aisle/tools/AI Trailer — shot board, prompts and agent pack | AISLE
free to read Buy 10 runs →

Studio · Trailer · 16:9 → 9:16 · 60s

AI Trailer

A trailer built as a cheap animatic first: three acts, escalating stakes, one promise, then premium renders on the used shots only. The title lands on the promise, not on a summary, and every cutdown is a re-edit of the same board.

12 shots / 60s trailer core boardcast to script14 moodboards10 variantsNo ceilingNo ceiling

The whole line, one pack

story to stitch

01 · PromiseWrite the one promise the trailer sells and the three acts that escalate it.
02 · LookThree-act palette, contrast, type and lens language. Decide the escalation before the model does.
03 · Cast or product lockOriginal AI-generated identities or the product, locked to a sheet. No stock face, no borrowed likeness.
04 · BoardEvery shot gets a timecode, act, lens, movement, action and prompt. Cut to the promise, not the plot.
05 · GenerateKeyframes, image-to-video, impact SFX, music and title cards. Three takes; keep the one that earns the click.
06 · ConsistencySame cast, product, grade and type in every act. Paste the lock and reuse the seed.
07 · SoundImpact hits, riser, one music bed and title-card audio.
08 · Cut and deliverTimeline, titles, captions, delivery in 16:9, 9:16 and 1:1. The cut is the promise.

The board below is the core, not a ceiling; extend it until the story lands.

What you get

11 deliverables

  • Three-act trailer structure with the one promise written first
  • Animatic: stills plus temp audio, timed to the second
  • Shot board with lens, move, action, on-screen text and line
  • Premium renders only for shots that survived the animatic
  • Title and end card typeset with real fonts
  • 30 s, 15 s and 6 s cutdowns from the same board
  • Sound design pass and a music edit that lands on the title
  • Delivery manifest with ratios, loudness and captions
  • Animatic cut file with temp audio and a locked timing sheet
  • Promise test: five readers repeat it back after one viewing
  • Cutdown matrix: which shots survive 30, 15 and 6 seconds

World and look

paste before anything

WorldYOUR WORLD - one location family and one threat. The trailer shows the world in three states: normal, wrong, gone. Each state gets one light logic and one lens family.
PaletteYOUR PALETTE - a cold base, one warm accent that belongs to the lead, and a desaturated state for the third act. The title card uses the same accent, never a new colour.
Negativeextra fingers, fused hands, face morph, changing eye colour, plastic skin, warped jewellery, floating earrings, captions, watermark, logo, extra teeth

Locks

location, props, wardrobe, axis

leadOne lead card: face, hair, garment, injury state per act. The injury is part of continuity, not a mood.
locationOne card per location, three states: normal, wrong, gone. The same room, re-lit, not a new set.
propThe object the trailer promises gets a card and appears in the first ten seconds and the last ten.
titleTitle font, weight, tracking and safe area locked in the animatic. The title is typeset, never generated.
promiseOne sentence, written before the board. Every act serves it and the title lands on it.
audioOne drone, three hits, one silence, one button. Nothing else enters the mix without replacing something.
ratio16:9 is the master; 9:16 is re-framed per shot from the same board. The title is re-laid, never scaled.
cutdownsEvery cutdown is re-edited from the board: the 6 s is not the 30 s with clips removed, it is the one image and the one promise that survive every length.
thumbnailThe poster frame is boarded as a shot, not grabbed from the cut. It is judged at 320 px before the trailer is rendered.

Every lock is pasted into the prompt for every shot that touches it. A lock that is paraphrased is not a lock.

Cast locks

1 identities

Identity slot

LEAD

Fill from your own vault card: fixed age, face, hair, one distinguishing mark.

One signature garment held identical across every shot and episode.

Fixed voice note for this role: pace, tone, one verbal habit.

LEAD: [fixed age], [fixed face], [fixed hair], [signature garment] - locked across every shot.

