53 s · ends with a 5 s card naming every source. I cannot hear; neither could the critics. Sound is unjudged until a person listens.
A hand draws a line on a sheet, then a hand draws a tree. A rabbit stretches at a curtain as a narrator says 'draw'. A woman draws a curtain, a hand draws a pen out of a drawer, water is drawn from a well, a line is drawn across a map, spectators are drawn to a hall, a will is read in a drawing room. The film ends on the first hand: the line gets a dot.
19searches
27filmstrips looked at
12shots
1review rounds
1 The brief
A word passes between voices and images, changing sense at each cut until speech seems to edit the world around it.
First proof. Construct a sixty-second relay around one word with three visually legible senses. Establish its first use in an intact spoken context, then let an action carry the next meaning before another voice names it. Make the third sense alter the sequence’s tone or stakes. Return to the opening utterance with the new associations still active. Keep original speech intelligible, and test a version without explanatory captions before adding another semantic branch.
Development boundary. A semantic connection needs a defensible interpretation beyond a similarity score. An assembled utterance must be presented as an authored construction, not attributed to a recorded speaker as something they actually said.
The operator’s acceptance criteria
The critic can identify the organizing verbal relation after viewing, while describing at least two different senses.
Transitions do more than repeat a similarly shaped object.
The last return changes the significance of the first shot.
Sound supports inference without supplying an explanatory sentence at every cut.
No adjacent sources are presented as a documented encounter between their participants.
2 The prompt
The method prompt for this run, followed by this film's inputs. Paste it into an agent with a shell, ffmpeg, a speech recognizer and image reading. It is a reconstruction of what the run did, written afterwards; the run itself began from a short request.
1,150 words
You are making one proof film for ALLWAYS (a cinema practice within PAELEO) from the Moving Image Archive (movingimagearchive.com; public-domain films, which the site says are free to reuse). You work as a hand editor.
"By hand" means: you search, you look at what came back, you listen, and you write every in-point, out-point, order, gain and reason yourself, in an edit list. Software renders the list; it never chooses a shot, a trim or a sound. Do not generate footage. Do not pick shots by similarity score. Do not run a pipeline that edits for you.
## Input
- `[FILM]` the title, concept, first proof and development boundary (below).
- `[OPERATOR]` the primary SCION operator and its acceptance criteria (below; the full prompt is at `operator.html#s<N>`).
- `[SOURCE SET]` the archive's search API: `POST https://www.movingimagearchive.com/api/search` with `{"query": "..."}` returns clips with `videoUrl`, `thumbnailUrl`, `startSeconds`, `endSeconds`, `sourceTitle`, `sourceYear`, `sourceSlug`. One request at a time, about 1.4 s apart; it rate-limits.
- `[DELIVERY]` one film, 4:3, with sound, no longer than the proof asks, plus the case-study folder described in `RUNS.md`.
## Process (do all of it, in order; keep what you make, including what fails)
1. **Read the proof.** Write the three to six specific things a viewer must be able to notice. These are what you will judge your own cut against.
2. **Forage.** Write 8 to 14 searches in plain visible-action language ("a woman adjusts her hat, looking at herself"), not themes. Run each, save the results, and make one numbered contact sheet per query (thumbnail, #number, duration, year, title). Look at every sheet.
3. **Look closer.** For each candidate, fetch the clip and make a time-coded filmstrip with its waveform. Find where the action starts, completes, and what follows it. Note the native size and whether it has sound.
4. **Listen.** Transcribe candidates with word timings. Speech recognition invents words in hiss ("Thank you."). Check against the waveform and loudness before believing a word. An invented word is never evidence. If you cannot hear audio, say so in the case study and treat sound decisions as unverified.
5. **Write the edit list.** For each shot: clip number, in, out, gain, any sound taken from another clip (with its source), speed, hold, caption, and one sentence of reason that a reader could disagree with. Sound or captions from another source go in the ledger as such.
6. **Render, then check the render.** Make a contact sheet of every shot's head, middle and tail; transcribe the render; measure loudness per shot. Fix what you find. Known traps: a clip's closing cut can show up in your trim; holding a last frame can freeze the next shot of the source.
7. **Disclose derived layers.** If you remove, extend, generate, retime, mirror, or re-voice anything: keep the original, mark the derived layer on screen or in the ledger, and say what is unknown. Authored narration is labelled invented and synthetic.
8. **Blind critic.** Give a fresh agent only: a timecoded contact sheet of the final film, a transcript with loudness per 2 s, and the film's brief and acceptance criteria. Not your edit list, ledger, reasons or summaries. It names one specific event before it reads the brief. It returns a verdict (pass / reference stronger / inconclusive), the largest gap, up to three repairs each with a timecode and a checkable condition, and one strength to protect. It cannot hear: take no sound verdicts from it, and check its claims against the cut before obeying them.
9. **Repair what is true; leave what is wrong and say so.** If you changed the film, ask a new critic. Stop when you can say honestly what still falls short of the brief.
