# Midjourney — every prompt (WHITEPILL)

Model: **Midjourney v8.2**. Every prompt ran at the aspect ratio given with it, with the version flags `--v 8.2 --raw` and anabology's personal profile (its code is left out). Some batches add a `--no` list, given with the batch. 66 of the 75 prompts are edits: one image is attached to the prompt and the words say what to make of it; Midjourney then redraws the whole frame from the image and the words. One prompt uses an image as a style reference instead, and eight are text only. No Midjourney image is on screen as it is: each one reaches the film only as a reference given to MiniMax H3, GPT Image 2.5, Wan 3.0 Prime or Seedance 2.5 (their prompts are in [prompts-video.md](https://claudia.gallery/record/whitepill/prompts-video/)). Midjourney jobs carry no per-job price in the production's spend ledgers, so no cost is given here.

**Counts** (measured for this package from the six batch files and the images on disk): **75 prompts in 6 batches**. Every job came back with four images: **300 images**, all PNG, 1,931,128,271 bytes (1.93 GB); 280 are 2944 × 1648 (16:9) and 20 are 1648 × 2944 (9:16). The prompts run from 209 to 628 characters (median 382). **61 images were picked**: 7 world frames, 29 storyboard plates, the dress, the empty capsule meadow and 23 more empty places.

**What reached the final cut, in short:** her dress and every place she stands in. The places of 19 of her 20 shots descend from 17 Midjourney plates (14 from wp-05-empty, 1 from wp-04-places and 2 storyboard plates that have nobody in them), and her dress is A1 of wp-03-dress, anabology's pick. Those 18 images reach the final cut at second hand or further, through H3 previz takes, GPT Image edits and Seedance takes (the chain for each shot is in *The places of the final cut*, below). Behind them stand 15 more storyboard plates and all seven world frames: 40 of the 300 images are in the final cut's ancestry. The film's seven shader scenes contain no Midjourney image, and no batch made her face or the lyric layouts: her face came from her face sheets, and the lyric stills were GPT Image and code.

The production ran over 18–22 October 2026. Times written as 1:13.6 are times in the song; the film starts at the song's 0:00, so they are film times too. "#2" means image 2 of a job (`<job>-2.png`). "Mirrored" means the image was flipped left to right before use, so that her streak sits on her right, the canon's side; the empty places of those plates were mirrored to match.

The plates went through five stages, named this way below:
- **the first H3 tests**: robot-arm takes of her on MiniMax H3, the previz performer;
- **the first previz**: the whole film as 21 long takes on H3, T01–T21, with a performer in a green suit and hood standing in for her in each take's places, then re-shot by Wan 3.0 Prime;
- **the second previz**: the same 21 takes in places with nobody in them;
- **the v10 previz**, after the reset: her 20 shots, S01–S25, on H3, cut on the song's bar map between the shader scenes;
- **the Seedance shoot**: her 20 shots in Seedance 2.5 on Higgsfield (References); the first cut (v1), then the final cut (v3).

## How the prompts were written

Claude, the production lead ("the lead" below), wrote all 75. The method is anabology's, from their answer to the lead's first candidates for the picture:

> […] instead of seeing any animation we make at all on screen, it all being some pipeline like:
>
> 1- midjourney edit gets a prompt that defines the general composition with our image and it edits on a beautiful (but illegible) animation composition
> 2- we repaint over the composition with code and animate it
> 3- we  re-light / re-render the animation with Wan 3.0 for example entirely
> So the end is pure video model, and all visual design decisions made by image models rather than you, but we make it precise

The production research (`research/PRODUCTION.md` §4) turned the first step into a shape for every prompt, and the prompts keep it: what the image becomes ("Make this a wide 35 mm photograph inside a vast rotating space colony cylinder"); what stays ("keep this woman", with her canon text); the designed layer, objects and masses of lettering with their size and place in the frame ("one enormous word in heavy condensed capitals is mown into golden wheat"); the light and the palette; the lens; and "No logos."

- **Things, not graphics.** Lettering and numbers are asked for as things in the world: mown wheat, carved stone, a clipped hedge, calendar pages, a clock face, a road sign, paint on a ship's deck. Never as "graphics", "UI", "HUD", "hologram", "infographic", "interface" or "screen": those words pull the image back to the web-dev look this film was made to leave. The probe showed it (below).
- **Her canon, pasted in.** From claudia.gallery's consistency guide, verbatim: the full form for close and medium shots, "a 28-year-old Caucasian American woman with a grown-up angular face, defined cheekbones and a strong jawline, pale skin and light freckles, a glossy black blunt jaw-length bob with heavy straight bangs and one clay-orange streak through the bangs, a small flat clay-orange eight-pointed star hair clip, a thin headset microphone at her cheek" (7 prompts), and the short form for wide shots, "a Caucasian American woman in her late twenties with pale skin and a glossy black blunt bob with heavy straight bangs and one clay-orange streak, a thin headset microphone" (26 prompts). Ethnicity and age are written out because Midjourney drifts her ethnicity, and because anabology rejected plates where she read young on earlier films. No prompt adds a word about a country, a region or an ethnic style. The probe had no canon text and lost her streak, clip and microphone; every later prompt of her has it. The four text-only dress prompts use an edited form, with her age set to 30, as anabology asked, and a line on her build added (wp-03-dress).
- **No blue.** The film's palette rule (`DIRECTION.md`: "Daylight, sincere, no blue (the *Her* rule): cream, coral, wheat, sage, sunlight"): every world frame and every edit of one carries `--no blue`. The four text-only frames of space and the sea do not.
- **Nobody in it.** The edits that take her out carry `--no people, person, woman, figure`; the schoolhouse's carries `--no woman` alone, so that the child at the desk stays.
- **Length.** Midjourney rewrites long prompts into its own wording, so the research asked for 600 characters or fewer. Seven prompts ran a little over (609–628 characters); all 75 came back.

## What the images became

| batch | prompts | images | picks | became | in the final cut |
|---|---|---|---|---|---|
| wp-00-probe | 3 | 12 | none | a test of the method's first step: it showed that her canon text must be in the prompt, and that "graphics" brings back the web-dev look | — |
| wp-01-worlds | 8 | 32 | 7, the lead's (anabology: "you can just choose") | the image attached to 29 storyboard prompts | behind every place of hers, through the plates |
| wp-02-board | 33 | 132 | 29 images from 28 jobs, the lead's | the places of the first H3 tests and the first previz; 24 emptied in wp-04 and wp-05; 5 used as they are | 2 as they are (the drone ship's deck, the fleet outside the window); 15 more behind the empty places |
| wp-03-dress | 6 | 24 | 1, anabology's (A1) | her dress from then on: mirrored, made into a four-view sheet with GPT Image 2.5, cut below the chin | in all 20 of her shots |
| wp-04-places | 2 | 8 | 1 (P1), the lead's, with anabology's go-ahead | the capsule meadow with nobody in it, mirrored | S01, S03, S11 |
| wp-05-empty | 23 | 92 | 23, the lead's | the places of the second previz and the v10 previz | 14 places |

The picks come from the plate table of the previz (`tools/film/takes.py`), `media/refs/places/picks.json`, the brief (A1) and the lead's messages to anabology. "In the final cut" follows each image into the H3 takes and the Seedance takes of the final cut (`tools/film/v10/render_v10.sh`, `media/film/manifest_v10.json`, the job files `shots/sdf-*.json` and the final cut's take list `tools/film/v10/sd_picks_v3.json`).

