“Put up with it”
Cold open on what her doctor said, then she does the opposite — and never argues back. The payoff is the dress, which is a clothes receipt rather than a body claim, and lands self-directed.
Two builds. Build A is the cheap volume test in the wall-of-text format — the viewer sees herself, not an aspiration, which is that format’s whole engine. Build B is the arc: black-screen open, montage, dress. Run A first at ~$1 a creative to find the winning text, then promote the winner into B.
Build A · wall-of-text · ~95 words, static, over a looping 6s clip
Caption: there is no secret, sorry. #over50 #menopausefitness · Background clip: Roxie, coral tee + slate joggers, living room, catching her breath after a set. Ordinary, 480p, unpolished. Not a body shot — this build is the mirror, not the aspiration.
Build B · the arc · revised again 2026-09-03
Roughly 19s. The rhetorical spine is a closed loop — the taunt, the refusal, the answer — and it is what lets the images stay wordless in between.
- Her, in the morning. 3s. Bedroom or hallway, ordinary light. Fitted tee, side profile or three-quarter. Her middle reads through the clothing — hand resting on it, or she simply stands and looks. Not posed, no hem lift, no bare midsection. The difference between a woman in her clothes and a “before” image is entirely in that choice.“You’re 51. You just have to put up with it.”
- Cut to black. 1.5s. Silent. Three words, centred, no motion.“who said so?”
This card does double duty — it is the refusal, and it retroactively frames shot 01’s line as a quoted taunt rather than the ad addressing the viewer. The personal-attributes risk resolves itself. - Hard cut to motion. Living room. 6s across three or four cuts. Sound comes up hard out of the silence — that jump is the engine of the whole ad. Real effort, slightly out of breath, no equipment in frame. Effort, not progress — do not slim her across the cuts.no text
- Neighbourhood, 2s. Optional. Morning walk with purpose, handheld or arm’s-length per the walkingwithnat play. Cut this first if the edit runs long.no text
- Bedroom mirror. The dress. 4s. Dressed for the night out, already in it — no getting-ready montage, no zip-up. She turns, checks the fit, one small adjustment at the waist. She likes what she sees and does not perform it. Hold longer than feels comfortable.no text — the shot is the payoff
- She picks up her bag and looks back once. 2.5s. The line that answers shot 02 and closes the loop.“still 51. still not putting up with it.”
Alternate close, if the night-out staging reads as dressing for someone: “i didn’t do this for tonight. tonight’s just when i noticed.” Warmer, more self-directed, but it drops the loop back to shot 02 — so prefer the primary unless testing says otherwise.
One line to check before this runs
“After 50 the thing you’re actually losing is muscle, and eating less takes more of it” is the closest thing in this flight to a physiological claim. It is directionally supported — the 101-RCT meta-analysis shows resistance training adds fat-free mass while cutting fat — but “menopause-specific science” framing without menopause-specific trial evidence is an upheld-ruling category with the ASA. Get the exact wording checked, or soften to “muscle is what goes first after 50” and drop the mechanism.
What the Meta risk actually is, stated accurately
An earlier version of this brief said a belly opening was “banned outright.” That is true of paired static before/after images. For a narrative video it is a judgment call at review — fitness advertisers run transformation montages on Meta constantly, and they are rejected inconsistently. Treat it as a rejection risk to manage, not a wall.
Four shooting choices carry most of the mitigation: clothed with no hem lift in shot 01, so it reads as a woman rather than a before-image; effort not progress through the montage, so it is a story rather than documented change; no time-stamp card between the two states, which is what converts two shots into a proof claim; and the waistband alternate shot in the same generation session, so one swap converts the whole ad to the safe version if it is rejected.
Run the belly cut on TikTok regardless — its policy differs, and the organic-post-plus-Promote model this format is built around is TikTok-native anyway.
The waistband alternate — keep it as the fallback
Shot 01 replaced by a close on hands at a fastening that will not do up, shot from behind or above the waist. The “before” then sits on a garment rather than a body, so the ad opens and closes on the same kind of evidence — a clothes receipt, which is the most believable proof format in the research. It is also worth testing on its own merits, not just as insurance: she has a specific memory of the button that stopped doing up, and the believability corpus is blunt about the alternative — “I hate being shown before/after pics in order to sell me something.”