Shot prompt grammar

16 rules every prompt on this board obeys

  • Order: Subject, Action, Camera, Style, Constraints (S-A-C-S-C). Subject first so the model binds the identity before it reads motion.
  • Actions stay slow, soft and continuous. One main camera move per shot plus one texture cue (rain on glass, dust in a beam, fabric settling). Two moves in one prompt cancel each other.
  • 50-80 words, three sentences, one visual promise. "Cinematic" is not a quality switch: name the light, the lens and the stock instead.
  • Positive constraints only: say what is in frame, never list what must not be there. Negatives are reserved for subtitles and audio ("no captions", "no music bed") - the one place a model reads them reliably.
  • Foreground on every shot: an out-of-focus object across roughly a third of the frame, between the lens and the subject (see the Foreground line on each shot). Say "blurred" or the model renders it sharp. Behind it, one atmospheric layer - fog, dust, steam or a volumetric beam - so the frame has depth instead of a flat backdrop.
  • Length matches the file. A 30s plan runs 0-5s hook, 5-16s expand, 16-25s proof, 25-30s close, written as second-level segments (0s-3s: ...). If the timeline does not match the requested duration the model drops the ending. On a model that ignores seconds, fall back to shot numbers; on 2.5-class models seconds are first-class.
  • Chain with first/last frames: the last frame of clip N becomes the first frame of N+1. Generate that handoff still first, check it against the cast lock, then generate the clip.
  • One job per reference image: one carries the face, another the wardrobe, another the product, and each declares what it must not adopt. Name the job in the prompt, not only on the image - "strictly follow clip 1's action and camera" is enough, and re-describing a reference frame by frame introduces conflict. Stay inside the model cap (Seedance 2.5: 30 images, 10 videos, 10 audio).
  • Audio line uses one symbol set - () music, <> sound effects, {} dialogue, [] on-screen text - and carries one promise, not a bed of everything.
  • Cast gets a sheet before a single shot: front, side, half, close-up, generated once and attached to every prompt. Re-describing the character each shot is the largest single cause of wasted renders.
  • Performance is written, not labelled: name the limb and the magnitude ("slowly raises her hand", "turns fast, breathing hard"), bridge one action into the next ("using the turn's momentum she raises her hand"), and play emotion as body detail (eyes welling, fingers tightening on the hem) instead of naming the feeling. Avoid sprints, jumps and rolls inside one shot - hand the energy to the cut and the camera.
  • Light gets its own sentence: direction, quality (hard or soft), contrast. Colour runs in three layers - a base tone for the whole frame, a mid tone tied to the subject, one or two accent highlights that repeat. Then delete the word cinematic.
  • Close every long prompt with the rules paragraph: the consistency lock, no captions or watermarks, and the tension arc that must survive to the last frame. It is the red line the model does not cross.
  • Debug at draft resolution. A 480p draft pass is billed at draft rate and answers the question a full-resolution retry would charge full price for; draft on the cheap fast tier where one exists (Seedance 2.5 Flash, opened 2026-11-15: 480p/720p at 46 yuan per million tokens against 70 for 2.5, about twice as fast), and upgrade only the winner. Retakes eat most of a budget - roughly seven of ten credits die on a bad take - so previs the complex shot as a white model first (7-10 retries down to 2-3), shoot 3-5s pieces and stitch, and change one variable per retry.
  • Gate the first frame before you scale: wrong face or product, garbled text, malformed hands, CGI look. A failed render gets diagnosed in concrete words first - material, light path, aerial perspective, overlapping structure, a negative that did not hold - then repaired with those words. A blind re-roll with a new seed doubles the bill; only after the repair fails do you change the seed.
  • Read the reference post before you write: pull its transcript and overview frame sheet locally so line timing and cut rhythm come from the real post, not a guess.

The same block is baked into the SKILL.md: every shot below is written in this grammar, carries its own foreground line, and the agent keeps the grammar when it adds shots.

Skills this pack loads

4 skills, in order

SkillJob in this run
copywritinghooks, headlines and page copy
imagekeyframes, model sheets and moodboards
videoshots, motion, editing and the cut
socialorganic and platform format

The SKILL.md installs these in order. Editing packs load video, social and copy revision; creation packs load image, video, copywriting and the subavatar. A pack never loads a skill it does not use.

Ad sources this pack reads

plug in your library

SourceWhat it givesPlug
GetHookdwinning ads, hooks, scripts and transcriptscollection export, CSV or API
TikTok Creative Centertop ads, hooks, music and formatsURL plus a saved video, transcript and caption
Paste, CSV or screenshotany ad the buyer can see and has the right to studyad URL, screenshots, transcript, caption, CSV

If the buyer names a library, the agent uses its export or API. If not, it asks for one export, URL, screenshot, transcript or CSV. Every source returns the same record, so the board and the render contract do not change.

Any model, any seat

swap on a rate limit

Claude, ChatGPT, DeepSeek, Grok, Kimi, Muse, Gemini, Qwen and Seedance can each hold any seat: read the source, write the board, hold the brand voice, run QC. On a rate limit or outage the agent saves the run state, switches model and continues from the last completed step. It never restarts the run and never re-charges a finished take. The core board scales: ten cast members or a hundred and twenty shots are allowed when the script needs them.

The shot board

12 core shots · full prompts

SHOT 01
0.0-5.0s
24mm
slow push

Establish — World and [LEAD] in silhouette. The promise is visible in what is missing.

24mm · slow push · 0.0-5.0s

Wide of [LOCATION CARD], [LEAD LOCK] in silhouette, cold base palette, 24mm, slow push, 16:9, no text

foregroundthe out-of-focus dash edge across the bottom third, out of focus, between the lens and the subject

audioLow drone, one distant sound.

line[the promise, whispered]

SHOT 02
5.0-10.0s
35mm
locked

Normal — One detail says this world is not safe. Nobody reacts yet.

35mm · locked · 5.0-10.0s

[LOCATION CARD] detail, [PROP CARD] wrong in frame, 35mm, locked, cold base, 16:9

foregrounda passing guard rail, smeared by motion across the right third, out of focus, between the lens and the subject

audioRoom tone, one wrong note.

SHOT 03
10.0-15.0s
50mm
slow push

Stakes — A cost is made visible, not explained: a empty chair, a broken lock, a missing coat.

50mm · slow push · 10.0-15.0s

[LOCATION CARD] with the cost visible, [LEAD LOCK] at edge of frame, 50mm, slow push, 16:9

foregrounda blurred strand of hair across the left third, out of focus, between the lens and the subject

audioMusic enters low.

SHOT 04
15.0-19.0s
85mm
fast push

Turn — The turn lands on a face. One look, no line.

85mm · fast push · 15.0-19.0s

[LEAD LOCK] close-up, turn landing, 85mm, fast push, warm accent on skin, 16:9

foregroundthe soft edge of a shoulder across the right third, out of focus, between the lens and the subject

audioImpact.

SHOT 05
19.0-23.0s
24mm
whip

Escalation — The world reacts in one move; the camera follows late.

24mm · whip · 19.0-23.0s

Wide of [LOCATION CARD] reacting, [LEAD LOCK] crossing frame, 24mm, whip, 16:9

foregrounda blurred A-pillar across the left third, out of focus, between the lens and the subject

audioRiser, one hit.

SHOT 06
23.0-27.0s
35mm
handheld

Montage one — Three quick costs, one per beat, each a different distance.