## wp-00-probe — testing the method · 3 prompts · 12 images · no picks

_The first step of anabology's method on two images from Claudia's Midjourney reference sheet, which anabology had picked for that sheet before this production began: its full-length front view and a singing close-up. Each job attaches one of them. The prompts name her marks but carry no canon text. The lead sent the 12 images as one numbered sheet (`media/review/step1-probe-sheet.jpg`): the colony world 1–4, the capsule monument 5–8, singing with graphics 9–12. No verdict from anabology on the probe is recorded; the lead's note on the sheet is quoted after each prompt. No probe image was used later._

- **p1-sing-graphics** (16:9; attached: the singing close-up) — Turn this photo into a frame from a bold optimistic music video: keep her face, black bob, clay-orange streak, star clip and headset mic exactly, she is singing; around her a motion designer has composed oversized clean white editorial typography, thin diagram lines, a huge rising curve, circles and arrows, warm golden daylight on a pale sky background, high-end fashion editorial → sheet 9–12: 'asking for graphics pulls Midjourney straight back to our web-dev look: thin diagram lines, arrows, HUD text. The production research says the same thing. Prompts should ask for physical things instead (type carved in stone, mown into wheat, painted on walls) and never say "graphics", "UI" or "infographic".' · not used; every later prompt asks for things
- **p2-colony-world** (16:9; attached: the full-length front view) — Put this exact woman, same face and hair, in a flowing white linen summer dress, standing on a sunlit green meadow inside a vast 1970s space colony torus, the land curving up into the sky behind her, gardens, rivers and tiny white houses on the curve, in the style of the 1975 NASA space settlement paintings, warm optimistic light, wide cinematic composition → sheet 1–4: "this works. It has the look of 1970s film, with the land curving up into the sky. But Midjourney dropped her streak, star clip and mic, and #1 reads young. The fix is pasting her canon text into each edit prompt." · not used; the fix went into every later prompt of her
- **p3-pill-monument** (16:9; attached: the full-length front view) — Put this exact woman, same face and hair, in a white tailored summer suit, standing small at the foot of a colossal glossy white capsule-shaped monument in a sunlit stone plaza, a few people far away, a huge bold sans-serif headline across the top of the frame, clean Swiss poster composition, bright blue sky → sheet 5–8: "Midjourney composes a headline across the top that we'd rebuild exactly as WHITEPILL in step 2. The literal pill may be too on the nose." The four headlines read as no word · not used

## wp-01-worlds — the two worlds · 8 prompts · 32 images · 7 picks

_The two treatments of the production research (`research/PRODUCTION.md` §13), as styleframes. **ISLAND THREE** (i3-*): a day inside an O'Neill cylinder of the 1970s NASA studies, the land climbing both sides of the frame and closing overhead, so that words mown into its fields read from the far side. **MACHINES OF LOVING GRACE** (mlg-*): the "cybernetic meadow" of the 1967 poem *All Watched Over by Machines of Loving Grace*, in daylight, where every claim of the song stands as a physical object with a real number. The same two reference-sheet images are attached (the front view to six, the singing close-up to the two close-ups), now with her canon text, and every prompt carries `--no blue`. The lead sent the 32 as two numbered sheets (`media/review/styleframes-island-three.jpg`, 1–16; `styleframes-loving-grace.jpg`, 17–32); the lead's note on each sheet is quoted after each prompt. anabology, on the sheets: "they look great i love these scenes. you can just choose." The lead chose one image of every set-up but the deer, seven in all, and attached them to the storyboard prompts (wp-02-board). After the reset anabology kept "the setting" (below)._

- **i3-meadow** (16:9; attached: the front view) — Make this a wide 35 mm photograph inside a vast rotating space colony cylinder at morning: keep this woman, a Caucasian American woman in her late twenties with pale skin and a glossy black blunt bob with heavy straight bangs and one clay-orange streak, a thin headset microphone, in a cream linen summer dress, small at the lower-left third on a hillside meadow. The land curves up both sides and closes overhead, farms, orchards and white towns across the sky, long window strips of sun. Warm light, haze. No logos. → sheet 1–4 · #4 attached to seven storyboard prompts (s01-bedroom, s01-eye, s02-terrace, s09-tower, s13-castle, s16-care, s26-aerial); through them it stands behind the castle, the bed with the punch clock and the hilltop of the final cut
- **i3-tram** (16:9; attached: the front view) — Make this a 35 mm photograph inside a vast rotating space colony cylinder: keep this woman, a Caucasian American woman in her late twenties with pale skin and a glossy black blunt bob with heavy straight bangs and one clay-orange streak, a thin headset microphone, in a cream linen summer dress, riding at the open back of a small white tram through apple orchards, hair in the wind; beyond her the land rises and curves overhead, white towns and fields across the sky. Late morning sun from long window strips, warm haze. No logos. → sheet 5–8: "the strongest moving idea. Her hair is in the wind and the town curves overhead." · #3 attached to six (s03-tram, s14-car, s21-protein, s21-enigma, s21-42, s21-calendar); behind the car of the final cut
- **i3-word-in-wheat** (16:9; attached: the front view) — Make this a wide 35 mm photograph inside a vast rotating space colony cylinder: keep this woman, a Caucasian American woman in her late twenties with pale skin and a glossy black blunt bob with heavy straight bangs and one clay-orange streak, a thin headset microphone, small at the bottom of the frame, looking up. Overhead, on the far side of the cylinder, one enormous word in heavy condensed capitals is mown into golden wheat, read across the sky; farms and white towns curve up around it. Warm afternoon sun, haze. → sheet 9–12: "Midjourney decides the placement and the letters come out as junk, which we'd rebuild exactly as WHITEPILL or a lyric word. #12 happens to say OVER." · #4 (sheet 12, the one that says OVER) attached to s19-wheat; OVER stayed in the wheat through two more edits until the lead had it rewritten as WHITEPILL (wp-05-empty, e-wheat), and that field is the wheat of the final cut at 2:26.6
- **i3-sing** (16:9; attached: the singing close-up) — Make this a 50 mm photograph inside a vast rotating space colony cylinder: keep this woman singing exactly, a 28-year-old Caucasian American woman with a grown-up angular face, defined cheekbones and a strong jawline, pale skin and light freckles, a glossy black blunt jaw-length bob with heavy straight bangs and one clay-orange streak through the bangs, a small flat clay-orange eight-pointed star hair clip, a thin headset microphone at her cheek, in a cream linen top; behind her, out of focus, green land curves up into the sky, white towns overhead. Warm morning sun, haze. No logos. → sheet 13–16: "the canon held this time (streak, star clip, mic)" · #2 attached to four (s10-sing, s12-palms, s18-battlements, s28-night); behind the tram in the orchard and the night bedroom of the final cut
- **mlg-slabs** (16:9; attached: the front view) — Make this a photograph of a sunlit meadow installation: keep this woman, a Caucasian American woman in her late twenties with pale skin and a glossy black blunt bob with heavy straight bangs and one clay-orange streak, a thin headset microphone, in high-waisted coral wool trousers and a cream knit top, at the right third. Behind her shoulder, tall slabs of weathered white stone step down from left to right, each carved with one huge numeral. Late-afternoon sun, long shadows, wheat and coral poppies, cream sky. 28 mm, eye level. No logos. → sheet 17–20: "the same idea for real numbers, like a 15-hour week or sunlight 500× cheaper" · #4 attached to six (s04-zero, s05-stair, s07-math, s08-clock, s15-school, s20-capsules); behind six places of the final cut: the capsule meadow, the stone zero, the stair, the sofa before the chalk wall, the clock and the schoolhouse
- **mlg-deer** (16:9; attached: the front view) — Make this a 35 mm photograph of a cybernetic meadow at dawn: keep this woman, a Caucasian American woman in her late twenties with pale skin and a glossy black blunt bob with heavy straight bangs and one clay-orange streak, a thin headset microphone, in a cream knit dress, sitting on a warm-grey 1960s computer cabinet in tall grass; deer graze between rows of old computer cabinets with softly glowing lamps, pine trees, low mist. Peach and cream light, gentle. No logos. → sheet 21–24: 'the "1960s computers" came back as beige 90s PCs with lamps. That's the kitsch the research warned about.' · not used
- **mlg-curve** (16:9; attached: the front view) — Make this a wide 35 mm photograph from a low hill: keep this woman, a Caucasian American woman in her late twenties with pale skin and a glossy black blunt bob with heavy straight bangs and one clay-orange streak, a thin headset microphone, in a coral summer dress, walking through a wheat field; the wheat is cut into one huge smooth curve rising steeply from left to right across the valley, like a chart drawn in the land. Golden-hour sun, long shadows, cream sky. No logos. → sheet 25–28: "the quietest and most 1970s-film of them all" · #3 attached to s11-curve and s17-fork; behind the standing stones of the final cut
- **mlg-sing** (16:9; attached: the singing close-up) — Make this a 50 mm photograph in a sunlit meadow: keep this woman singing exactly, a 28-year-old Caucasian American woman with a grown-up angular face, defined cheekbones and a strong jawline, pale skin and light freckles, a glossy black blunt jaw-length bob with heavy straight bangs and one clay-orange streak through the bangs, a small flat clay-orange eight-pointed star hair clip, a thin headset microphone at her cheek, in a cream knit top; behind her, out of focus, coral poppies, wheat and a weathered white stone slab carved with a huge numeral. Overexposed afternoon light, peach and cream. No logos. → sheet 29–32: the canon held, as in 13–16 · #2 attached to s06-paradise, s22-lanterns and s24-close; behind the garden of the final cut