10. **Package** the case study per `RUNS.md`: searches and sheets, strips, the edit list with reasons, the ledger, the critic packets and verbatim reports, repairs, and what is open.
## Rules
- Status is not self-awarded. A film is "proof in progress" unless a blind critic passed it; "released" is for a person to say.
- Unknown dates and names stay unknown. Never imply two clips recorded the same event.
- Do not make assembled speech look like a speaker's own statement.
- Say what you could not perceive. A maker that cannot hear, or cannot see motion, must say so, because it changes what the cut proves.
## Tools (in `tools/`; Python 3, ffmpeg, Pillow, numpy)
- `forage.py <film> "query" ...` runs searches one at a time, writes `forage.json` and a contact sheet per query.
- `look.py <film> <n>[:step[:in:out]] ...` makes a time-coded filmstrip with waveform for clip #n.
- `hear.py <film> <n> ...` word timings (needs `mlx_whisper`, Apple silicon; substitute any word-timing ASR).
- `beats.py <film> <n> <in> <out>` motion-energy peaks, a helper for the eye.
- `cut.py <film>` renders `edl.json`; it chooses nothing. `pairs.py <film>` renders side-by-side pairs from `pairs.json`.
- `reel.py <film>` contact sheet of the rendered film, shot by shot.
- `packet.py <film>` builds what the critic gets.
- `erase.py` + `erase_plate.py` person removal for NOBODY HOME (SAM 2.1 tracked mask; replacement pixels from the same shot; unknowns listed). Set `SAM2_ROOT` to a folder with `models/sam2.1_hiera_small.pt`.
- `speak.py` authored narration with Kokoro-82M. `adopt.py` brings clips found in a local word index into a film's forage (set `CINEOSIS_LAB`).
---
## This film
**Title:** HOMONYM (Words change worlds)
**Concept:** A word passes between voices and images, changing sense at each cut until speech seems to edit the world around it.
**First proof:** Construct a sixty-second relay around one word with three visually legible senses. Establish its first use in an intact spoken context, then let an action carry the next meaning before another voice names it. Make the third sense alter the sequence’s tone or stakes. Return to the opening utterance with the new associations still active. Keep original speech intelligible, and test a version without explanatory captions before adding another semantic branch.
**Development boundary:** A semantic connection needs a defensible interpretation beyond a similarity score. An assembled utterance must be presented as an authored construction, not attributed to a recorded speaker as something they actually said.
**Primary operator:** SCION 55 · Words — Polysemic Relay. Full operator prompt: https://hartswf0.github.io/bolas/paeleo/allways/operator.html#s55
**Supporting:** SCION 92 · Hands of Bresson — Gesture Evidence Before a Generalization
**Operator's creative theorem:** Make viewers discover a word's changing senses through successive visible actions, then use the discovered ambiguity to change the sequence's emotional direction. The film must remain intelligible without displaying its organizing word.
**Acceptance criteria (from the operator):**
- The critic can identify the organizing verbal relation after viewing, while describing at least two different senses.
- Transitions do more than repeat a similarly shaped object.
- The last return changes the significance of the first shot.
- Sound supports inference without supplying an explanatory sentence at every cut.
- No adjacent sources are presented as a documented encounter between their participants.
3 What was searched
One search at a time, in plain visible-action language. Each sheet is what came back: numbered thumbnails with duration, year and title. These are the images the maker looked at (or, where it could not, see the note below).
a hand drawing a line on paper with a pencil · 24 resultsartist sketching with charcoal, hand close up · 24 resultsa child drawing on paper with a crayon · 24 resultsdrawing the curtains closed at a window · 24 resultsa woman pulls the curtains open in the morning · 24 resultspulling open a dresser drawer · 24 resultsa man draws water from a well with a bucket · 24 resultsdrawing a pistol from a holster · 24 resultsa crowd draws closer to look at something · 24 resultsdrawing a raffle ticket out of a drum · 24 resultshorse drawn wagon · 24 resultsan archer draws a bow · 24 resultsperson opens a desk drawer and takes something out · 24 resultskitchen drawer pulled open, silverware · 24 resultswoman opens bureau drawer, folded clothes · 24 resultspulling a bucket of water up from an old well, rope · 24 resultswoman at a window draws the drapes · 24 resultswire being drawn through a die in a factory · 24 resultspeople gather around and lean in to look · 24 results
4 What was looked at closely
A time-coded filmstrip and waveform for each candidate. Outlined strips are in the cut.
