 3 qualifying cards carried VERBATIM from script v1.2) Runtime: delivered A 202.2s / B 202.2s / C 201.8s vs band 153-207s (headroom 4.8/4.8/5.2s) and ceiling 240s (headroom 37.8s) Production-limitation flags: 10 (no-avatar-layer simplification, animation QC, scale-LCD legibility, one-build chart, time-lapse registration, anonymization, product legibility, runtime headroom, gym location, real shed hair)

**Adapted-convention / compat skips (flagged, not silently dropped):** Beat 1A/1B/1C hook variants · integer contiguous cumulative timecodes on a donor-duration-share basis with a separate delivered column · CID KD_C0090 pre-allocated at storyboard (E3 / S2 / S4 convention) · donor PINNED as anchor cornerstone with **no rotation re-query** · nine-beat discipline preserved · 3D animation routed via the layer stack · **SKIPPED per Echo convention:** rotation-log append (§2g), concept_pipeline append (7c), library_changes style-family log, Notion / board / publish (§7d) — handled at package · save path redirected to the Echo session tree.

---

1. The D09 no-presenter branch needs its own named path in the skill. Every one of the v2 skill's Step-1 Fork-1 options assumes a performer exists; four of the six are avatar-tool routes. A D09 board with a standing no-presenter order has to reinterpret the per-beat Avatar-tool prompt field as a real-shoot / asset spec, which this board does explicitly. Three Echo boards have now done this by hand (S2 partially, S4 partially, V2 completely). Recommend a declared VO-only / prop-led production target in Fork 1, with the per-beat field renamed or dual-labelled so it stops reading as a missing avatar prompt.
2. A measured-artifact concept needs a first-class artifact-continuity surface. The chart is on screen in six of nine beats, ages across five states, and carries the ad's entire proof. Neither the v2 output schema nor the Echo precedent has a slot for that, so this board invented one (the chart state-by-state table plus a per-beat Chart state field). Recommend promoting an artifact-continuity table to a named optional section for any D09 or prop-led board — a co-star with no continuity spec becomes B-roll by the second week of production.
3. The Withheld-Number Law is a reusable pattern, and it is where two house rules meet. "No invented numbers" (copy-side, product_facts §6) and "the withheld number is the open loop" (craft-side) turned out to impose the same picture constraint. The rule generalizes: any measured-artifact ad should board its instrument so that no value is ever legible, because a legible value is simultaneously a fabricated claim and a closed loop. Worth writing into the storyboard checklist as a named check for prop-demo boards.
4. All-SEED pinned runs concentrate the anchor pool on the batch's own donors. Three of five anchors here are this run's donors. Nothing is wrong with the boards that result — arguably they are better, since every anchor is a measured format match — but the rotation and diversity machinery in §2a.0 / §2c has no meaning on a pinned run and is being skipped by hand each time. Recommend the Echo branch state this explicitly: on a PINNED run, rotation is inapplicable, the anchor set is the donor plus the concept's own cited rows, and the diversity note is a declaration rather than a check.
5. Hook-beat duration on a donor-share basis removes the per-variant timecode shift. The E3 and S2 boards both carried per-variant body shifts (−1s, −3s) because their hook beats were word-derived. This board's hook beat is the donor's duration-share, so a one-word difference between hooks is absorbed inside the hook and the body boundaries are identical on all three cuts. That is strictly better for an editor and strictly better for the parser. Recommend it as the default rule for Echo boards whose scripts derive beat durations from donor shares.