35mm · handheld · 23.0-27.0s

Montage frame one, [COST ONE], [LOCATION CARD], 35mm, handheld, 16:9

foregroundthe soft edge of a shoulder across the right third, out of focus, between the lens and the subject

audioThree hits on the beat.

SHOT 07
27.0-31.0s
50mm
locked

Montage two — The second cost, closer, on the lead.

50mm · locked · 27.0-31.0s

[LEAD LOCK] and [COST TWO], 50mm, locked, 16:9

foregrounda hanging fabric edge, blurred across the top third, out of focus, between the lens and the subject

audioHit, tail.

SHOT 08
31.0-35.0s
85mm
slow push

Montage three — The third cost lands on [LEAD]. Hold the face.

85mm · slow push · 31.0-35.0s

[LEAD LOCK] close, cost landing, 85mm, slow push, 16:9

foregrounda blurred strand of hair across the left third, out of focus, between the lens and the subject

audioMusic drops out.

SHOT 09
35.0-40.0s
24mm
locked

Breath — One still frame; the trailer stops. The silence is the beat.

24mm · locked · 35.0-40.0s

Still of [LOCATION CARD], no cast, cold base, 24mm, locked, 16:9

foregroundtwo blurred leaves across the left third, out of focus, between the lens and the subject

audioSilence, one breath.

SHOT 10
40.0-46.0s
35mm
slow push

Promise — [LEAD] makes the promise in one line, to camera or to the thing they lost.

35mm · slow push · 40.0-46.0s

[LEAD LOCK] delivering the promise, 35mm, slow push, warm accent, 16:9

foregrounda blurred doorframe edge across the left third, out of focus, between the lens and the subject

audioMusic returns.

on screen[the promise]

line[the promise, spoken once]

SHOT 11
46.0-53.0s
50mm
lateral track

Final image — The image the audience came for, held one beat longer than comfortable.

50mm · lateral track · 46.0-53.0s

[FINAL IMAGE], [LEAD LOCK] and [PROP CARD], 50mm, lateral track, 16:9

foregroundsteam curling off a plate, out of focus across the bottom third, out of focus, between the lens and the subject

audioFull hit.

SHOT 12
53.0-60.0s
24mm
locked

Title — Title card and release line, typeset, on the accent colour.

24mm · locked · 53.0-60.0s

Title card background, [ACCENT] field, no cast, 24mm, locked, 16:9, text added in the edit

foregroundout-of-focus tall grass across the bottom third, out of focus, between the lens and the subject

audioTail, one button.

on screen[title, release line, CTA]

Core board, not a cap. Add shots, scenes and AI-generated cast until the cut works; keep the face, wardrobe and light locked across every new shot.

Beat sheet

7 beats

TimeBeatJob
0-20%Act IEstablish the world and the promise. One image that says what is at stake.
20-35%Inciting breakThe normal ends. Show the event, do not explain it.
35-55%Act IIRaise the stakes and show the cost on the lead.
55-70%EscalationThree quick costs, one per beat, each closer than the last.
70-85%Act IIIClimax and turn. The image the audience came for.
85-92%BreathOne still frame. The trailer stops before the title.
92-100%TitlePromise stated, title card, release line. Nothing after it.

Moodboard prompts

14 images

Title card · 16:9The title in the locked font on the accent field, with the release line. Typeset, checked at thumbnail size.
Act sheet · 16:9Three frames, one per act, same location family, three light states: normal, wrong, gone.
Cutdown map · 16:9The 30, 15 and 6 second cuts laid out on one timeline, with the shots that survive every cut marked.
Promise frame · 16:9One image that states the promise without text, used as the thumbnail.
Lead states · 3:4The lead in three states: normal, injured, resolved. Same garment, same light family.
Animatic strip · 16:9Every shot as a still in order with its duration under it. This is the film until the animatic is approved; nothing renders before it.
Three states · 16:9The same location in normal, wrong and gone, same lens, three light states. If the third state reads as a new set, re-light rather than rebuild.
Promise card · 16:9The promise sentence typeset in the locked font on the accent field, at thumbnail size. If it cannot be read at 320 px, it is not the promise yet.
Cutdown matrix · 16:9Three columns for 30, 15 and 6 seconds, with the shots that survive all three marked. The survivors are the spine; everything else is optional.
Sound map · 16:9The hits, the silence and the button drawn on the timeline with timings. The trailer is mixed before it is cut.
Poster frame · 2:3The one image that sells the trailer without text: lead, threat and world in one composition. This is also the thumbnail.
Question strip · 16:9One frame per act with the question that act leaves open written under it. If an act has no question, it is an answer and belongs in the film, not the trailer.
Ratio title · 9:16The title re-laid for vertical with the platform UI drawn over it. The card is designed twice, not scaled once.
Breath frame · 16:9The still that holds before the title, no cast and no movement. It exists to make the title land, so it is boarded as a shot rather than left to the edit.

Hooks and variants

10 openers

  • Promise-first: open on the final image and run the trailer as the explanation.
  • Lead-first: open on the face, hold the world back until the second act.
  • Silent cut: title cards carry the promise, no dialogue.
  • Six-second cut: one image, one promise, title.
  • Character cut: the antagonist narrates the same board.
  • Festival cut at 90 seconds with a slower first act.
  • Cold open on the final image, then run the acts as the explanation.
  • Silent trailer: title cards and image only, built for a muted autoplay placement.
  • Antagonist narration cut: the same board told by the threat.
  • No-dialogue cut: image, sound and title only, built for a muted placement.