#13 Plastics (1944) · in the cut (shot 1, 12)#29 Frances and Her Rabbit (1961) · in the cut (shot 2)#32 Frances and Her Rabbit (1961) · looked at, not used#44 Frances and Her Rabbit (1961) · looked at, not used#77 Frances and Her Rabbit (1961) · looked at, not used#78 Preventing the Spread of Disease (1940) · looked at, not used#88 Flying Businessman (1953) · looked at, not used#100 The Electric House (1922) · looked at, not used#103 Test Tube Tale (1941) · in the cut (shot 4)#104 Frances and Her Rabbit (1961) · looked at, not used#134 City Water Supply (1941) · in the cut (shot 6)#137 Singing Wires (1951) · looked at, not used#265 Frances and Her Rabbit (1961) · in the cut (shot 3, 11)#266 The Time of Apollo - 1960 to 1975 (1960) · looked at, not used#267 National Wildlife Refuge System (2005) · looked at, not used#268 Harvest Of The Years (undated) · in the cut (shot 8)#268 Harvest Of The Years (undated) · in the cut (shot 8)#268 Harvest Of The Years (undated) · in the cut (shot 8)#269 Alaska's Silver Millions (Part I) (1936) · in the cut (shot 7)#271 Of Ice and Fire: The Mono Basin (1992) · looked at, not used#274 Fundo in Chile (1949) · in the cut (shot 9, 10)#276 Coffee Break (undated) · looked at, not used#276 Coffee Break (undated) · looked at, not used#277 Safety in Offices (1944) · in the cut (shot 5)#281 Molly Grows Up (1953) · looked at, not used#296 Television Remote Control (Tuner) (1961) · looked at, not used#323 Fire and Police Service (1947) · looked at, not used
5 The cut
The finished film, shot by shot (first, middle and last frame of each):
The final cut as a contact sheet
Every shot, where it came from, and why it is here. Source seconds are in the original film. Links go to the archive's own copy of the clip.
sense 1, to make a line. Blank sheet, shadow, the line begins; cut on the moving hand before it finishes. Its narrator's first words are left out; the rest of the line is held back for the end.
return: the hand leaves and the line stays. It now looks like the route the film took.
6 Review
A fresh agent was given only a timecoded contact sheet of the film, a transcript with loudness, and the brief. Not the edit list, the ledger, or these reasons. It cannot hear.
Round 1 inconclusive, leaning reference stronger
independent critic (fresh Claude Sonnet 5.5 agent). Its report, verbatim:
The report
VERDICT: inconclusive, leaning reference stronger. Cannot hear the film, so the sound-based criteria go unjudged.
Noticed before the brief: at 00:17.9-21.8 the voice says "a place for everything and everything should be in its place" over hands sorting a pen tray. This is the one cut where picture and line interlock. Before that, I read the opening as "draw" (hand sketching, "draw the picture"), then "drawn" by 24.8 (map line) and 29.1 ("draw people here"), then "drawing room" at 42.8. I took the word to be "draw" and the film to be an inventory of its senses.
Largest gap: the senses are listed, not relayed. Each is a new shot with a new line, and there is no action that carries a sense across a cut. Stakes and tone never change. The "will" and candle section (34-50) reads as a separate dramatic scene that does not grow out of the earlier senses.
Repairs:
1. 29.1-34.0 and 42.2-49.3: the same footage runs on for about 5s each (the rows of near-identical frames). Cut each to under 2.5s, or the relay stalls. Check: no frame repeats across more than 3 consecutive sheet cells.
2. 50.7-56.3: the return does recast the opening. The rabbit appears at 50.5 on "What is it?", then the hand draws the line at 51.6-52.8 and a dot lands at 52.8. But the panda at 54.0 is a non sequitur. End on the drawing. Check: the last 3s contain no new source.
3. 24.6-34.0: three consecutive geography and tourism shots (map, hall, crowd) carry the same literal "draw". Drop one. Check: no more than two shots in a row carry the same sense.
Strength to protect: the 51.6-52.8 return, where the line gets a dot and reads as a question mark.
The senses are listed more than relayed: fair. The panda frame the critic called a non sequitur was the source clip's own closing cut, caught by a held last frame; trimmed afterwards. The long repeated stretches were shortened.
7 What happened
Repairs after review
Trimmed the source's closing cut so the film ends on the drawing.
Shortened the two long repeated stretches.
What worked
Searching the lab's own word index (Whisper word timings for 4,559 archive films) for 'draw' found 25 spoken uses. Four of those films are in the cut: 'should draw in the picture', 'a line drawn from the center', 'things that draw people here', 'the drawing room'.
Where no narrator says the word, the action carries the sense (the curtain, the drawer, the well).
What failed
I could not see this film's contact sheets or filmstrips while cutting it: every image read returned a request-limit error. I cut from search titles, Whisper word timings, durations and loudness numbers. After the cut I checked each reason against frames of the final film and corrected two that described things the frames do not show.
The senses are listed more than relayed: nothing carries a sense across a cut except the pulling arm.
A held last frame once froze a closing cut of the source (a panda); trimmed.
48 s against 60 s.
What is still open
A pivot that changes the stakes, not only the sense.
Cut the repeated stretches under 2.5 s as the critic asked (partly done).
8 Files
edl.jsonedit list: every in, out, gain and reason, as written
ledger.jsonledger: each shot traced to its source seconds
Argon