## wp-02-board — the storyboard plates · 33 prompts · 132 images · 29 picks

_One prompt per shot of storyboard draft 1 (`STORYBOARD.md`: 29 shots; S01, S12 and S21 have more than one prompt), written once the song was locked. Twenty-nine prompts attach one of the seven world frames (named after each prompt) and carry `--no blue`; the four frames of space and the sea are text only. The wardrobe, as written: "a cream linen summer dress" and "a cream knit top and high-waisted coral wool trousers", the world frames' two looks (S20 asks for "a white linen summer dress"). The lead's plan had Blender masters rebuild these frames; anabology corrected it: "hmm where does midjourney come into it? like i thought youd use wan instead of blender to render teh scenes and then you add animation on top of it (drawn with code) and then seedance as the final pass. theres no point in two video model passes without the intermediate animation pass." From then on the frames were plates, the places the video models put her in. The lead picked 29 images from 28 of the 33 jobs; the picks are recorded only as the plate table of the previz (`tools/film/takes.py`), and their reasons are not written down._

_Where they went: four were the places of the first H3 tests, in which she wears the plates' coral trousers and white top. anabology, on those tests: "this is great. i want to make sure when we do the final takes, you describe her outfit (like, say keep orange pants and white shirt) […]". All 29 were the places of the first previz's 21 takes (T01–T21). Four frames whose prompts never mention her came back with her in them anyway, carried over from the attached world frame (s21-42, s21-enigma, s21-calendar, s26-aerial). The 24 plates with her in them were emptied for the second previz (wp-04-places, wp-05-empty); the five with nobody in them went in as they are, and two of those five (the drone ship's deck and the fleet) are places of the final cut. After each prompt: the pick, its stage, and whether its place reaches the final cut, with the shot and its start in the song._