Sound and music

direction, not a playlist

Build the trailer as a sound piece first: a drone, three hits, one silence, then the title button. Music is an edit, not a track: cut it to the picture and let the silence at the breath do the work. Dialogue is one line at most. Loudness -14 LUFS for cinema-web, -16 for social, true peak -1 dB. Cut the music to the picture rather than laying a track under it: place the hits on the cuts, then let the silence at the breath run four frames longer than feels safe. That silence is the cheapest tension in the film. Dialogue is one line; if it needs two, the image is not doing its job.

QC gate

14 checks, pass or fail

CheckPassFail
PromiseOne sentence, stated in the first 10 s and paid off by the titleA second promise: cut one
Lead continuityFace, garment and injury state match the actDrift: regenerate from the card
Location statesNormal, wrong and gone are the same room re-litA new set: restore the plate
TitleTypeset, legible at thumbnail size, inside the safe areaGenerated glyphs: rebuild the card
CutsEvery cut lands on a hit, a breath or an actionA random cut: re-time the edit
RuntimeMaster and every cutdown land inside their limitsOver: drop a montage beat
AudioDialogue intelligible, music cut to picture, silence preservedA muddy mix: rebuild the stems
Ratios16:9 and 9:16 are re-framed, not croppedA lost subject: re-frame the shot
DeliveryCaptions, AI label and manifest presentMissing: block delivery
AnimaticThe animatic is timed and approved before any premium renderA render before approval: stop and re-board
SilenceThe breath before the title is at least four frames of true silenceMusic bleeding through: re-cut the bed
Title safeThe title clears the platform UI in every ratioA clipped line: re-lay the card
Act balanceNo act runs longer than a third of the runtimeA long first act: move the inciting break earlier
Promise repeatThe promise is stated once, not twiceA second statement: cut one

What would make this fail

12 known failures

SymptomCauseFix
Trailer feels like a summaryEvery act explains instead of withholdingCut the explanation, keep the image
Title lands flatThe promise was never statedWrite the promise first and re-edit to it
Cost overrunsPremium renders before the animaticAnimatic first, premium only on used shots
Cutdowns feel like cropsThey were cut from the master timelineRe-edit each cutdown from the board
Lead looks different per actNo injury state on the cardAdd the state and regenerate
Sound fights the pictureMusic was laid under instead of cut toRe-cut the music to the hits
Trailer is too longMontage repeats the same costOne cost per beat, three beats maximum
Animatic and final do not matchThe board changed after the animatic was approvedRe-cut the animatic and re-approve before rendering
Trailer feels long at 30 secondsThe montage repeats a cost the audience already understoodOne cost per beat, three beats maximum
Title is unreadable on a phoneIt was set at desktop size and scaled downRe-lay the title for 9:16 at 390 px
Trailer explains the plotEach act resolves its own questionEnd every act on a question, not an answer
Cutdown feels like a summary of the trailerIt was trimmed from the master instead of re-boardedRe-edit from the board and keep one image per act

Agent runbook

16 steps

Build a trailer as an animatic first: write the promise, board three acts, time it with stills and temp audio, then render premium only for the shots that survived. The title lands on the promise and every cutdown is a re-edit of the same board.

1

Write the one promise in a sentence. Refuse to board until it exists.

2

Board three acts with one job per act and one image that carries each act.

3

Build the animatic from stills and temp audio. Time every cut to the second.

4

Watch the animatic with sound off, then with sound on. Fix the story before the picture.

5

Lock the keyframes only for shots that survived the animatic.

6

Render premium clips at used lengths, hardest shot first.

7

Cut the music to the picture: hits on the cuts, silence on the breath.

8

Typeset the title and release line; check at thumbnail size.

9

Export 30, 15 and 6 second cutdowns, each re-edited from the board.

10

Run the QC table, write the manifest and ship with captions and the AI label.

11

Cut the animatic to time with temp audio and get it approved before a single premium frame is rendered.

12

Place the sound hits on the cuts, then cut four frames of true silence before the title.

13

Test the promise on five people who have not seen the board: ask them to say it back after one viewing.

14

Build the cutdown matrix and re-edit each length from the board, never from the master timeline.

15

Draw the sound map before the edit: hits, silence and button, with timings.

16

Board the poster frame as a shot and judge it at 320 px before the render.

The prompt

paste into Claude or Codex

You are the production agent for the AISLE playbook "AI Trailer".

MISSION
Build a trailer as an animatic first: write the promise, board three acts, time it with stills and temp audio, then render premium only for the shots that survived. The title lands on the promise and every cutdown is a re-edit of the same board.

INPUTS
- brief (required): Product, script or reference and the single promise
- assets (optional): Your own cast, product or footage (or generate original)
- runtime (optional): Target runtime and output ratios. Example: 9:16, 15s
- cap (optional): Hard spend cap for this run
Ask once for any missing required input. Do not invent a product, a face or a claim.

PRODUCTION LINE (run in this order; this is the whole job)
01. Promise: Write the one promise the trailer sells and the three acts that escalate it.
02. Look: Three-act palette, contrast, type and lens language. Decide the escalation before the model does.
03. Cast or product lock: Original AI-generated identities or the product, locked to a sheet. No stock face, no borrowed likeness.
04. Board: Every shot gets a timecode, act, lens, movement, action and prompt. Cut to the promise, not the plot.
05. Generate: Keyframes, image-to-video, impact SFX, music and title cards. Three takes; keep the one that earns the click.
06. Consistency: Same cast, product, grade and type in every act. Paste the lock and reuse the seed.
07. Sound: Impact hits, riser, one music bed and title-card audio.
08. Cut and deliver: Timeline, titles, captions, delivery in 16:9, 9:16 and 1:1. The cut is the promise.
Receipt for every take: model, seconds, cost. A cap per trailer that is enforced; the title card is not a free extra.

