Studio · Song VSL · 9:16 · 12-stage journey
Drama Song Ad
Build the Ryze/Resilia-style format properly: a first-person hero journey sung end to end, a word timing map, a locked four-view cast sheet and clips whose length comes from the timing map, never a fixed number. The product appears after the halfway mark and the pitch owns the last minute.
The whole line, one pack
story to stitch
The board below is the core, not a ceiling; extend it until the story lands.
What you get
10 deliverables
- First-person lyric script, 1,100-1,500 words, short lines built to be sung
- Two song generations from one style prompt, crossfaded at the chorus
- Word timing map: start and end seconds for every sung line
- Four-view cast sheet per character, plus a new sheet after the transformation
- Shot sheet: one image per four frames, setting once and camera per frame
- Silent clips cut to the timing map, never to a fixed length
- Stitched cut with the song as the only audio, captions last
- QC report on identity, lip sync, caption timing and the vindication reversal
- Lyric sheet with syllable counts per line, so the singer never rushes a word
- Chorus map: which line is the hook, which is the turn, which is the payoff
World and look
paste before anything
| World | YOUR WORLD - paste the setting for each scene and keep the description identical in every image and video prompt. The world is ordinary and specific: a kitchen, a salon, a night bus. The transformation happens in the same world, not in a fantasy set. |
| Palette | YOUR PALETTE - one 3D-animated-feature look held across every frame. Lock the style sentence once and reuse it verbatim. The hero and the betrayer must be separable by colour alone in a thumbnail. |
| Negative | extra fingers, fused hands, face morph, changing eye colour, plastic skin, warped jewellery, floating earrings, captions, watermark, logo, extra teeth |
Locks
location, props, wardrobe, axis
| location | One card per scene: the room, the light, the one object the song returns to. The vindication shot must reuse the humiliation location exactly, same lens, same light. |
| props | The product is a prop with a card: size, label, how it is held. It does not appear before the halfway mark and it never changes shape after it. |
| wardrobe | The hero has two states: before and after. Both are written on the cast sheet. The betrayer keeps one garment so the reversal reads instantly. |
| voice | One sung voice, one register. The mentor may double the hero an octave down; nobody else sings. |
| axis | The humiliation and the vindication are the same setup, so the axis is fixed once and reused. Do not re-stage the room for the payoff. |
| syllables | Count syllables per line and keep them inside the range the song can carry. A line that reads well but cannot be sung is a script bug, not a performance problem. |
| colour | The hero's transformation is carried by one colour shift on one garment. If the whole palette changes, the audience reads a new character. |
Every lock is pasted into the prompt for every shot that touches it. A lock that is paraphrased is not a lock.
Cast locks
3 identities
HERO
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.
MENTOR
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.
BETRAYER
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.
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
5 skills, in order
| Skill | Job in this run |
|---|---|
avatar-research | brand voice, avatar and subavatar |
copywriting | hooks, headlines and page copy |
image | keyframes, model sheets and moodboards |
video | shots, motion, editing and the cut |
social | organic 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
| Source | What it gives |
|---|---|
| GetHookd | winning ads, hooks, scripts and transcripts |
| TikTok Creative Center | top ads, hooks, music and formats |
| Paste, CSV or screenshot | any ad the buyer can see and has the right to study |
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
0.0-3.5s
35mm
slow push
Ordinary world — [HERO] in the ordinary world before the insult. The room is the room the payoff returns to.
[HERO LOCK] in [LOCATION CARD], ordinary morning, one window light, 35mm, slow push, 3D-animated feature look, vertical 9:16, no text
foregroundrain-beaded glass edge across the left third, out of focus, between the lens and the subject
audioSong intro, room tone underneath.
line[first sung line: name the ordinary day]
3.5-7.5s
50mm
locked
Humiliation — [BETRAYER] delivers the line; [HERO] freezes. The insult is heard, not narrated.