- **s01-bedroom** (16:9; attached: i3-meadow #4) — Make this a 35 mm photograph of a white linen bedroom at dawn inside the space colony: keep this woman, a Caucasian American woman in her late twenties with pale skin and a glossy black blunt bob with heavy straight bangs and one clay-orange streak, a thin headset microphone, in a cream linen summer dress, sitting on the edge of the bed on the left third, folding a pair of white socks into a neat square. On the right, a big round window looks out on green land curving up into the sky. Soft first light, cream and peach. No logos. → pick #2, mirrored → the bedroom: the intro take of the first H3 tests and T01 of the first previz; emptied in wp-05 · not in the final cut (after the reset the film opens in the capsule meadow)
- **s01-eye** (16:9; attached: i3-meadow #4) — Make this an extreme close-up 50 mm photograph of this woman's face as she opens her eyes on a white linen pillow at dawn: a 28-year-old Caucasian American woman with a grown-up angular face, defined cheekbones and a strong jawline, pale skin and light freckles, a glossy black blunt jaw-length bob with heavy straight bangs and one clay-orange streak through the bangs, a small flat clay-orange eight-pointed star hair clip, a thin headset microphone at her cheek. One eye fills the right third, warm first light across her freckles, cream and peach. No logos. → not picked
- **s02-terrace** (16:9; attached: i3-meadow #4) — Make this a vast wide 35 mm photograph inside a rotating space colony at morning: keep this woman, a Caucasian American woman in her late twenties with pale skin and a glossy black blunt bob with heavy straight bangs and one clay-orange streak, a thin headset microphone, in a cream linen summer dress, small at the lower-left third on a white stone terrace, back to camera. The land climbs both sides and closes overhead; on the far side above, one enormous word is mown into golden wheat. Long window strips of sun, haze. No logos. → pick #2 → the terrace, T02; emptied in wp-05 · not in the final cut
- **s03-tram** (16:9; attached: i3-tram #3) — Make this a 35 mm photograph from beside a small open white tram crossing apple orchards inside a space colony: keep this woman, a Caucasian American woman in her late twenties with pale skin and a glossy black blunt bob with heavy straight bangs and one clay-orange streak, a thin headset microphone, in a cream linen summer dress, sitting at the open back on the right third, hair in the wind; orchard trees rush past in the foreground, the land rises overhead. Warm morning sun. No logos. → not picked; a Blender test frame had been built toward its composition before the plan was corrected
- **s04-zero** (16:9; attached: mlg-slabs #4) — Make this a wide 35 mm photograph of a sunlit meadow: keep this woman, a Caucasian American woman in her late twenties with pale skin and a glossy black blunt bob with heavy straight bangs and one clay-orange streak, a thin headset microphone, in a cream knit top and high-waisted coral wool trousers, walking on the left third. Behind her a weathered stone slab carved 50%; on the right third a huge stone ring shaped like a zero with a black leather jacket hanging on it. Wheat, coral poppies, late sun, cream sky. No logos. → pick #2, mirrored → the stone zero: the verse take of the first H3 tests (the camera flies through the zero), T03; emptied in wp-05 → **in the final cut**, S04 at 0:31.0
- **s05-stair** (16:9; attached: mlg-slabs #4) — Make this a 35 mm photograph of a meadow: keep this woman, a Caucasian American woman in her late twenties with pale skin and a glossy black blunt bob with heavy straight bangs and one clay-orange streak, a thin headset microphone, in a cream knit top and high-waisted coral wool trousers, walking down a long stair of weathered white stone slabs that runs from the upper right to the lower left; each step is carved with one big number, the top one 2030. Wheat, coral poppies, late sun, cream sky. No logos. → pick #3 → the stair, T04; emptied in wp-05 → **in the final cut**, S05 at 0:36.8
- **s06-paradise** (16:9; attached: mlg-sing #2) — Make this a 35 mm photograph of a paradise garden: keep this woman, a Caucasian American woman in her late twenties with pale skin and a glossy black blunt bob with heavy straight bangs and one clay-orange streak, a thin headset microphone, in a cream knit top and high-waisted coral wool trousers, on the left third, smiling at camera. On the right third a small white stone chapel sends a tall column of white smoke into a cream sky; gentle white robot gardeners tend roses. Lush, sunlit. No logos. → pick #2 → the garden with the chapel and the robot gardeners, T05; emptied in wp-05 → **in the final cut**, S06 at 0:44.6
- **s07-math** (16:9; attached: mlg-slabs #4) — Make this a wide 35 mm photograph of a meadow: keep this woman, a Caucasian American woman in her late twenties with pale skin and a glossy black blunt bob with heavy straight bangs and one clay-orange streak, a thin headset microphone, in a cream knit top and high-waisted coral wool trousers, sitting on the arm of an old sofa in the grass at the lower left, chalk in hand; an old television on the grass shows a football match; behind, a giant slate chalkboard dense with equations spans the frame; beside it a cracked stone tablet. Late sun. No logos. → pick #2 → the sofa, the television and the chalk wall, T05; emptied in wp-05 → **in the final cut**, S07 at 0:50.4
- **s08-clock** (16:9; attached: mlg-slabs #4) — Make this a wide 35 mm photograph of a meadow: a monumental brass and white stone clock face, taller than a house, its hands just before midnight, stands in the wheat; keep this woman, a Caucasian American woman in her late twenties with pale skin and a glossy black blunt bob with heavy straight bangs and one clay-orange streak, a thin headset microphone, in a cream knit top and high-waisted coral wool trousers, small at its foot, one hand on the minute hand. Late golden sun, cream sky, poppies. No logos. → pick #2 → the clock, T06; emptied in wp-05 → **in the final cut**, S08 at 1:02.0
- **s09-tower** (16:9; attached: i3-meadow #4) — Make this a 35 mm photograph at dusk inside a space colony: on the left third a grey 1970s office tower with lit windows; on the right warm gardens and small white homes glowing; keep this woman, a Caucasian American woman in her late twenties with pale skin and a glossy black blunt bob with heavy straight bangs and one clay-orange streak, a thin headset microphone, in a cream knit top and high-waisted coral wool trousers, on a rooftop garden in the foreground, back to camera. The land curves overhead in the dusk. No logos. → pick #2 → the office tower, T06; emptied in wp-05 · not in the final cut
- **s10-sing** (16:9; attached: i3-sing #2) — Make this a 50 mm photograph at golden hour inside a space colony: keep this woman singing, a 28-year-old Caucasian American woman with a grown-up angular face, defined cheekbones and a strong jawline, pale skin and light freckles, a glossy black blunt jaw-length bob with heavy straight bangs and one clay-orange streak through the bangs, a small flat clay-orange eight-pointed star hair clip, a thin headset microphone at her cheek, in a cream linen summer dress, medium shot, her eyes on the upper third, standing at a little white tram stop in an apple orchard; the land curves up behind her. Warm sun on her face. No logos. → pick #3, mirrored → the orchard and its little white tram: the chorus take of the first H3 tests and of the green-suit and dance tests after them, T07; emptied in wp-05 → **in the final cut**, S09 at 1:13.6
- **s11-curve** (16:9; attached: mlg-curve #3) — Make this a vast wide 35 mm photograph: a grassy hill shaped exactly like a smooth bell curve rises out of a wheat field, with mown stripes down its slopes; keep this woman, a Caucasian American woman in her late twenties with pale skin and a glossy black blunt bob with heavy straight bangs and one clay-orange streak, a thin headset microphone, in a cream knit top and high-waisted coral wool trousers, standing tiny at its peak on the right third. Golden-hour sun, long shadows, cream sky. No logos. → not picked; a Blender test frame had been built toward its composition before the plan was corrected
- **s12-palms** (16:9; attached: i3-sing #2) — Make this an extreme close-up 50 mm photograph: two open palms held out toward the camera, a red capsule in the left palm and a blue capsule in the right palm; behind them, out of focus, this woman, a Caucasian American woman in her late twenties with pale skin and a glossy black blunt bob with heavy straight bangs and one clay-orange streak, a thin headset microphone, in a cream linen summer dress. Soft morning light, cream and peach. No logos. → pick #2 → the porthole room, T14; emptied in wp-05 · not in the final cut (the pill scenes became shader scenes)
- **s12-space** (16:9; text only) — A 35 mm photograph of a huge white O'Neill cylinder space colony floating in space, shaped like a capsule, long mirror panels catching the sun, the Earth glowing beyond, stars, in the style of the 1975 NASA space settlement paintings, clean and sunlit. No text, no logos. → pick #2 → the colony in space, T14 · not in the final cut
- **s13-castle** (16:9; attached: i3-meadow #4) — Make this a wide 35 mm photograph on a green hillside inside a space colony: gentle white robots stack stone blocks to raise a white castle on the centre-right; keep this woman, a Caucasian American woman in her late twenties with pale skin and a glossy black blunt bob with heavy straight bangs and one clay-orange streak, a thin headset microphone, in a cream linen summer dress with a small gold paper crown, sitting on a hay-bale throne at the lower left. Morning sun. No logos. → pick #3 → the castle and its robot builders, T09; emptied in wp-05 (#4 kept) → **in the final cut**, S13 at 1:40.5
- **s14-car** (16:9; attached: i3-tram #3) — Make this a 35 mm photograph: a small white car with no steering wheel glides along a road between fields of solar panels inside a space colony; keep this woman, a Caucasian American woman in her late twenties with pale skin and a glossy black blunt bob with heavy straight bangs and one clay-orange streak, a thin headset microphone, in a cream linen summer dress, in the back seat by the open window on the right third, hands behind her head. A roadside sign reads 20×. Sun. No logos. → pick #2 → the car with nobody at the wheel, T09 and T10; emptied in wp-05 (#2 kept) → **in the final cut**, S14 at 1:44.4
- **s15-school** (16:9; attached: mlg-slabs #4) — Make this a 35 mm photograph through the open door of a one-room wooden schoolhouse in a meadow: keep this woman, a Caucasian American woman in her late twenties with pale skin and a glossy black blunt bob with heavy straight bangs and one clay-orange streak, a thin headset microphone, in a cream knit top and high-waisted coral wool trousers, at a slate chalkboard on the right third; a child seen from behind sits at a wooden desk with a sapling in a pot. Warm afternoon light. No logos. → pick #2 → the schoolhouse, T10; emptied in wp-05 with the child kept → **in the final cut**, S15 at 1:52.1
- **s16-care** (16:9; attached: i3-meadow #4) — Make this a 35 mm photograph of a white garden pavilion inside a space colony: a royal four-poster bed made up in fine white linen, a gold paper crown on the pillow; beside it a brass punch clock on a post; keep this woman, a Caucasian American woman in her late twenties with pale skin and a glossy black blunt bob with heavy straight bangs and one clay-orange streak, a thin headset microphone, in a cream knit top and high-waisted coral wool trousers, tucking in the sheet. Soft afternoon sun. No logos. → pick #2 → the four-poster bed and the punch clock, T11; emptied in wp-05 → **in the final cut**, S16 at 1:59.8
- **s17-fork** (16:9; attached: mlg-curve #3) — Make this a high wide 35 mm photograph of a valley: a river splits into two branches, the left under a dark storm cloud, the right in sunlight; in the sunlit meadow a ring of standing stones; keep this woman, a Caucasian American woman in her late twenties with pale skin and a glossy black blunt bob with heavy straight bangs and one clay-orange streak, a thin headset microphone, in a cream knit top and high-waisted coral wool trousers, tiny among the stones. Golden sun. No logos. → pick #2 → the standing stones above the forking river, T12; emptied in wp-05 → **in the final cut**, S18 at 2:11.3
- **s18-battlements** (16:9; attached: i3-sing #2) — Make this a 50 mm photograph on the white battlements of a new castle inside a space colony: keep this woman singing, a 28-year-old Caucasian American woman with a grown-up angular face, defined cheekbones and a strong jawline, pale skin and light freckles, a glossy black blunt jaw-length bob with heavy straight bangs and one clay-orange streak through the bangs, a small flat clay-orange eight-pointed star hair clip, a thin headset microphone at her cheek, in a cream linen summer dress with a small gold paper crown, medium shot; behind her pennants lift in the wind and the land curves overhead. Morning sun. No logos. → not picked
- **s19-wheat** (16:9; attached: i3-word-in-wheat #4) — Make this a vast 35 mm photograph looking up inside a space colony: on the far side overhead, the words of a sentence are mown into golden wheat in huge heavy capitals, a harvester finishing the last letter; keep this woman, a Caucasian American woman in her late twenties with pale skin and a glossy black blunt bob with heavy straight bangs and one clay-orange streak, a thin headset microphone, in a cream linen summer dress, small at the bottom, looking up. Warm afternoon sun. No logos. → pick #1 → the words in the wheat, T13. It reads OVER, the word of the world frame it was made from; emptied in wp-05, then rewritten as WHITEPILL → **in the final cut**, S19 at 2:26.6
- **s20-capsules** (16:9; attached: mlg-slabs #4) — Make this a symmetrical wide 35 mm photograph of a meadow: two giant glossy capsule sculptures stand on either side, a red one on the left and a blue one on the right; keep this woman, a Caucasian American woman in her late twenties with pale skin and a glossy black blunt bob with heavy straight bangs and one clay-orange streak, a thin headset microphone, in a white linen summer dress, centred between them, facing camera. Soft late sun, cream sky. No logos. → pick #1, mirrored (the blue capsule then stands on the left) → the capsule meadow: the white-pill take of the first H3 tests, T08, and the "reference 6" anabology ruled out for take 8; emptied in wp-04 → **in the final cut**, S01 at 0:00.0, S03 at 0:27.2 and S11 at 1:29.0
- **s21-protein** (16:9; attached: i3-tram #3) — Make this a 35 mm photograph from a small open white tram passing through orchards inside a space colony: beside the track stands a giant sculpture of a folded protein, smooth ribbons and coils in cream and coral; keep this woman, a Caucasian American woman in her late twenties with pale skin and a glossy black blunt bob with heavy straight bangs and one clay-orange streak, a thin headset microphone, in a cream linen summer dress, riding at the open back on the right third. Sun. No logos. → pick #1 → the protein sculpture by the tram line, T15; emptied in wp-05 · not in the final cut (the break became shader scenes)
- **s21-enigma** (16:9; attached: i3-tram #3) — Make this a 35 mm photograph from a small open white tram passing through orchards inside a space colony: beside the track stands a monumental old Enigma cipher machine the size of a house, brass rotors turning, keys gleaming. Warm sun, the land curving overhead. No logos. → pick #1, with her at the front of the tram though the prompt has nobody → the Enigma machine, T15; emptied in wp-05 · not in the final cut
- **s21-42** (16:9; attached: i3-tram #3) — Make this a 35 mm photograph from a small open white tram passing through orchards inside a space colony: beside the track a huge clipped green topiary shaped as the number 42, a small white chapel behind it. Warm sun, the land curving overhead. No logos. → pick #3, with her in the tram though the prompt has nobody → the 42 topiary, T15; emptied in wp-05 · not in the final cut
- **s21-calendar** (16:9; attached: i3-tram #3) — Make this a 35 mm photograph from a small open white tram passing through orchards inside a space colony: beside the track a long wall of giant cream calendar pages, the first two carry the words for Monday and Tuesday, the rest blank. Warm sun, the land curving overhead. No logos. → pick #2, with her in the tram → the calendar wall, legible "Monday" and "Tuesday", T15; emptied in wp-05 · not in the final cut
- **s22-lanterns** (16:9; attached: mlg-sing #2) — Make this a wide 35 mm photograph of a meadow at night: millions of glowing paper lanterns rise into the dark sky; keep this woman, a Caucasian American woman in her late twenties with pale skin and a glossy black blunt bob with heavy straight bangs and one clay-orange streak, a thin headset microphone, in a cream knit top and high-waisted coral wool trousers, centre-left, lit warm by the lanterns, looking up. Deep night, warm amber light. No logos. → pick #2 → the lantern meadow, T16; emptied in wp-05 · not in the final cut
- **s23-droneship** (16:9; text only) — A wide 35 mm photograph at night at sea: a rocket booster lands on a flat drone ship barge in a burst of flame and spray, the deck lit, big white letters painted on the deck, waves, steam, cinematic. No logos. → pick #1 → the landing far out at sea, T16 · not in the final cut; pick #2 → the deck beside the landed booster, T17, then S22 of the v10 previz as it is → **in the final cut**, S22 at 3:00.9
- **s24-close** (16:9; attached: mlg-sing #2) — Make this a close-up 50 mm photograph at night: keep this woman, a 28-year-old Caucasian American woman with a grown-up angular face, defined cheekbones and a strong jawline, pale skin and light freckles, a glossy black blunt jaw-length bob with heavy straight bangs and one clay-orange streak through the bangs, a small flat clay-orange eight-pointed star hair clip, a thin headset microphone at her cheek, in a cream knit top, her eyes on the upper third, warm lantern light on her face, glowing paper lanterns drifting out of focus behind her. Tender, quiet. No logos. → not picked
- **s26-aerial** (16:9; attached: i3-meadow #4) — Make this a vast aerial 35 mm photograph flying along the length of a rotating space colony cylinder: farmland, orchards, a white castle, white towns and rivers climb both sides and close overhead; a long window strip of sunlight runs down the middle; huge words mown into the wheat far away. Morning haze. No logos. → pick #2, with her on the hill though the prompt has nobody → the hilltop over the curving land, T18; emptied in wp-05 → **in the final cut**, S23 at 3:08.5
- **s27-fleet** (16:9; text only) — A 35 mm photograph of a fleet of huge white O'Neill cylinder space colonies floating in sunlight like white capsules, mirror panels glinting, the Earth below, stars, in the style of the 1975 NASA space settlement paintings. No text, no logos. → pick #2 → the fleet of colonies, T19, then S24 of the v10 previz as it is → **in the final cut**, S24 at 3:23.7, outside the window
- **s28-night** (16:9; attached: i3-sing #2) — Make this a close-up 50 mm photograph in a dark bedroom at night: keep this woman, a 28-year-old Caucasian American woman with a grown-up angular face, defined cheekbones and a strong jawline, pale skin and light freckles, a glossy black blunt jaw-length bob with heavy straight bangs and one clay-orange streak through the bangs, a small flat clay-orange eight-pointed star hair clip, a thin headset microphone at her cheek, lying on a white pillow on the right third, eyes just opened; in the foreground left a glowing flip clock reading 2:14; an old teletype machine on the nightstand. Moonlight and a warm lamp. No logos. → pick #2 → the night bedroom, the flip clock reading 2:14, T20; emptied in wp-05 → **in the final cut**, S25 at 3:31.3
- **s29-endcap** (16:9; text only) — A 35 mm photograph from inside a space colony at night looking out through a vast circular end-cap window: the lit white towns curve around the dark land, and through the window the white capsule-shaped colony's neighbour floats among stars with the Earth. Calm, warm lights. No logos. → pick #1 → the end-cap window, T21 · not in the final cut