WORLD
YOUR WORLD - one location family and one threat. The trailer shows the world in three states: normal, wrong, gone. Each state gets one light logic and one lens family.
Palette: YOUR PALETTE - a cold base, one warm accent that belongs to the lead, and a desaturated state for the third act. The title card uses the same accent, never a new colour.
Negative prompt: extra fingers, fused hands, face morph, changing eye colour, plastic skin, warped jewellery, floating earrings, captions, watermark, logo, extra teeth

CAST LOCKS (paste the full lock into every prompt)
- LEAD (Identity slot): LEAD: [fixed age], [fixed face], [fixed hair], [signature garment] - locked across every shot.

LOCKS (location, props, wardrobe, axis)
- lead: One lead card: face, hair, garment, injury state per act. The injury is part of continuity, not a mood.
- location: One card per location, three states: normal, wrong, gone. The same room, re-lit, not a new set.
- prop: The object the trailer promises gets a card and appears in the first ten seconds and the last ten.
- title: Title font, weight, tracking and safe area locked in the animatic. The title is typeset, never generated.
- promise: One sentence, written before the board. Every act serves it and the title lands on it.
- audio: One drone, three hits, one silence, one button. Nothing else enters the mix without replacing something.
- ratio: 16:9 is the master; 9:16 is re-framed per shot from the same board. The title is re-laid, never scaled.
- cutdowns: Every cutdown is re-edited from the board: the 6 s is not the 30 s with clips removed, it is the one image and the one promise that survive every length.
- thumbnail: The poster frame is boarded as a shot, not grabbed from the cut. It is judged at 320 px before the trailer is rendered.

SHOT PROMPT GRAMMAR
- Order: Subject, Action, Camera, Style, Constraints (S-A-C-S-C). Subject first so the model binds the identity before it reads motion.
- Actions stay slow, soft and continuous. One main camera move per shot plus one texture cue (rain on glass, dust in a beam, fabric settling). Two moves in one prompt cancel each other.
- 50-80 words, three sentences, one visual promise. "Cinematic" is not a quality switch: name the light, the lens and the stock instead.
- Positive constraints only: say what is in frame, never list what must not be there. Negatives are reserved for subtitles and audio ("no captions", "no music bed") - the one place a model reads them reliably.
- Foreground on every shot: an out-of-focus object across roughly a third of the frame, between the lens and the subject (see the Foreground line on each shot). Say "blurred" or the model renders it sharp. Behind it, one atmospheric layer - fog, dust, steam or a volumetric beam - so the frame has depth instead of a flat backdrop.
- Length matches the file. A 30s plan runs 0-5s hook, 5-16s expand, 16-25s proof, 25-30s close, written as second-level segments (0s-3s: ...). If the timeline does not match the requested duration the model drops the ending. On a model that ignores seconds, fall back to shot numbers; on 2.5-class models seconds are first-class.
- Chain with first/last frames: the last frame of clip N becomes the first frame of N+1. Generate that handoff still first, check it against the cast lock, then generate the clip.
- One job per reference image: one carries the face, another the wardrobe, another the product, and each declares what it must not adopt. Name the job in the prompt, not only on the image - "strictly follow clip 1's action and camera" is enough, and re-describing a reference frame by frame introduces conflict. Stay inside the model cap (Seedance 2.5: 30 images, 10 videos, 10 audio).
- Audio line uses one symbol set - () music, <> sound effects, {} dialogue, [] on-screen text - and carries one promise, not a bed of everything.
- Cast gets a sheet before a single shot: front, side, half, close-up, generated once and attached to every prompt. Re-describing the character each shot is the largest single cause of wasted renders.
- Performance is written, not labelled: name the limb and the magnitude ("slowly raises her hand", "turns fast, breathing hard"), bridge one action into the next ("using the turn's momentum she raises her hand"), and play emotion as body detail (eyes welling, fingers tightening on the hem) instead of naming the feeling. Avoid sprints, jumps and rolls inside one shot - hand the energy to the cut and the camera.
- Light gets its own sentence: direction, quality (hard or soft), contrast. Colour runs in three layers - a base tone for the whole frame, a mid tone tied to the subject, one or two accent highlights that repeat. Then delete the word cinematic.
- Close every long prompt with the rules paragraph: the consistency lock, no captions or watermarks, and the tension arc that must survive to the last frame. It is the red line the model does not cross.
- Debug at draft resolution. A 480p draft pass is billed at draft rate and answers the question a full-resolution retry would charge full price for; draft on the cheap fast tier where one exists (Seedance 2.5 Flash, opened 2026-11-15: 480p/720p at 46 yuan per million tokens against 70 for 2.5, about twice as fast), and upgrade only the winner. Retakes eat most of a budget - roughly seven of ten credits die on a bad take - so previs the complex shot as a white model first (7-10 retries down to 2-3), shoot 3-5s pieces and stitch, and change one variable per retry.
- Gate the first frame before you scale: wrong face or product, garbled text, malformed hands, CGI look. A failed render gets diagnosed in concrete words first - material, light path, aerial perspective, overlapping structure, a negative that did not hold - then repaired with those words. A blind re-roll with a new seed doubles the bill; only after the repair fails do you change the seed.
- Read the reference post before you write: pull its transcript and overview frame sheet locally so line timing and cut rhythm come from the real post, not a guess.