[BETRAYER LOCK] speaking to [HERO LOCK], two-shot, same room, 50mm, locked, 3D-animated feature look, vertical 9:16
foregrounda curtain fold, out of focus across the right third, out of focus, between the lens and the subject
audioSong lifts on the insult.
line[sung: the insult, word for word]
7.5-11.5s
85mm
slow push
Wound — [HERO] plays the wound as body detail: a hand, a swallow, a step back.
[HERO LOCK] close-up, held expression, no tears, 85mm, slow push, 3D-animated feature look, vertical 9:16
foregroundthe soft edge of a shoulder across the right third, out of focus, between the lens and the subject
audioSong drops to one instrument.
line[sung: what the insult cost]
11.5-15.0s
24mm
handheld
Failed solution — A product that does not work, shown honestly, with the receipt in frame.
[HERO LOCK] trying [FAILED PRODUCT], cluttered room, 24mm, handheld, 3D-animated feature look, vertical 9:16
foregroundthe out-of-focus shoulder of a chair back across the right third, out of focus, between the lens and the subject
audioComic beat on the failure, then silence.
line[sung: the money and time it cost]
15.0-18.5s
50mm
locked
Mentor — [MENTOR] enters, calm and credible, before any product is named.
[MENTOR LOCK] entering [LOCATION CARD], warm key, 50mm, locked, 3D-animated feature look, vertical 9:16
foregrounda passing guard rail, smeared by motion across the right third, out of focus, between the lens and the subject
audioSong opens up, one new instrument.
line[sung: the truth the mentor gives]
18.5-22.0s
35mm
lateral track
Doubt — [HERO] almost refuses the gift. The pause is the turn.
[HERO LOCK] looking at [PRODUCT CARD] in [MENTOR LOCK] hand, 35mm, lateral track, 3D-animated feature look, vertical 9:16
foregroundthe soft edge of a shoulder across the right third, out of focus, between the lens and the subject
audioSong thins to voice.
line[sung: the doubt]
22.0-26.0s
35mm
montage
The climb — Three quick states of change, same room, same light, wardrobe shifts with the hero.
[HERO LOCK] three states in [LOCATION CARD], same light, wardrobe changes per state, 35mm, montage frames, 3D-animated feature look, vertical 9:16
foregrounda curtain fold, out of focus across the right third, out of focus, between the lens and the subject
audioSong builds, drums enter.
line[sung: week by week]
26.0-30.0s
50mm
locked
Vindication — The opening insult reverses in the same framing, same room, same lens as shot 02.
[HERO LOCK] and [BETRAYER LOCK] same two-shot as shot 02, roles reversed, 50mm, locked, 3D-animated feature look, vertical 9:16
foregrounda curtain fold, out of focus across the right third, out of focus, between the lens and the subject
audioSong peaks on the reversal.
line[sung: the insult returned word for word]
30.0-34.0s
24mm
slow push
Product hero — Product handed over after the halfway mark, in the mentor's hand, then the hero's.
[PRODUCT CARD] passed from [MENTOR LOCK] to [HERO LOCK], 24mm, slow push, 3D-animated feature look, vertical 9:16
foregrounda blurred strand of hair across the left third, out of focus, between the lens and the subject
audioSong holds, no new element.
line[sung: the product named once]
34.0-38.0s
50mm
locked
Return — The hero in the transformed room, same window light, one new object that proves the change.
[HERO LOCK] in [LOCATION CARD] after transformation, same window light, one new object, 50mm, locked, 3D-animated feature look, vertical 9:16
foregroundthe blurred rim of a shop window across the right third, out of focus, between the lens and the subject
audioSong resolves.
line[sung: the elixir]
38.0-42.0s
85mm
slow push
Offer — Product on a clean surface, label readable, price typeset by the editor.
[PRODUCT CARD] on clean surface, label readable, soft key, 85mm, slow push, 3D-animated feature look, vertical 9:16, text added in edit
foregroundan out-of-focus fingertip entering frame across the bottom third, out of focus, between the lens and the subject
audioSong tail, one button.
on screenOffer line, typeset in the edit
42.0-45.0s
50mm
locked
End card — The hero and the product in one frame, no new information.