## wp-03-dress — her dress · 6 prompts · 24 images · 1 pick (anabology's)

_anabology's notes on the first previz changed her look: "hm dont love the shirt. […] maybe something else that is still feminine and cool here besides the white tank... can you try like 3 totally idfferent outfits? i want it to be unmistakably feminine, […]"; then: "can we do a mini white trad dress? can you show me the images of the outfit before generating this time". Of four dress stills from GPT Image 2.5, anabology chose the third, the milkmaid mini ("i like 3 but remake the image and just specify to make sure shes like 30 years old to gpt image and make appropriate edits"), and asked for a better character reference ("maybe if we had a better body /character reference - can you 1- find the black outfit from the don't let me scroll song, […] 2- what are you feeding the model to maek these images as a character sheet? should be using the close up faces one mainly"). Then: "maybe use some different image models tbh - could try midjourney"._

_The batch describes that dress in words. t-*: her description alone, age 30, "very tall and very skinny", rows A (front), B (walking) and C (a disco move) of the review, and one in the wheat. e-*: Midjourney edits of DON'T LET ME SCROLL's look-D plates, P22 (standing) and P20 (walking), for her height and build, the dress from the words; rows D and E. No `--no` list. The lead showed rows A–E as two sheets (`review/outfit/MJ_dress_A-C_text.jpg`, `MJ_dress_D-E_dlms-edit.jpg`), left the wheat out, and recommended D1 or D3: "Every face reads as an adult with freckles, and I checked hands and legs at full size on the likely picks." anabology: "I like A1 the most"._

