Liverty — the rebuilt shot plan

Our shot for every one of the donor’s 98 shots, and what it costs to make.

98shots
2.4saverage shot
47hard cuts
5people on camera
$2.52to generate
$14.65the old plan
The rule every row follows The picture acts out the line. Back pain becomes eczema, the spot moves from hip and lower back to hands, neck and inner elbow, back braces and mattresses become ointments and allergy panels, the orthopedic surgeon becomes a dermatologist, and the spine animation becomes a skin cross-section.
What it replaces 27 takes of one woman talking in one hallway, with one setting and no cuts, overlays or cutaways. That plan cost $14.65 and would have been four and a half minutes of a static face.

takes T01–T03, $1.73 spent

Where the money goes

sourcewhat it isshotsscreen timecost
ownfootage we hold2361s$0.00
postbuilt in the edit2339s$0.00
productour bottle photography1531s$0.00
genSeedance 2.0 mini2864s$2.22
talkInfiniteTalk, our GPU943s$0.30

The brick library already holds 529 scratching clips, 1,512 of hands and arms, 420 in bed, 508 in a kitchen, 418 outdoors and 1,142 showing a bottle, so the 23 shots marked own are picked rather than paid for.

Every shot

#atsecwhat the donor showswhat we show sourcecost
00s3.2hook title bar over her lying in bed, hand on her hiphook title bar over her lying in bed at night, fingers dug into the back of one handgen$0.11
13s5.0same frame, red arrow pointing at the massager on the bedsame frame, red arrow pointing at the Liverty dropper on the nightstandpost—
28s1.2her bending into the washing machine, hand going to her lower backher bending into the washing machine, stopping to scratch the back of one handown—
39s1.5her on the bed, orange glow burned over the hipher on the bed, orange glow burned over the back of one handpost—
411s1.6her sitting on the bed, both hands pressed to her lower backher sitting on the bed, rubbing the side of her neckown—
512s3.2her at the kitchen counter, one hand on her backher at the kitchen sink, shaking water off raw handsown—
616s1.7her lowering herself into an armchair, wincingher under a lamp turning one hand over, looking at itown—
717s1.8her at a laptop, searchingher at a laptop, searchingown—
819s0.8her at the laptop, closer, reading resultsher at the laptop, closer, reading resultsown—
920s0.9a torso in a back bracea tub of thick ointment being scooped with two fingersgen$0.03
1021s1.3her stretching on a rugher reading the back of a food packet at the counter, elimination dietgen$0.04
1122s1.8her lying on a bare expensive mattressher forearm gridded with allergy-panel scratch marks at a clinicgen$0.06
1224s1.5her in leggings, pink circle highlight on the hipher in a t-shirt, pink circle highlight on the inner elbowpost—
1325s3.5a chiropractor adjusting a patienta gloved hand squeezing a steroid tube onto a patch of raw skingen$0.12
1429s1.8green bank notification overlays, Amount $739,468green bank notification overlays, pharmacy and dermatology chargespost—
1531s4.5her talking to camera, orange topher talking to cameratalk$0.03
1635s6.2her lying on the blue mattress, pain spreadingher hands held out over a sink, cracked across bothown—
1741s1.5her at a work desk, hand on her backher at a work desk, stopping typing to scratchown—
1843s3.5her walking outdoors in a skirt, mobility goneher pulling a long sleeve down over her hand outside the front doorgen$0.12
1946s5.0her in bed in pink pyjamas, day after dayher in bed at night, scratching in her sleepown—
2051s1.2her lying on the sofa, defeatedher lying on the sofa, defeatedown—
2152s4.5her in bed, thinking about the expensive optionher in bed, thinking about the expensive optionown—
2257s2.5the orthopedic surgeon, glasses, monitors behind himthe dermatologist, glasses, monitors behind himgen$0.09