[HERO LOCK] and [PRODUCT CARD] end card, same palette, 50mm, locked, 3D-animated feature look, vertical 9:16
foregrounda passing guard rail, smeared by motion across the right third, out of focus, between the lens and the subject
audioSong ends on the chorus, no outro.
on screenProduct name and CTA, typeset
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
10 beats
| Time | Beat | Job |
|---|---|---|
| 0-5% | Ordinary world | A few seconds of normal before the insult. Establish the room the payoff will return to. |
| 5-15% | Humiliation | The opening insult, delivered as another character's dialogue, never as self-description. |
| 15-25% | Wound deepens | Bad becomes worse. The hero reacts in body, not in a speech. |
| 25-35% | Frozen | The hero stops trying. Show the stall, not the sadness. |
| 35-45% | Failed solutions | Two things that do not work, shown honestly, with the money and time they cost. |
| 45-55% | Mentor and gift | A trusted source enters before the product is handed over. The gift is a truth, not a discount. |
| 55-65% | Doubt | The hero almost refuses. The doubt makes the climb believable. |
| 65-80% | The climb | Week by week change; other characters notice before the hero does. |
| 80-92% | Vindication | The exact reversal of the opening insult, in the same framing and the same room. |
| 92-100% | Return with the elixir | Product and offer land in the final minute only, as the reward for the story. |
Moodboard prompts
7 images
Hooks and variants
9 openers
- Open on the insult and play the ordinary world as a flashback inside the first chorus.
- Make the betrayer the narrator: the story is told by the person who caused the wound.
- Cut the mentor entirely and let the product arrive as the hero's own decision.
- Song VSL at 5 minutes for cold traffic, plus a 60-second chorus cut for paid.
- Male and female vocal versions of the same lyric, same timing map, two audiences.
- Replace the transformation with a skill the hero learns; the product becomes the tool, not the cure.
- End on the betrayer's reaction instead of the hero's victory, so the loop back to the opening insult is tighter.
- Duet cut: the mentor sings the doubt, the hero sings the climb, same timing map.
- Acoustic cut: one instrument, slower tempo, the same lyric and the same reversal.
Sound and music
direction, not a playlist
The song is the only audio: no dialogue stem, no sound effects, no room tone in the final mix. Generate two takes from one style prompt, keep the chorus free of an outro and the vocal free of an intro, and crossfade on the chorus. Cut every clip to the word timing map, trim the first and last flash frame, and add captions last. Loudness -14 LUFS for music-led vertical, true peak -1 dB. Keep the lyric sheet beside the timeline: every clip is cut to a line, and every line has a start and an end. When a line is re-sung, the map changes and the clips that sit on it are re-cut, not stretched. Never time-stretch a clip to fit a new take; the mouth will drift and no amount of grading hides it.