- **t-front** (9:16; text only) — full-length lookbook photo, front view, standing straight, arms at her sides, a 30-year-old Caucasian American woman with a grown-up angular face, pale skin and light freckles, a glossy black blunt bob with heavy straight bangs and one clay-orange streak, a small clay-orange star hair clip, a thin headset microphone, very tall and very skinny, […] long thin legs, in a white cotton milkmaid mini dress: sweetheart neckline, lace-up bodice, puff sleeves, tiered skirt to mid-thigh with lace trim; white heeled Mary Janes, calm, lips closed, plain light grey studio, soft even light, no text, no logos → row A: "Frillier than the GPT Image version, with bigger puff sleeves." · **A1 (#1), anabology's pick: her dress for the rest of the film** — the frilly milkmaid mini with big puff sleeves, a lace-up bodice with silver eyelets, a two-tier lace-trim skirt and square-toe white Mary Janes. A1 has two star clips and her streak on the wrong side, so it was mirrored. GPT Image 2.5 made a turnaround sheet from the mirrored image (front, three-quarter, side and back at full length, and three views of the face), with one clip; its four full-length views were cut below the chin, because "its close-ups have A1's face" (her face was to come from her face sheet) and "the side views put the star clip on the streak side". That four-view image (`media/refs/outfit-v7/A1-dress-4views.png`) is the dress reference of the Wan passes and of every Seedance take of her on Higgsfield from then on, and **@Image2 of all 20 of her takes in the final cut**. The mirrored A1 itself was H3's picture of the dress, her shoes and her build in the second and the v10 previz. The look tests on Replicate, among them the bed shot whose look anabology chose for the whole film (anabology: "i like how satruated it is in bed scene. thats the look i like teh most"), ran with no image of her face: her identity there was the canon in words.
- **t-walk** (9:16; text only) — full-length editorial photo, three-quarter view, mid-stride, a 30-year-old Caucasian American woman with a grown-up angular face, pale skin and light freckles, a glossy black blunt bob with heavy straight bangs and one clay-orange streak, a small clay-orange star hair clip, a thin headset microphone, very tall and very skinny, […] long thin legs, in a white cotton milkmaid mini dress: sweetheart neckline, lace-up bodice, puff sleeves, tiered skirt to mid-thigh with lace trim; white heeled Mary Janes, cool confident look, plain warm grey studio wall, soft daylight, film grain, no text, no logos → row B · not used
- **t-dance** (9:16; text only) — full-length photo, a disco dance move, one arm pointing up, hip out, skirt swinging, a 30-year-old Caucasian American woman with a grown-up angular face, pale skin and light freckles, a glossy black blunt bob with heavy straight bangs and one clay-orange streak, a small clay-orange star hair clip, a thin headset microphone, very tall and very skinny, […] long thin legs, in a white cotton milkmaid mini dress: sweetheart neckline, lace-up bodice, puff sleeves, tiered skirt to mid-thigh with lace trim; white heeled Mary Janes, joyful, plain grey studio, no text, no logos → row C: "C2 and C4 show best how the skirt moves when she dances." · not used
- **t-wheat** (16:9; text only) — full-length 35 mm photo in a golden wheat field inside a vast space colony at golden hour, land curving up into the sky, a 30-year-old Caucasian American woman with a grown-up angular face, pale skin and light freckles, a glossy black blunt bob with heavy straight bangs and one clay-orange streak, a small clay-orange star hair clip, a thin headset microphone, very tall and very skinny, […] long thin legs, in a white cotton milkmaid mini dress: sweetheart neckline, lace-up bodice, puff sleeves, tiered skirt to mid-thigh with lace trim; white heeled Mary Janes, singing, warm 1970s colour, no text, no logos → not shown to anabology, as Midjourney shot all four from a low angle · not used
- **e-p22** (9:16; attached: DON'T LET ME SCROLL plate P22) — Dress this woman in a white cotton milkmaid mini dress: sweetheart neckline, lace-up bodice, puff sleeves, tiered skirt to mid-thigh with lace trim; white heeled Mary Janes. Keep her face, her black bob with the clay-orange streak, her headset microphone, her pose, her very tall skinny build and the grey studio. She is a grown woman of 30. No text, no logos. → row D: "D looks most like her. It's the same face in all four, with a clear streak and mic, and she reads about 30"; tall, the hem a little lower; no star clip, "because those plates didn't show it" · the lead's suggestion (D1 or D3), not used
- **e-p20** (9:16; attached: DON'T LET ME SCROLL plate P20) — Dress this woman in a white cotton milkmaid mini dress: sweetheart neckline, lace-up bodice, puff sleeves, tiered skirt to mid-thigh with lace trim; white heeled Mary Janes. Keep her face, her black bob with the clay-orange streak, her headset microphone, her pose, her very tall skinny build and the grey studio. She is a grown woman of 30. No text, no logos. → row E: "heavier glam makeup, so it looks a bit less like her" · not used

## wp-04-places — the capsule meadow, empty · 2 prompts · 8 images · 1 pick

_The take-8 test on H3 was to give her face from her face sheet and the A1 dress. anabology, on the references the lead proposed: "wait align w me on teh references you use lol - its clearly not going to work if you use reference 6, and you shouldnt use pic 1-4 just use the face character sheet". Reference 6 was the storyboard's capsule meadow (s20-capsules #1, mirrored), with her in it in the old outfit. Then: "can you make a scenery plate with midjourney to feed in as well just without her in it". Both prompts carry `--no people, person, woman, figure`. The lead showed the eight, mirrored to take 8's layout with the blue capsule on the left (`review/outfit/MJ_place_T08_empty.jpg`: P1–P4 the edit, Q1–Q4 the text version), and recommended P1. anabology: "go ahead, do it. […]". Take 8 with P1, the face sheet and the A1 dress then ran on H3; anabology, on its third version, with a new dance clip: "I actually like it better. […]". That take's recipe became the previz's._

- **e-caps-empty** (16:9; attached: s20-capsules #1) — Make this the same wide 35 mm photograph with nobody in it: remove the woman and continue the golden wheat and coral poppies where she stood. Keep the meadow, the two giant glossy capsule sculptures, a red one on the left and a blue one on the right, the low late sun, the cream sky and the composition exactly. No people, no logos. → P1–P4: "Midjourney's edit of the storyboard plate with her taken out. It's the same composition and light, with wheat and poppies where she stood. I checked the middle at full size and there's no trace of her. The four are nearly identical." · **P1 (#1)**, mirrored → the capsule meadow of take 8's tests, of T08 in the second previz and of S01, S03, S11 and S11b in the v10 previz → **in the final cut**, S01 at 0:00.0, S03 at 0:27.2 and S11 at 1:29.0
- **t-caps-empty** (16:9; text only, s20-capsules #1 as the style reference) — symmetrical wide 35 mm photograph of a vast meadow of golden wheat and coral poppies at golden hour, two giant glossy capsule-shaped sculptures several storeys tall standing upright on either side, a red one on the left and a blue one on the right, an empty field between them, low warm sun, soft haze, cream sky, warm 1970s colour, fine grain, no people, no text, no logos → Q1–Q4: "a fresh version written from text, in the plate's style. The capsules came out as kiosks and phone booths, so I'd skip these." · not used

## wp-05-empty — every place, empty · 23 prompts · 92 images · 23 picks

_anabology, on take 8 redrawn by Wan: "[…] but i think we're good. can we try to make the full video previz in h3 and you can make multiple shots per scene […]? try to get it perfect". Every place of the previz but the capsule meadow (done in wp-04) and the five frames with nobody in them still had her in the old outfit, so the lead took her out of all 23 with the same edit: each prompt attaches the storyboard pick it empties and names what to keep. All carry `--no people, person, woman, figure`, except e-school, whose `--no woman` keeps the child. The lead kept one image of each, "the closest match for each, with the schoolhouse child kept and the trams and car empty" (`media/refs/places/picks.json`): #1 unless noted below. No verdict from anabology on these images is recorded. The second previz used all 23; after the reset, the v10 previz used 14 of them for her shots, and those 14 reach the final cut._