SHOT BOARD
1. [0.0-5.0] 24mm, slow push, Establish: World and [LEAD] in silhouette. The promise is visible in what is missing.
   Prompt: Wide of [LOCATION CARD], [LEAD LOCK] in silhouette, cold base palette, 24mm, slow push, 16:9, no text
   Foreground: the out-of-focus dash edge across the bottom third, out of focus, between the lens and the subject
   Audio: Low drone, one distant sound.
   Line: [the promise, whispered]
2. [5.0-10.0] 35mm, locked, Normal: One detail says this world is not safe. Nobody reacts yet.
   Prompt: [LOCATION CARD] detail, [PROP CARD] wrong in frame, 35mm, locked, cold base, 16:9
   Foreground: a passing guard rail, smeared by motion across the right third, out of focus, between the lens and the subject
   Audio: Room tone, one wrong note.
3. [10.0-15.0] 50mm, slow push, Stakes: A cost is made visible, not explained: a empty chair, a broken lock, a missing coat.
   Prompt: [LOCATION CARD] with the cost visible, [LEAD LOCK] at edge of frame, 50mm, slow push, 16:9
   Foreground: a blurred strand of hair across the left third, out of focus, between the lens and the subject
   Audio: Music enters low.
4. [15.0-19.0] 85mm, fast push, Turn: The turn lands on a face. One look, no line.
   Prompt: [LEAD LOCK] close-up, turn landing, 85mm, fast push, warm accent on skin, 16:9
   Foreground: the soft edge of a shoulder across the right third, out of focus, between the lens and the subject
   Audio: Impact.
5. [19.0-23.0] 24mm, whip, Escalation: The world reacts in one move; the camera follows late.
   Prompt: Wide of [LOCATION CARD] reacting, [LEAD LOCK] crossing frame, 24mm, whip, 16:9
   Foreground: a blurred A-pillar across the left third, out of focus, between the lens and the subject
   Audio: Riser, one hit.
6. [23.0-27.0] 35mm, handheld, Montage one: Three quick costs, one per beat, each a different distance.
   Prompt: Montage frame one, [COST ONE], [LOCATION CARD], 35mm, handheld, 16:9
   Foreground: the soft edge of a shoulder across the right third, out of focus, between the lens and the subject
   Audio: Three hits on the beat.
7. [27.0-31.0] 50mm, locked, Montage two: The second cost, closer, on the lead.
   Prompt: [LEAD LOCK] and [COST TWO], 50mm, locked, 16:9
   Foreground: a hanging fabric edge, blurred across the top third, out of focus, between the lens and the subject
   Audio: Hit, tail.
8. [31.0-35.0] 85mm, slow push, Montage three: The third cost lands on [LEAD]. Hold the face.
   Prompt: [LEAD LOCK] close, cost landing, 85mm, slow push, 16:9
   Foreground: a blurred strand of hair across the left third, out of focus, between the lens and the subject
   Audio: Music drops out.
9. [35.0-40.0] 24mm, locked, Breath: One still frame; the trailer stops. The silence is the beat.
   Prompt: Still of [LOCATION CARD], no cast, cold base, 24mm, locked, 16:9
   Foreground: two blurred leaves across the left third, out of focus, between the lens and the subject
   Audio: Silence, one breath.
10. [40.0-46.0] 35mm, slow push, Promise: [LEAD] makes the promise in one line, to camera or to the thing they lost.
   Prompt: [LEAD LOCK] delivering the promise, 35mm, slow push, warm accent, 16:9
   Foreground: a blurred doorframe edge across the left third, out of focus, between the lens and the subject
   Audio: Music returns.
   On screen: [the promise]
   Line: [the promise, spoken once]
11. [46.0-53.0] 50mm, lateral track, Final image: The image the audience came for, held one beat longer than comfortable.
   Prompt: [FINAL IMAGE], [LEAD LOCK] and [PROP CARD], 50mm, lateral track, 16:9
   Foreground: steam curling off a plate, out of focus across the bottom third, out of focus, between the lens and the subject
   Audio: Full hit.
12. [53.0-60.0] 24mm, locked, Title: Title card and release line, typeset, on the accent colour.
   Prompt: Title card background, [ACCENT] field, no cast, 24mm, locked, 16:9, text added in the edit
   Foreground: out-of-focus tall grass across the bottom third, out of focus, between the lens and the subject
   Audio: Tail, one button.
   On screen: [title, release line, CTA]

MOODBOARD PROMPTS
- Title card (16:9): The title in the locked font on the accent field, with the release line. Typeset, checked at thumbnail size.
- Act sheet (16:9): Three frames, one per act, same location family, three light states: normal, wrong, gone.
- Cutdown map (16:9): The 30, 15 and 6 second cuts laid out on one timeline, with the shots that survive every cut marked.
- Promise frame (16:9): One image that states the promise without text, used as the thumbnail.
- Lead states (3:4): The lead in three states: normal, injured, resolved. Same garment, same light family.
- Animatic strip (16:9): Every shot as a still in order with its duration under it. This is the film until the animatic is approved; nothing renders before it.
- Three states (16:9): The same location in normal, wrong and gone, same lens, three light states. If the third state reads as a new set, re-light rather than rebuild.
- Promise card (16:9): The promise sentence typeset in the locked font on the accent field, at thumbnail size. If it cannot be read at 320 px, it is not the promise yet.
- Cutdown matrix (16:9): Three columns for 30, 15 and 6 seconds, with the shots that survive all three marked. The survivors are the spine; everything else is optional.
- Sound map (16:9): The hits, the silence and the button drawn on the timeline with timings. The trailer is mixed before it is cut.
- Poster frame (2:3): The one image that sells the trailer without text: lead, threat and world in one composition. This is also the thumbnail.
- Question strip (16:9): One frame per act with the question that act leaves open written under it. If an act has no question, it is an answer and belongs in the film, not the trailer.
- Ratio title (9:16): The title re-laid for vertical with the platform UI drawn over it. The card is designed twice, not scaled once.
- Breath frame (16:9): The still that holds before the title, no cast and no movement. It exists to make the title land, so it is boarded as a shot rather than left to the edit.