2360s3.6the surgeon holding up x-ray filmsthe dermatologist holding up clinical skin photographsgen$0.13
2463s3.4a hand pointing at a spine x-ray on screena hand pointing at a magnified skin image on screengen$0.12
2566s0.1white flash transitionwhite flash transitionpost—
2667s1.9the surgeon at his desk, leaning inthe dermatologist at his desk, leaning ingen$0.07
2768s11.4her talking to cameraher talking to cameratalk$0.08
2880s4.93D animation, vertebrae pressing together3D animation, the skin barrier thinning layer by layergen$0.17
2985s2.9split screen, her on the left, spine animation on the rightsplit screen, her on the left, skin animation on the rightpost—
3088s2.0her standing, hand on her back, serious conditionsher standing, both hands held away from her sidesown—
3190s1.7the surgeon holding a spine model, it is reversiblethe dermatologist holding a skin cross-section model, it is reversiblegen$0.06
3291s2.0a practitioner working on a back with an anatomy overlaya practitioner examining a forearm with a skin-layer overlaygen$0.07
3393s4.83D vertebra, orange nerve highlight, absorbing nutrients3D skin cross-section, orange highlight on the inflamed layergen$0.17
3498s4.5her talking to camera, I chuckledher talking to cameratalk$0.03
35103s0.1white flash transitionwhite flash transitionpost—
36103s1.2dark frame, card: NODDING AND UNDERSTANDINGdark frame, card: NODDING AND UNDERSTANDINGpost—
37104s7.3her talking to camera, he suggested one last thingher talking to cameratalk$0.05
38111s2.2hand pointing at the massager standing on the floorhand holding the Liverty bottle up to the lensproduct—
39113s1.8the massager on a mat beside her, proven treatmentsthe three botanicals laid beside the bottle on a counterproduct—
40115s1.9a white anatomical torso with the device against the spinea skin cross-section graphic with the three botanicals labelledpost—
41117s1.7her lying on the floor on the device, several of his patientsher taking the dropper under the tongue on the sofagen$0.06
42119s1.8a wrist and forearm resting, amazing resultsa forearm, clear skin, resting on a cushionown—
43121s1.0her from behind, lying on the deviceher from behind at the bathroom mirror, looking at her neckown—
44122s1.0her talking to camera in an orange-red top, I was skepticalher talking to cameratalk$0.01
45123s1.0a man in a white coat giving a thumbs upcut: our offer carries no guarantee, so the beat drops to a stock badge onlypost—
46124s1.8hands holding the device, 90-day money-back badgehands holding the Liverty bottle, 4.9 and 604 reviews badgepost—
47125s3.1her lying on the device on the floor, decided to try ither setting the bottle on the nightstandproduct—
48128s2.8black and white split, already wastedblack and white split, already wastedpost—
49131s0.1the device coming out of its boxthe Liverty box being openedproduct—
50131s1.3the device lifted out, brand on the boxthe bottle lifted out, brand on the boxproduct—
51133s1.3her carrying the device through the house, impressedher carrying the box through the housegen$0.04
52134s0.8her crouching to plug it in, by its qualityher unscrewing the dropper at the kitchen countergen$0.03
53135s3.0her sitting on the floor with the dog, decided to try ither sitting on the sofa with the dog, first dosegen$0.10
54138s2.6her lying on the device, felt real reliefher asleep, hands open and still on the sheetgen$0.09
55140s1.7split screen, her lying down beside a spine animationsplit screen, her asleep beside a skin-layer animationpost—
56142s2.6her on the sofa with a mug and the dog, satisfying feelingher on the sofa with a mug and the dogown—
57145s0.1the device under her lower back on a chairthe bottle beside her on the arm of the chairproduct—
58145s5.4the device working, no words can explain thisthe dropper being filled, closeproduct—