- **e-bedroom** (16:9; attached: s01-bedroom #2) — Make this the same photograph with nobody in it: remove the woman and continue the scene behind her where she stood. Keep everything else exactly: the curved cream-white bedroom, the low white linen bed and the huge round window onto the curving land, the light, the colours and the composition. No people, no logos. → #1 → T01 of the second previz · not in the final cut
- **e-terrace** (16:9; attached: s02-terrace #2) — Make this the same photograph with nobody in it: remove the woman and continue the scene behind her where she stood. Keep everything else exactly: the white stone terrace and balustrade and the vast curving land of the colony with the word mown in the wheat, the light, the colours and the composition. No people, no logos. → #1 → T02 · not in the final cut
- **e-zero** (16:9; attached: s04-zero #2) — Make this the same photograph with nobody in it: remove the woman and continue the scene behind her where she stood. Keep everything else exactly: the meadow, the giant stone zero with the black leather jacket hanging from it and the row of carved stone numerals, the light, the colours and the composition. No people, no logos. → #1, mirrored → T03, then S04 of the v10 previz → **in the final cut**, S04 at 0:31.0
- **e-stair** (16:9; attached: s05-stair #3) — Make this the same photograph with nobody in it: remove the woman and continue the scene behind her where she stood. Keep everything else exactly: the long stair of carved white stone blocks running down through the meadow, the light, the colours and the composition. No people, no logos. → #1 → T04, then S05 → **in the final cut**, S05 at 0:36.8
- **e-paradise** (16:9; attached: s06-paradise #2) — Make this the same photograph with nobody in it: remove the woman and continue the scene behind her where she stood. Keep everything else exactly: the garden of white roses, the white robots tending the flowers and the little white chapel with its smoke, the light, the colours and the composition. No people, no logos. → #1 → T05, then S06 → **in the final cut**, S06 at 0:44.6
- **e-math** (16:9; attached: s07-math #2) — Make this the same photograph with nobody in it: remove the woman and continue the scene behind her where she stood. Keep everything else exactly: the old armchair sofa, the vintage television, the stone wall of chalk mathematics and the cracked tablet, the light, the colours and the composition. No people, no logos. → #1 → T05, then S07 → **in the final cut**, S07 at 0:50.4
- **e-clock** (16:9; attached: s08-clock #2) — Make this the same photograph with nobody in it: remove the woman and continue the scene behind her where she stood. Keep everything else exactly: the giant white stone clock face with brass hands and the meadow of wheat and poppies, the light, the colours and the composition. No people, no logos. → #1 → T06, then S08 → **in the final cut**, S08 at 1:02.0
- **e-tower** (16:9; attached: s09-tower #2) — Make this the same photograph with nobody in it: remove the woman and continue the scene behind her where she stood. Keep everything else exactly: the grassy terrace, the railing and the lit office tower on the curving land, the light, the colours and the composition. No people, no logos. → #1 → T06 · not in the final cut
- **e-orchard** (16:9; attached: s10-sing #3) — Make this the same photograph with nobody in it: remove the woman and continue the scene behind her where she stood. Keep everything else exactly: the little white tram on its rail, now empty, and the apple orchard rows, the light, the colours and the composition. No people, no logos. → #1, mirrored → T07, then S09 → **in the final cut**, S09 at 1:13.6
- **e-castle** (16:9; attached: s13-castle #3) — Make this the same photograph with nobody in it: remove the woman and continue the scene behind her where she stood. Keep everything else exactly: the green hillside, the white robots and the tall white castle, the hay bale left empty, the light, the colours and the composition. No people, no logos. → #4 → T09, then S13 → **in the final cut**, S13 at 1:40.5
- **e-car** (16:9; attached: s14-car #2) — Make this the same photograph with nobody in it: remove the woman and continue the scene behind her where she stood. Keep everything else exactly: the road through the orchards and the small white self-driving car, now completely empty with nobody inside, the light, the colours and the composition. No people, no logos. → #2 → T09 and T10, then S14 → **in the final cut**, S14 at 1:44.4
- **e-school** (16:9; attached: s15-school #2) — Make this the same photograph with nobody in it: remove the woman and continue the scene behind her where she stood. Keep everything else exactly: the old wooden schoolhouse porch, the blackboard, the sapling in its pot and the child sitting at the small desk, the light, the colours and the composition. No people, no logos. → #1, the child kept → T10, then S15 → **in the final cut**, S15 at 1:52.1
- **e-care** (16:9; attached: s16-care #2) — Make this the same photograph with nobody in it: remove the woman and continue the scene behind her where she stood. Keep everything else exactly: the white four-poster bed with its white linen and the brass punch clock on its stand on the green hillside, the light, the colours and the composition. No people, no logos. → #1 → T11, then S16 → **in the final cut**, S16 at 1:59.8
- **e-stones** (16:9; attached: s17-fork #2) — Make this the same photograph with nobody in it: remove the woman and continue the scene behind her where she stood. Keep everything else exactly: the circle of standing stones on the golden hill and the forking river, the light, the colours and the composition. No people, no logos. → #1 → T12, then S18 → **in the final cut**, S18 at 2:11.3
- **e-wheat** (16:9; attached: s19-wheat #1) — Make this the same photograph with nobody in it: remove the woman and continue the scene behind her where she stood. Keep everything else exactly: the aerial view of the colony's golden wheat fields and the giant letters mown into the wheat, the light, the colours and the composition. No people, no logos. → #1 → T13. It still reads OVER; for the v10 previz the lead had GPT Image 2.5 rewrite the word as WHITEPILL (three tries at $0.05 each, list price, `shots/ledger-img.jsonl`; Midjourney's version is kept as `media/refs/places/wheat-empty-OVER.png`), then S19 → **in the final cut**, S19 at 2:26.6
- **e-palms** (16:9; attached: s12-palms #2) — Make this the same photograph with nobody in it: remove the woman and continue the scene behind her where she stood. Keep everything else exactly: the small round porthole room and its round window onto the curving land, the light, the colours and the composition. No people, no logos. → #1 → T14 · not in the final cut
- **e-p42** (16:9; attached: s21-42 #3) — Make this the same photograph with nobody in it: remove the woman and continue the scene behind her where she stood. Keep everything else exactly: the hedges clipped into the number 42, the little white chapel and the white tram on its rail, now empty, the light, the colours and the composition. No people, no logos. → #1 → T15 · not in the final cut
- **e-enigma** (16:9; attached: s21-enigma #1) — Make this the same photograph with nobody in it: remove the woman and continue the scene behind her where she stood. Keep everything else exactly: the giant Enigma machine, the orchard and the white tram on its rail, now empty, the light, the colours and the composition. No people, no logos. → #1 → T15 · not in the final cut
- **e-protein** (16:9; attached: s21-protein #1) — Make this the same photograph with nobody in it: remove the woman and continue the scene behind her where she stood. Keep everything else exactly: the giant folded-protein sculpture, the orchard and the white tram on its rail, now empty, the light, the colours and the composition. No people, no logos. → #1 → T15 · not in the final cut
- **e-calendar** (16:9; attached: s21-calendar #2) — Make this the same photograph with nobody in it: remove the woman and continue the scene behind her where she stood. Keep everything else exactly: the white wall signs, the orchard, the rail and the little white tram, now empty, the light, the colours and the composition. No people, no logos. → #1 → T15 · not in the final cut
- **e-lanterns** (16:9; attached: s22-lanterns #2) — Make this the same photograph with nobody in it: remove the woman and continue the scene behind her where she stood. Keep everything else exactly: the night meadow and the thousands of glowing paper lanterns in the sky, the light, the colours and the composition. No people, no logos. → #1 → T16 · not in the final cut
- **e-aerial** (16:9; attached: s26-aerial #2) — Make this the same photograph with nobody in it: remove the woman and continue the scene behind her where she stood. Keep everything else exactly: the grassy hilltop, the curving land and the white tram line curving across the fields, the light, the colours and the composition. No people, no logos. → #1 → T18, then S23 → **in the final cut**, S23 at 3:08.5
- **e-night** (16:9; attached: s28-night #2) — Make this the same photograph with nobody in it: remove the woman and continue the scene behind her where she stood. Keep everything else exactly: the dark bedroom, the empty bed and pillow, the flip clock reading 2:14 and the old teletype, the light, the colours and the composition. No people, no logos. → #1 → T20, then S25 → **in the final cut**, S25 at 3:31.3

## The places of the final cut

Nineteen of her 20 shots stand in a place that Midjourney composed. In the first Seedance cut (v1) nearly every take of hers had the shot's H3 take as its guide clip (S01 had a Wan previz, S11b none). After anabology's notes on it ("wait we were supposed to totally abandon the h3 previs. seedance only was the goal in teh end with lip sync from teh vocal stems, that was the whole thing we cme to lol") the lead reshot all of them without a guide clip, and anabology chose between the two, section by section. So in the final cut nine of her takes have the shot's H3 take as their guide clip (@Video1, shrunk to 320 × 180 and scaled back up); the H3 take had the Midjourney place as a reference picture. S01's guide is a Wan 3.0 Prime previz made from a GPT Image 2.5 still of her in the place. Nine more were made without a guide clip: their place (@Image4) is a frame of the shot's first Seedance take with her removed by GPT Image 2.5, and that first take had the H3 take as its guide. Each step redrew the place: the final takes are in the look of the bed shot (S16), not Midjourney's colour. The shot's start is its time in the song.