Why the closing line is a refusal, not a feeling
Her own words are “I want to feel sexy, and sensual… mostly for myself” — and a night out reintroduces exactly the reading that undercuts it: that she got in shape for an occasion, or for someone at it. The fix is to divide the labour. Let the image carry how she looks, and make the words carry the refusal. “Still 51, still not putting up with it” answers the black card, closes the loop back to the opening taunt, and never makes a claim about her body — so the dress can be as good as it likes without the ad promising anything.
The dress is the safer beat, not the riskier one
A dress reveal feels like a bigger claim than a midriff shot but is the more defensible of the two. “The dress fits” is checkable and sits explicitly on the believable-at-12-weeks list, alongside strength gains and energy. A midriff is a body claim, which invites the forensic reading she applies to every transformation. Keep the dress fitted at the waist and ordinary everywhere else — a gown tips it into spectacle and loses the receipt.
“The quiet house”
The empty-nest pivot. Note this is a separate storyline from the frail-mother one, not a variant — there, she fears becoming her mother; here, she is the mother, and the job has just ended.
Shot list
- Open on the empty room. A doorway into a bedroom that is too tidy. No person in frame for the first two seconds. Let the quiet land before she speaks.
- Roxie, seated, talking head. Living room, sage loungewear, afternoon light. Four-dollars recipe — measured delivery, word-level pauses written into the prompt.
- Cutaway on “so I started training.” Mid-session, living room, unglamorous. Two seconds, no more.
- Back to talking head for the last two lines. The turn on “I'm just me again” needs her face doing the work, so hold it.
- “Turns out she was still in there” — land it, then cut to the end card on the beat. Do not let it breathe; the cut is the punctuation.
Why “not to lose weight” is load-bearing
It is the costly-honesty move — the ad conceding the thing an ad would never concede. It also flips the category’s default: decades of female fitness marketing was built on loss (eat less, shrink, snap back), and the research is unambiguous that gain language outperforms loss language with this audience. “Strong enough to hold everything up” is the gain framing doing its job, and it arrives without a single number attached, which is what keeps the whole script compliance-free.
“Pick her up”
The strongest of the three, and the only one where a before/after is legal to build — because it demonstrates a capability, not a weight loss.
Shot list
- Low angle, from child height. Small arms reaching up into frame. We never see a face. Roxie crouches, gets her hands under the child’s arms.“week 1”
- She can’t. Half a lift, then she sets her down and straightens up slowly, hand going to her lower back. This is the whole ad — shoot it until it is honest, not dramatic.“she’s two. she wants up. every single time.”
- Living room, mid-session. Three fast cuts, unglamorous, no equipment. Sweat, effort, ordinary clothes.no text
- Same doorway, same light, same framing as shot 01. Matching the frame is what makes it read as evidence rather than as two clips.“week 12”
- She scoops her up cleanly and spins. One motion, no strain on her face. Sets her down.no text — the shot is the claim
- Garden or hallway, from behind. Small legs running ahead, Roxie following, keeping up. Motion blur, handheld.“muscle is what decides whether i can still do this when she’s ten.”
Why this one is safe when a body before/after is not
Meta’s prohibition is specifically on weight-loss before/after imagery. What shots 01 and 05 pair is a capability: she could not lift a child, and now she can. That is a functional demonstration, which is the format the research says she actually believes — process receipts over outcome photos, objective and checkable rather than a silhouette she has to take on faith.
It also stands on the strongest evidence in the whole corpus. Strength gains of 20–30% in 12 weeks for women 50+ come from a 101-RCT meta-analysis of 5,697 postmenopausal women, and the falls and functional-independence evidence is high-certainty Cochrane across 108 trials. Nothing else in this flight is that well substantiated.
“Same weight”
Added after the belly finding. On the entry metric that actually measures scroll-stopping — views per video across 279 titles — belly beats strength framing 4×. This is the belly hook nobody in the category is running.
Everyone runs “lose belly fat.” Her own words are “it’s not just the weight gain, it’s WHERE I’m gaining it” — she does not experience this as getting fatter, she experiences it as her body rearranging itself. Fuse that with the one green-rated claim that absolves her of the scale and the ad writes itself.
Build A · wall-of-text · ~100 words, static, over a looping 6s clip
Caption: nobody tells you it moves. #over50 #midlife · Background clip: Roxie, coral tee + slate joggers, kitchen or hallway, ordinary. 480p. No body shot — the text is doing all the work and a midriff reveal here would undercut the confessional tone.