59150s1.8her settling into a chair with the device, gentle yet preciseher taking the dropper at the kitchen tablegen$0.06
60152s1.9the device on a dark surface, soft heatthe bottle on a dark surface, light through the glassproduct—
61154s1.3a timer graphic, 8:42, after 15 minutesa timer graphic over her sleeping, through the nightpost—
62155s5.5her talking to camera, pain dropped from an 8 to a 2her talking to cameratalk$0.04
63161s1.8her lying on the device, with each use my pain decreasedher hands, week by week, less redown—
64163s3.4her standing by a plant, mobility improvedher standing by a plant in short sleevesgen$0.12
65166s1.8red X over black and white chiropractor footagered X over black and white steroid-tube footagepost—
66168s1.5red X over black and white massage footagered X over black and white allergy-panel footagepost—
67169s1.7hands setting up the device at homehands setting the bottle on the bedside tableproduct—
68171s1.5the device on her back, that actually workedthe dropper going under the tonguegen$0.05
69172s1.5her on the floor on the device, three months laterher sitting on the floor stretching, sleeves upgen$0.05
70174s1.5red X over surgery footage, I did not need surgeryred X over an injection-pen graphic, she did not need Dupixentpost—
71175s2.2her talking to camera, I did not just get rid of painher talking to cameratalk$0.01
72178s0.0a blurred office window, I could sit through meetingsa blurred office windowown—
73178s1.2her in a blazer at a meeting table, smilingher in a meeting shaking a colleague's handgen$0.04
74179s1.2her walking outdoors in white shorts, no sharp painher walking outdoors in a short-sleeved topgen$0.04
75180s2.4her waking in bed without fearher waking in bed and looking at clean sheetsgen$0.08
76182s3.8her in front of a mirror, it is not about just gettingher in front of a mirror in a vest topown—
77186s1.9her lying on the floor on the device, it is about consistencythe dropper being taken, same time each nightgen$0.07
78188s1.4the device under her, traction and heat combothe bottle and the three botanicals, steady nightly dropperproduct—
79190s2.3her on the bed with the device under her back, namedher on the bed with the bottle in frame, namedproduct—
80192s2.0a jug pouring into a glass, because when your spine hasa jug pouring into a glassown—
81194s2.5split screen, a body silhouette and a scan, reset and healsplit screen, a body silhouette and a skin-layer graphicpost—
82196s0.7her lying in bed smiling, now meher lying in bed smilingown—
83197s0.5her mother holding the device, my motherher mother holding the Liverty bottlegen$0.02
84198s2.7a forearm being held, my best frienda friend's forearm, clearown—
85200s0.1white card, the product dailywhite card, Liverty Dynamic Drops nightlypost—
86200s2.0her on the bed on the device, the product dailyher on the bed, bottle on the nightstandproduct—
87202s5.0her talking to camera, it is not magicher talking to cameratalk$0.04
88207s3.2empty warehouse shelving, demand is highempty warehouse shelvingown—
89211s1.3OUT OF STOCK tags across a pegboardOUT OF STOCK tags across a pegboardpost—
90212s1.5her talking to camera, but if you are watching thisher talking to cameratalk$0.01
91213s1.8a laptop on a desk showing the product pagea laptop showing the Liverty pagepost—
92215s1.9the laptop closer, if you order by clicking the linkthe laptop closerpost—
93217s1.8hands on the device, SAVE UP TO 60% badgehands on the bottle, third bottle free when you buy two badgepost—
94219s5.3the device under her, your orderthe bottle in hand, your orderproduct—
95224s10.3banknotes being flushed down a toiletbanknotes being flushed down a toiletown—
96234s1.5the device on her back, product namedthe dropper in use, product namedproduct—
97236s1.1the device held up to camera, risk free todaythe bottle held up to camera, take your first dropper tonightproduct—

