Pants Gap, Same Person, One Move

One winner, one weaker sibling with a better tail, one different job

Read. A is the winner by a mile and the one to build first. B is A's weaker sibling with a better tail (product proof, then a CTA), so its tail gets grafted onto A. C is a different job: it is saved and forwarded (reposts outnumber comments 135 to 1) rather than watched, the same engine as the wall-of-text ads.

"6 months of just walking" — the garment does the measuring

shot

Bedroom, phone propped at chest height, one framing for both halves

talent

Young woman (~25), brown crop tee, grey wide sweatpants, hair up

caption

6 months of
just walking   fixed, centred, ~62% frame height, whole video

audio

Pop track; "I've been waiting all night for this moment" lands on the cut

cuts

4.47s (before → after, hard, on the beat) · 7.93s (→ end card)

Why it works

• Same room, same clothes, one cut. The sameness is the proof of "same person"; the cut is the whole edit.
• "Just walking" is the lowest barrier in the corpus (no-equipment promises average 1.53M views per video, top of the table). "6 months" is a timeframe she believes.
• The viewer can check the receipt against her own wardrobe. Nothing is asserted.

What does not transfer

• The subject is about 25 and the payoff is a large weight loss. Our audience reads a big delta on a young body as "not me" and a big delta on an older body as Ozempic.
• "Just walking" is the wrong method for our proposition. Her words: it's not the weight, it's where it's going. After 50 the fat moves to the middle and it is muscle that moves it back; walking alone is what she has already tried.
• The paired weight-loss reveal is exactly the Meta-prohibited format. A pants gap is a weight-loss image however it is framed.
• A 4.8 laurel we do not have. No fabricated ratings.

Remake A · "6 months of 15 minutes at home"

Roxie, 55, in a real bedroom (bed, dresser, door, morning light), phone propped. Fitted tee and wide-leg soft sweatpants in both halves so the garment can measure. Before-Roxie derived from after-Roxie (house sequencing rule): fuller middle, waistband folded under the stomach, flatter expression, ponytail with roots. After-Roxie at the "great for 55" calibration: soft definition, real skin, not a six-pack.

Production

1. Stills, Nano Banana Pro, ~4cr each. After-frame first from the blush or red charsheet re-dressed in tee and sweatpants (the charsheet is an input to the image model only). Before-frame with the after-frame passed as a second reference for framing only; the body delta written at the scene-frame level relative to clothing fit, never as a number.
2. Gate, 0cr. 50% alpha-blend of before and after. Head size, camera height, door line and bed edge must coincide or the cut reads as two people.
3. Clips, MiniMax H3 start-frame, 6s each, use ~4s. Before mime; A1 gap; A2 floor rise. The floor rise is a full-body arc, so budget Kling non-turbo with start and end frames as the fallback, two attempts.
4. Post. Pick our track first and place the cut on its drop. Caption burned in Avenir Black, white with a black stroke, at 62% height. Append the end card. 9:16 master plus a 4:5 feed crop.

"same person btw" — profile first, then the receipts

shots

Bathroom (before) → living room (after) → phone over geraniums → App Store

talent

Same woman as A; wet hair and side profile before, blonde and patterned leggings after

caption

same person btw   small, mid-frame, on the before only

audio

Music only, no vocal

cuts

5.87s · 9.17s · 12.67s

B does 5% of A's views. Its hook is weaker and its policy problem is identical, but the tail is the part A is missing.

Remake B · graft the tail onto A, keep a standalone for organic

Zero fabricated numbers: the receipt is sessions done, which is what the product actually counts. No store page, no rating; 50Queen is web-only.

Linda — one move and a testimonial that takes longer to read than the loop

shot

One locked shot, 8.48s. Living room, wood floor, cream ottoman, curtains, framed art; camera low at floor level, side-on

talent

Woman about 60, grey bob, white tank, teal shorts, grey mat, 2 kg dumbbells

move

Glute bridge with a dumbbell chest press: legs to tabletop, hips lift, dumbbells press up; feet come down at the end

overlay

Two white boxes, top third, whole duration (below)

audio

Music, one sung line ("rise up") at 5.2s

The caption sells the rest: "firmer tummy and lifted glutes without squats, try this move for 14 days", a seven-line benefit checklist, "2kg each dumbbell", 12 reps × 3, walk daily, sleep 7–8 hours.