Build B · the literal winning formula · belly + minutes + low barrier + over 50
The corpus states the formula outright, and three of its four components are the top three rows of the title-promise table. You already have the chair-workout asset adjacent to this.
- Roxie on a dining chair, side on. Ordinary room, ordinary clothes, feet flat. Nothing athletic about the staging.“10 minutes. Seated.”
- She starts moving. Controlled, unglamorous, clearly doable by someone who has not exercised in years.“for the part of you that changed at 50.”
- Two or three cuts of the same session. Different angles, same chair. No floor work, no equipment in frame.“no floor. no equipment. nothing that hurts your knees.”
- Hands at her waistband, checking the fit. Not a reveal — a measurement. Clothed, matter-of-fact.“the scale won’t show this. your waistband will.”
The phrase to keep out of every variant
“Menopause belly” fails twice over. Compliance: the ASA banned MenoDaily partly for treating bloating as a menopause symptom — naming a symptom and selling a fix is a treatment claim, and even a product name can imply one. Performance: the scale player keeps “menopause” out of every one of its 15 longest-running ads, because the word drops you into a polluted auction and publicly tags a stigmatised identity.
The contract this hook creates
An age-specific promise is a contract — deliver generic content afterwards and she calls it fraud. A belly promise binds the same way, and it is honourable here only because the product genuinely does move visceral fat. But the funnel has to carry it through: the landing page and week 1 need to speak to her middle specifically. Opening on belly and then pivoting to full-body strength on arrival is the bait-and-switch she is primed to spot.
The child rules
Ad 03 needs a child on camera, which is the one genuinely hard production problem in this flight. Generated children are a moderation minefield, and your charsheets already show Higgsfield refusing on far milder framing than this.
- Never a clear face. Not cropped-at-the-last-second — composed from the start so the face is never the subject.
- The four safe framings: small arms reaching up into frame from below; the child from behind; a low angle shot from child height with only the lower body in view; legs running ahead, motion-blurred.
- The lift itself: shoot from behind Roxie’s shoulder, so what reads is her effort and the weight, not the child’s identity.
- This matches the register you already have. The no-face rule covers both the abs mirror-check and the walking-selfie plays — the child simply extends it, so nothing here fights the established look.
- If generation keeps refusing: the fallback is implication — she crouches, opens her arms toward camera, and we cut on contact. Weaker, but it survives moderation and still carries the beat.
Do not attempt a workaround
If a model refuses on child imagery, that is the correct behaviour and not a prompt-engineering problem. Take the implication fallback rather than looking for phrasing that gets around the refusal. A child in a paid ad also carries real usage-rights exposure that a generated adult character does not — worth a decision from you on whether this ad ships with a child at all, or ships as the implication cut.
Order and budget
Revised 2026-09-03 after the belly finding. Cheapest learning first, most expensive production last — and the strongest entry hook moved to the front, ahead of what had been the lead.
- Ad 04, Build A first — ahead of everything else. Belly is the strongest entry hook in the corpus (1.31M views/video against strength framing's 0.32M), it is green-rated on evidence, and the relocation angle is unclaimed by any competitor. At ~$1 a creative, run 10–20 text variants before anything expensive is generated.
- Ad 02 in parallel. The talking-head recipe is locked, no new charsheet, no wardrobe, no compliance sign-off — the lowest-friction thing on the list. It can be in market while the belly copy is still learning.
- Ad 01, Build A next. The GP hook sits mid-table on the entry metric (“what nobody tells you” runs 0.46M/video), but comment-engagement data backs the dismissal vein hard — so test it cheaply rather than skipping it. Promote a winner into Build B.
- Ad 04 Build B and Ad 03 last. The seated build reuses existing chair-workout staging so it is cheap to add once Build A has a winning line. Ad 03 stays last because the child problem needs solving first, and because it opens the 55+ track, which the research says to run after restoration has found its audience.
What this flight deliberately leaves out
All three of these are restoration-track ads except Ad 03. That is intentional — the research puts the bulk of a 50+ Meta audience at the distress peak between 48 and 54, where purchase urgency is highest. The dismissal vein still has a second, higher-ceiling ad in it (the “too old to be promoted” story, 13,000 likes) and the untested wedges — the week-13 honesty ad and the cognition hook — are both still unbuilt. Those are the second flight, not omissions.