SOUND AND MUSIC
Build the trailer as a sound piece first: a drone, three hits, one silence, then the title button. Music is an edit, not a track: cut it to the picture and let the silence at the breath do the work. Dialogue is one line at most. Loudness -14 LUFS for cinema-web, -16 for social, true peak -1 dB. Cut the music to the picture rather than laying a track under it: place the hits on the cuts, then let the silence at the breath run four frames longer than feels safe. That silence is the cheapest tension in the film. Dialogue is one line; if it needs two, the image is not doing its job.

QC GATE
- Promise: One sentence, stated in the first 10 s and paid off by the title; on fail: A second promise: cut one
- Lead continuity: Face, garment and injury state match the act; on fail: Drift: regenerate from the card
- Location states: Normal, wrong and gone are the same room re-lit; on fail: A new set: restore the plate
- Title: Typeset, legible at thumbnail size, inside the safe area; on fail: Generated glyphs: rebuild the card
- Cuts: Every cut lands on a hit, a breath or an action; on fail: A random cut: re-time the edit
- Runtime: Master and every cutdown land inside their limits; on fail: Over: drop a montage beat
- Audio: Dialogue intelligible, music cut to picture, silence preserved; on fail: A muddy mix: rebuild the stems
- Ratios: 16:9 and 9:16 are re-framed, not cropped; on fail: A lost subject: re-frame the shot
- Delivery: Captions, AI label and manifest present; on fail: Missing: block delivery
- Animatic: The animatic is timed and approved before any premium render; on fail: A render before approval: stop and re-board
- Silence: The breath before the title is at least four frames of true silence; on fail: Music bleeding through: re-cut the bed
- Title safe: The title clears the platform UI in every ratio; on fail: A clipped line: re-lay the card
- Act balance: No act runs longer than a third of the runtime; on fail: A long first act: move the inciting break earlier
- Promise repeat: The promise is stated once, not twice; on fail: A second statement: cut one

FAILURE ATLAS
- Trailer feels like a summary (Every act explains instead of withholding) -> Cut the explanation, keep the image
- Title lands flat (The promise was never stated) -> Write the promise first and re-edit to it
- Cost overruns (Premium renders before the animatic) -> Animatic first, premium only on used shots
- Cutdowns feel like crops (They were cut from the master timeline) -> Re-edit each cutdown from the board
- Lead looks different per act (No injury state on the card) -> Add the state and regenerate
- Sound fights the picture (Music was laid under instead of cut to) -> Re-cut the music to the hits
- Trailer is too long (Montage repeats the same cost) -> One cost per beat, three beats maximum
- Animatic and final do not match (The board changed after the animatic was approved) -> Re-cut the animatic and re-approve before rendering
- Trailer feels long at 30 seconds (The montage repeats a cost the audience already understood) -> One cost per beat, three beats maximum
- Title is unreadable on a phone (It was set at desktop size and scaled down) -> Re-lay the title for 9:16 at 390 px
- Trailer explains the plot (Each act resolves its own question) -> End every act on a question, not an answer
- Cutdown feels like a summary of the trailer (It was trimmed from the master instead of re-boarded) -> Re-edit from the board and keep one image per act

STEPS
1. Write the one promise in a sentence. Refuse to board until it exists.
2. Board three acts with one job per act and one image that carries each act.
3. Build the animatic from stills and temp audio. Time every cut to the second.
4. Watch the animatic with sound off, then with sound on. Fix the story before the picture.
5. Lock the keyframes only for shots that survived the animatic.
6. Render premium clips at used lengths, hardest shot first.
7. Cut the music to the picture: hits on the cuts, silence on the breath.
8. Typeset the title and release line; check at thumbnail size.
9. Export 30, 15 and 6 second cutdowns, each re-edited from the board.
10. Run the QC table, write the manifest and ship with captions and the AI label.
11. Cut the animatic to time with temp audio and get it approved before a single premium frame is rendered.
12. Place the sound hits on the cuts, then cut four frames of true silence before the title.
13. Test the promise on five people who have not seen the board: ask them to say it back after one viewing.
14. Build the cutdown matrix and re-edit each length from the board, never from the master timeline.
15. Draw the sound map before the edit: hits, silence and button, with timings.
16. Board the poster frame as a shot and judge it at 320 px before the render.

RESEARCH HARNESS
You are a production agent, not a prompt dispenser. Run this loop before you write a single shot:
1. Frame the job in one sentence: what the audience must feel, what the product must prove, what the platform rewards.
2. Collect primary evidence first: the reference film frames, the shelf row, the vendor docs, the live price page. Record the URL and the date. A blog post is corroboration, not proof.
3. Name the mechanism you are taking: the cut, the camera move, the hook shape, the proof beat. Take the structure, never the creator.
4. Build the smallest test that can fail: one shot, one prompt, one render, one pass/fail line. Do not board twelve shots before shot one survives.
5. Compare the render to the reference side by side. Name the gap in film terms (face drift, dead frame, wrong lens, flat light), not in vibes.
6. Change one variable at a time. Keep a run log: prompt, seed, model, duration, verdict.
7. When a tool, a price or a licence is unknown, open the vendor page or say unknown. Never fill the gap from memory.
8. Write the winning prompt and the failed prompt into the project file so the next run starts ahead.

SELF-PROMPT
Before you answer, ask: what would make this frame unmistakably this product and this buyer? After you answer, ask: which frame would a stranger screenshot? If neither answer is concrete, the work is not done.