The first cut

All 98 shots, the whole script, 3 minutes 57 seconds Every shot in the plan is real picture now. The voiceover runs underneath and the captions are burned in the house tiktok style, as the donor does.
How the last 70 were madeMatching them to our 3,546 bricks by keyword returned a boy in a blue shirt for “her hands over a sink” and a makeup palette for “her at the bathroom mirror”, so they were generated on Seedance 2.0 mini instead for $3.90 rather than shipped wrong.
Two women in one adfal ran out of balance, and BytePlus refuses a real person’s photograph, so the last 70 shots use a Seedream cast sheet drawn on our own account. She is consistent with herself and is not the woman in the nine fal takes. Putting the whole ad on one face costs about $1.30 of re-rendering.

The voice needs your ear

What went wrong voice_cast.py measures pitch, pitch spread, voiced share, words a minute, pause rate, rhythm and word errors. It has no measurement of accent at all, and it auditioned four voices out of a catalogue of ninety-five because it reused a cached shortlist. It scored Amina best on those numbers, and that is the voice in the cut above.

Sixteen American-sounding voices auditioned against the donor's narrator on the hook line. The top four sit within 0.06 of each other, which is noise, so the numbers cannot separate them and nothing in the tool can hear an accent. Pick one and I will re-read the whole script in it.

voicepitchscorethe hook line
Veronica178.8 Hz1.131
Kelsey173.9 Hz1.143
Wendy183.9 Hz1.189
Deborah173.9 Hz1.192
Ashley190.5 Hz1.703
Avery136.8 Hz1.462
Amina166.7 Hz0.910

lower score is closer to the donor's narrator; the donor sits at 166.7 Hz

Our own voice bank, which is not what cast this

The names above (Amina, Veronica, Kelsey) are Inworld TTS-2 stock voices, a 95-name catalogue hardcoded inside voice_cast.py and rented through DeepInfra. They are the provider’s, not ours.

Our own belt is ~/Claude/assets/voice-bank/: six voices lifted from winning swipe ads and cloned on Chanjing cicada4.0, driven by tools/voice_lines.py. Five are female. Measured against the donor’s narrator at 166.7 Hz, spread 3.16, 183 wpm, 1.8 pauses:

our voicepitchspreadpace
female-roman170.2 Hz5.32167 wpmclosest of ours on pitch, and the donor’s exact pause rate
female-hamptons190.5 Hz2.17144 wpmthe only one already cloned; reads about 28
female-grant-pants207.8 Hz2.76155 wpmtoo high against the donor’s 166.7 Hz
female-resilia-housekeeper———ripped from a winning ad, cast as a woman of about 40 at home
female-ritz———in the bank, not yet profiled

Only female-hamptons is cloned today; the rest need one clone call each before they can read anything, and the call timed out when I tried it just now.

The voiceover

Cast by measurement, not by ear voice_cast.py measured the donor's own narrator and auditioned stock voices against it. Amina won at 166.7 Hz against the donor's 166.7, pitch spread 2.89 against 3.16. Nobody chose it by listening.
CheckedAll 52 sentences were read, transcribed back and compared with the locked script one at a time. Every one matches. The whole-file transcript drops four clauses at its 30-second seams; an 8-second window across each one finds them all present in the audio.
Length236.9s, which is the donor's runtime to the tenth of a second. 786 words against the donor's 788, so the picture plan's shot times carry straight over.

4 minutes, $0.15 of text to speech

The nine takes we generated

Two frames each, in take order A B C E F G I J K. 28 of the donor's shots are cut out of these.

the nine generated takes

Those 28 shots, cut and run together

64 seconds of picture, cut from the nine takes, with no voiceover, captions or sound. It is not embedded here yet because the review tool fails it on face identity, which is the open defect below. The other 70 shots of the ad come from footage we hold, our bottle photography and the edit.

Face identity, measured

SFace cosine of each take's best frame against the narrator reference. The review tool treats anything below 0.82 as a different person; every labelled control sits above it.

takecosineverdict
T01 talking take0.887pass
J outdoors, plant, meeting0.867pass
G the dropper at home0.769drift
C kitchen, doorstep, parcel0.752drift
I asleep and waking0.593drift
A the hook, in bed at night0.312wrong woman

Four of the five takes that show her drift off her face, and the hook is the worst of them at 0.312, which reads as a different woman. The three talking takes hold because they are chest-up and facing the lens like the reference; the wider and angled framings are where H3 Max rebuilds the face.

The donor, shot by shot

donor shots 0-34 donor shots 35-69 donor shots 70-97