Why it works

• She is the viewer. Grey hair, a real living room, a mat. No transformation, no before. The whole trust budget goes on resemblance.
• The overlay takes about ten seconds to read against an 8.5s loop. Rewatch, watch-time signal. Same engine as the wall-of-text ads.
• "Without squats" removes the knees objection in two words. One move is the lowest possible barrier.
• Reposts 814 against 6 comments: it is saved to do later and sent to a friend. "2kg" and "12 reps × 3" are specificity, and specificity is believability.

What does not transfer

• Dumbbells. Roxie never uses them; a bodyweight version says "no equipment", which is the top-ranked promise in the corpus anyway.
• "Belly is flatter after 3 weeks" is amber to red. Firmness and waist changes are green within 12 weeks; three weeks is not defensible. Move the timeframe to four weeks and the claim to feel and fit.
• The movement library has no supine bridge (nearest is the seated knee-drops), so this is new footage, not a library demo.

Remake C · "4 weeks in"

Roxie on the mulberry mat in the living-room set, cherry-red racerback set (already chosen for today's aliciae6 build, so wardrobe stays continuous across the flight), camera low and side-on like Linda's. Glute bridge with an overhead reach: arms sweep from hips to overhead as the hips lift. Glutes, deep core, shoulders: the same muscle list Linda claims, no dumbbells. Eight to nine seconds, locked, three slow reps.

Claims check: "feel stronger", "middle feels firmer", "clothes sit differently" are green (feel and fit, unquantified). No "flatter", no "belly", no weeks-to-result promise beyond the honest "4 weeks in".

Production

1. Still, ~4cr. Roxie supine on the mat, hips lifted mid-bridge, arms overhead, side-on at floor level. Two-step: character first on a plain ground from the red charsheet, then character plus the living-room plate.
2. Motion. Kling non-turbo with a start frame (bridge down) and end frame (bridge up) gives a clean rep arc; loop the rep three times in post. Kling Motion Control driven by Linda's footage is acceptable for a replication spike only and must be re-driven before spend.
3. Post. The two white boxes drawn in PIL, a Seed Audio instrumental, 9:16 and 4:5.

Policy and claims

• "Menopause belly" never appears. Hooks say "the middle after 50" or "my middle at 55". "Menopause" stays out of cold hooks; it may appear once in a caption body.
• Paired body before/after (A1, standalone B) is Meta-prohibited weight-loss imagery. Those run organic on TikTok and Instagram. Meta gets A2 with B's tail, and C. The one precedent for running a clone on Meta anyway is the 03-belly-clone-control static, run as a labelled control; if Eugene wants the same experiment here, A1 goes in labelled, with its own kill rule and the expectation of a review flag.
• No numbers on the body. No pounds, inches, or "flat". Green wording: feel stronger, middle feels firmer, clothes fit, same weight different body, exercise targets deep belly fat even when the scale barely moves (unquantified).
• Timeframes. Six months on the transformation (A); "4 weeks in" on the feel claim (C). The 12-week program is a product fact on the end card, never a results promise.
• Music. Source tracks for the spike cut only. Paid masters get our Seed Audio bed or a licensed track, cut on its own drop.
• End card claims ("Just 15 minutes a day", "Get my plan now") are already live in the shipped chair-workout ad. Re-verify the quiz URL before flight.

Cost and order

H3 pricing moved from 5 to 58 credits per second inside one day last week. Probe the current rate before the first clip; the table shows both ends.

Order. A first (the 1.5M-view structure, and its before/after pair feeds B). C second (cheapest, fastest, a different job). B's tail is assembled onto A last.

Test plan

Kill rule as before: 50 leads, CPL against control, checkout rate second.

Four decisions before anything is generated