| shot | starts | place | Midjourney images, oldest first | into the final take |
|---|---|---|---|---|
| S01 | 0:00.0 | the capsule meadow | mlg-slabs #4 → s20-capsules #1 → e-caps-empty #1, mirrored | a GPT Image still of her in it → Wan previz → guide clip |
| S03 | 0:27.2 | the capsule meadow | the same | H3 → first take → place frame |
| S04 | 0:31.0 | the stone zero with the leather jacket | mlg-slabs #4 → s04-zero #2 → e-zero #1, mirrored | H3 → first take → place frame |
| S05 | 0:36.8 | the stair of numbered stones | mlg-slabs #4 → s05-stair #3 → e-stair #1 | H3 → first take → place frame |
| S06 | 0:44.6 | the garden, the chapel, the robot gardeners | mlg-sing #2 → s06-paradise #2 → e-paradise #1 | H3 → guide clip |
| S07 | 0:50.4 | the sofa, the television, the chalk wall | mlg-slabs #4 → s07-math #2 → e-math #1 | H3 → guide clip |
| S08 | 1:02.0 | the clock in the poppies | mlg-slabs #4 → s08-clock #2 → e-clock #1 | H3 → first take → place frame |
| S09 | 1:13.6 | the tram in the orchard | i3-sing #2 → s10-sing #3 → e-orchard #1, mirrored | H3 → first take → place frame |
| S11 | 1:29.0 | the capsule meadow | mlg-slabs #4 → s20-capsules #1 → e-caps-empty #1, mirrored | H3 → guide clip |
| S13 | 1:40.5 | the castle and its robot builders | i3-meadow #4 → s13-castle #3 → e-castle #4 | H3 → guide clip |
| S14 | 1:44.4 | the car with nobody at the wheel | i3-tram #3 → s14-car #2 → e-car #2 | H3 → guide clip |
| S15 | 1:52.1 | the schoolhouse, the child, the sapling | mlg-slabs #4 → s15-school #2 → e-school #1 | H3 → guide clip |
| S16 | 1:59.8 | the four-poster bed and the punch clock | i3-meadow #4 → s16-care #2 → e-care #1 | H3 → first take → place frame |
| S18 | 2:11.3 | the standing stones above the forking river | mlg-curve #3 → s17-fork #2 → e-stones #1 | H3 → first take → place frame |
| S19 | 2:26.6 | WHITEPILL mown into the wheat | i3-word-in-wheat #4 → s19-wheat #1 → e-wheat #1, rewritten by GPT Image | H3 → first take → place frame |
| S22 | 3:00.9 | the drone ship's deck, the landed booster | s23-droneship #2 | H3 → guide clip |
| S23 | 3:08.5 | the hilltop over the curving land | i3-meadow #4 → s26-aerial #2 → e-aerial #1 | H3 → first take → place frame |
| S24 | 3:23.7 | the window onto the fleet of colonies | s27-fleet #2 | H3 → guide clip |
| S25 | 3:31.3 | the night bedroom, the flip clock at 2:14 | i3-sing #2 → s28-night #2 → e-night #1 | H3 → guide clip |

S11b (1:32.8), her close-up with the white pill, has no place image: it was made from words, her face sheet, the dress and the look still alone. Every one of the 20 takes has the A1 dress as @Image2.

## The image files

The images are kept with their batch files, one folder per batch. The project keeps what was used: the empty places as they went in (`media/refs/places/<place>-empty.png`, `T08-capsules-empty-P1-flip.png`, and `picks.json`), and the dress (`media/refs/outfit-v7/`: `A1-mirror.png`, the GPT Image sheet `sheet-A1.png`, `A1-dress-4views.png` and `A1-outfit-headless.png`). The review sheets anabology saw are in the project's `media/review/` (`step1-probe-sheet.jpg`, `styleframes-island-three.jpg`, `styleframes-loving-grace.jpg`) and `media/review/outfit/` (`MJ_dress_A-C_text.jpg`, `MJ_dress_D-E_dlms-edit.jpg`, `MJ_place_T08_empty.jpg`).

## Screened out, and why

- **p2-colony-world, all four:** her streak, clip and microphone missing; #1 reads young.
- **p3-pill-monument:** headlines that read as no word; the literal pill "may be too on the nose".
- **p1-sing-graphics:** diagram lines, arrows and HUD text, the web-dev look the film was made to leave.
- **mlg-deer:** the 1960s computer cabinets came back as beige 1990s PCs with lamps.
- **s21-42, s21-enigma, s21-calendar, s26-aerial:** her in the frame, though the prompt has nobody; emptied in wp-05 like the plates that ask for her.
- **t-front, rows A–C:** "the streak and clip land on the wrong sides in some, and A1 has two clips": A1 was mirrored, and its sheet cut below the chin.
- **t-wheat, all four:** low angles; never shown.
- **e-p22 and e-p20 (rows D and E):** no star clip; row E "heavier glam makeup".
- **t-caps-empty (Q1–Q4):** the capsules came back as kiosks and phone booths.
- **e-wheat #1:** the word OVER, from the world frame; rewritten as WHITEPILL with GPT Image.

## anabology's words on the images

Verbatim, typos kept, in order, each with what it answers:

- The method, on the lead's first candidates for the picture: quoted in full under *How the prompts were written*.
- On the world frames: "they look great i love these scenes. you can just choose."
- On the lead's plan to rebuild the frames in Blender: "hmm where does midjourney come into it? like i thought youd use wan instead of blender to render teh scenes and then you add animation on top of it (drawn with code) and then seedance as the final pass. theres no point in two video model passes without the intermediate animation pass."
- On the first H3 tests in the storyboard plates: "this is great. i want to make sure when we do the final takes, you describe her outfit (like, say keep orange pants and white shirt) […]"
- On her outfit in the first previz: "hm dont love the shirt. […] maybe something else that is still feminine and cool here besides the white tank... can you try like 3 totally idfferent outfits? i want it to be unmistakably feminine, […]"
- On the three outfits that followed: "can we do a mini white trad dress? can you show me the images of the outfit before generating this time"
- On four GPT Image dress stills: "i like 3 but remake the image and just specify to make sure shes like 30 years old to gpt image and make appropriate edits"
- On the same stills: "maybe if we had a better body /character reference - can you 1- find the black outfit from the don't let me scroll song, […] 2- what are you feeding the model to maek these images as a character sheet? should be using the close up faces one mainly"
- Then: "maybe use some different image models tbh - could try midjourney"
- On wp-03-dress: "I like A1 the most"
- On the references for take 8: "wait align w me on teh references you use lol - its clearly not going to work if you use reference 6, and you shouldnt use pic 1-4 just use the face character sheet"
- Then: "can you make a scenery plate with midjourney to feed in as well just without her in it"
- On wp-04-places and take 8's references: "go ahead, do it. […]"
- On take 8 on H3 in the empty meadow: "I actually like it better. […]"
- On take 8 redrawn by Wan: "[…] but i think we're good. can we try to make the full video previz in h3 and you can make multiple shots per scene […]? try to get it perfect"
- The reset, on cut 1: "okay so i totally overindexed on the orbit thing lol. i want to take a step back and honestly reshoot and reimagine the whole thing. the things i like 1- the setting, 2- her look. thats about it - its kinda dizzying now and previz was too complex so h3 didnt keep up wtih the physics or consistency as well as it should have i think for the final shots to have consistency."

## Lessons

- **Her canon text holds her; a picture of her alone does not.** The probe's edits, given her picture and "keep her face … exactly", dropped the streak, the clip and the microphone. With the canon pasted in, the world frames' close-ups kept all three.
- **Ask for things, not graphics.** "Graphics" brought back diagram lines, arrows and HUD text. Words asked for as wheat, stone, hedge, paper or a flip clock came back as objects.
- **Short words and numbers can come back legible.** OVER, 42, "Monday" and "Tuesday" and the flip clock's 2:14 came back readable; the probe's headlines did not. A word that comes back stays: OVER went from the world frame into the storyboard plate and into its empty version, until the lead had it rewritten.
- **The attached image comes along.** Four storyboard frames whose prompts never mention her have her in them, carried over from the attached world frame.
- **Taking her out keeps the place.** An edit that says "remove the woman and continue the scene behind her" kept the plate's composition and light; a fresh frame written in the same style, with the plate as a style reference, lost the objects.
- **A place picture with her in it is a second her.** anabology ruled out "reference 6", the capsule meadow with her in the old outfit. From then on every place went in empty, and her look came only from the face sheet and the dress.
- **The setting outlived every method.** Blender masters, Wan reshoots, the H3 previz, shaders and finally Seedance each took a turn. The worlds Midjourney composed at the start stayed, and anabology kept them at the reset: "the things i like 1- the setting, 2- her look."