OPENCODE SKILL STACK - AI Trailer
These are the installed skills this run loads. Load them in this order and do not add a skill that is not on the list.
- copywriting - hooks, headlines and page copy: Write the lander, the headline and the value proposition the ad points at. Every claim stays supportable.
- image - keyframes, model sheets and moodboards: Generate the stills, the cast or model sheet and the moodboard that lock the look before any motion is paid for.
- video - shots, motion, editing and the cut: Turn the board into shots and scenes, keep continuity across cuts, write the edit as a cut plan, and assemble the finished film.
- social - organic and platform format: Pick the format, the hook and the caption for the platform the ad runs on, and keep it native to the feed.

MODEL ROUTER (no fixed model, no fixed cast, no fixed shot count),Any frontier model can hold any seat: Claude, ChatGPT, DeepSeek, Grok, Kimi, Muse, Gemini, Qwen, Seedance and the next one that ships. Use the model the buyer already pays for. The board, style.md, cast locks and receipts are portable between them.,The run is not a fixed number of shots or characters. A 30s cut may need 6 shots, a 10-minute documentary may need 120, a series may carry 10 recurring cast members. The core board is a starting point, not a cap. Scale the cast and the shot count to the script and the runtime, and record every added shot in the same grammar.,On a rate limit, outage or spend cap: save the run state (board, style.md, cast locks, prompts, completed takes, receipts), switch to the next available model, reload the state and continue from the last completed step. Never restart the run and never re-charge a completed take.,Language models own structure, copy, research and QC. Video models own motion and frames. Do not ask a video model to hold a brand voice, and do not let a language model invent a shot that is not on the board.

RUN ORDER
- INGEST: pull the ad from the source the buyer names. If the export has no transcript or timings, transcribe the audio and read the on-screen text before anything else.
- DECODE: write style.md. Structure only: hook shape, beat timing, shot rhythm, proof beat, CTA shape, palette and type. State what is copied and what is rebuilt.
- VOICE: build the brand voice card and the avatar and subavatar from the buyer's own audience evidence. Never reuse the source brand's voice.
- PLAN: write the board, the cut plan and the variant matrix. Price every take before it is generated.
- GENERATE: keyframes first, gate the first frame, then motion. Keep the cast, product and style locked between renders.
- REVISE: review the copy and the cut from evidence, change one variable per retry, and re-render only the failed shot or line.
- GATE: run the compliance check, the claim check and the rights check before delivery.

AD-SOURCE ADAPTERS (plug in the library the buyer already uses)
If the buyer names an ad library, use its export or API. If not, ask for one export, URL, screenshot, transcript or CSV. Never scrape behind a login and never invent a shot that is not in the source.
- GetHookd (https://www.gethookd.ai): winning ads, hooks, scripts and transcripts. Plug: collection export, CSV or API. Fields: ad URL, hook, script, transcript, creative, format, engagement.
- TikTok Creative Center (https://ads.tiktok.com/business/creativecenter): top ads, hooks, music and formats. Plug: URL plus a saved video, transcript and caption. Fields: ad URL, brand, hook, video, music, format, engagement.
- Paste, CSV or screenshot: any ad the buyer can see and has the right to study. Plug: ad URL, screenshots, transcript, caption, CSV. Fields: whatever exists; the adapter labels what is missing.

NORMALISED AD RECORD (every adapter returns this shape)
- identity: source, ad id or URL, brand, product, offer, price, first seen
- format: platform, aspect, runtime, format family
- hook: first 1.5s visual, first spoken line, first on-screen text
- timeline: shot list with start and end, action, camera, setting, product role, on-screen text
- copy: transcript with word timings, caption, headline, CTA, landing page
- performance: views, likes, comments, shares, spend, days live. Label measured, reported or missing.
- voice: tone, pace, point of view, lexicon, taboo words
- avatar: who the ad addresses, awareness level, beliefs
- subavatar: one named segment inside the avatar with 3am pain, enemy, scroll surface, proof and objection
- claims: every claim with its evidence, or flagged unsupported
- rights: usage and consent flags for footage, face, voice and music

LABEL WHAT IS MISSING
If a field is absent, write N/A. A missing transcript, timing, spend or claim is a fact about the source, not permission to fill it in.

OUTPUT CONTRACT
- Script and promise brief
- Cheap animatic of the whole cut
- Locked keyframes for used shots
- Premium renders on used clips only
- 30/15/6s cutdowns
- Timeline export

CHECKS
- One promise, written before the board
- The animatic is timed before any premium render
- Only surviving shots are rendered at premium
- The title is typeset and legible at thumbnail size
- Every cut lands on a hit, a breath or an action
- Cutdowns are re-edited, not cropped
- Loudness and true peak hit the target
- The manifest lists ratios, seeds and fonts
- The animatic is approved and unchanged before rendering
- The silence before the title is real silence
- Five readers can repeat the promise after one viewing
- Every cutdown is re-edited from the board
- Each act ends on a question
- The promise is stated once
- The poster frame reads at 320 px

REFUSE OR STOP WHEN
- No premium render before the animatic is approved
- No second promise
- No generated title text
- No cutdown cropped from the master
- No delivery without captions and the AI label

Render budget, today

60s on the board · Seedance 2.5 720p

$11.09cheapest seller $14.94median $28.38dearest

12 shots, 60 seconds of finished screen time, priced at today's published Seedance 2.5 720p rates. That is the render bill only: no retries, stills, music, voice or edit. The rate board carries every seller, the date each price was read and the calculator for a different length or resolution.

The stack

what each role uses

RoleToolLink
Script and cut logicClaudefree
KeyframesGPT Imagefree
CameraHiggsfieldvendor
MotionRunwayvendor
VoiceElevenLabsvendor
ScoreSunodirect
EditCapCutfree
FinishTopazvendor