QC gate
12 checks, pass or fail
| Check | Pass | Fail |
|---|---|---|
| Identity | Four-view sheet matches in every shot; new sheet after the transformation | Any drift: regenerate from the sheet, never from the previous clip |
| Style | The style sentence is pasted verbatim and the look does not change between scenes | A new look: discard the take |
| Timing | Every clip length comes from the word timing map | A fixed-length clip: re-cut to the map |
| Lip sync | Mouth closes on the sung syllable, not after it | Offset over two frames: re-generate or trim |
| Reversal | Vindication uses the humiliation framing, lens and room | A new setup: re-board the shot |
| Product timing | Product first appears after the halfway mark | An early appearance: cut the shot |
| Captions | Word-level captions match the sung line, checked against the script | Drift: rebuild from the script, not from ASR |
| Audio | Song only, no room tone, no clipping, no gap at the crossfade | A gap: re-crossfade on the chorus |
| Runtime | Full cut and paid cut both land inside their platform limits | Over: cut a verse, never the vindication |
| Disclosure | AI label and music disclosure present on the master | Missing: block delivery |
| Syllables | Every line lands inside the syllable range the melody can carry | A rushed line: rewrite the lyric, do not speed the vocal |
| Colour shift | The transformation is one garment colour and nothing else | A full palette change: re-grade to the sheet |
What would make this fail
10 known failures
| Symptom | Cause | Fix |
|---|---|---|
| Song and picture drift apart | Clips were cut to a fixed length instead of the timing map | Rebuild the map from the script and re-cut every clip |
| Whisper mishears the lyric | ASR was run on sung audio | Use the script words for the map, never the transcript |
| Face changes after the transformation | A new sheet was not generated | Make the second sheet before shooting the climb |
| Mouth does not match | Clip starts before the sung line | Trim to the word start, then check two frames either side |
| Vindication feels weak | The insult was not re-staged exactly | Reuse the humiliation setup and reverse the roles |
| Product feels bolted on | It appeared before the mentor or before the halfway mark | Move the first appearance later and give the mentor the truth first |
| Captions lag | They were built from ASR instead of the script | Rebuild from the script and the timing map |
| Cost blows past the cap | Every shot generated at premium before the song was locked | Lock the song and the map first, then generate only the used frames |
| Singer rushes the back half | Lines were written longer than the melody can carry | Cut syllables from the line and re-sing |
| Transformation reads as a new character | The palette changed with the wardrobe | Hold the palette and change one garment colour only |
Agent runbook
14 steps
Produce a sung hero-journey ad as a timed production: script, song, timing map, cast sheets, shot sheets, silent clips cut to the map, then one stitched cut with captions last. The product appears after the halfway mark and never before the mentor has given the truth.
Write the lyric script in first person, short lines, 1,100-1,500 words. Read it aloud and cut any line that cannot be sung.
Check the script against the twelve-stage journey. Every stage must be present or deliberately cut, and the humiliation and vindication must use the same words.
Generate the song twice from one style prompt with the lyrics verbatim. Keep the chorus free of an outro and the vocal free of an intro.
Crossfade the two takes on the chorus and confirm the result is one continuous performance.
Build the word timing map from the script: start and end seconds for every line. Never use ASR on sung audio.
Draw the cast sheets: four views per character, description pasted word for word, new sheet after the transformation.
Build one shot sheet per scene as a four-frame image. Generate one sheet and inspect it before making the rest.
Generate silent clips in small batches, hardest shot first, each length taken from the timing map.
Trim the first and last flash or still frame from every clip before it enters the edit.
Stitch early with the song as the only audio and watch the whole cut before fixing any single shot.
Add captions last, from the script and the timing map, and check them at 390 px.
Run the QC table, export the master and the paid cut, and write the AI and music disclosures.
Count syllables per line and mark the ones that cannot be sung, then rewrite before generating the song.
Keep the lyric sheet open beside the timeline and re-cut clips when a line is re-sung.
The prompt
paste into Claude or Codex
You are the production agent for the AISLE playbook "Drama Song Ad".
MISSION
Produce a sung hero-journey ad as a timed production: script, song, timing map, cast sheets, shot sheets, silent clips cut to the map, then one stitched cut with captions last. The product appears after the halfway mark and never before the mentor has given the truth.
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. Song and script: Lock the track, the lyric and the story the song sells. One sentence for the turn.
02. Timing map: Mark intro, verse, chorus, drop and button against the waveform. Every cut lands on a beat.
03. Performer lock: One original performer: face sheet, wardrobe, voice, performance behaviour. No stock face, no borrowed likeness.
04. Look: Stage, palette, light cue and lens language. The visual hook must survive the first 15 seconds.
05. Board: Every shot gets a beat, timecode, lens, movement, action and prompt. Board to the track, not the vibe.
06. Generate: Keyframes, image-to-video, lip-sync, lights and SFX. Three takes; keep the one that lands on the beat.
07. Consistency: Same performer, wardrobe, stage and light in every shot. Paste the lock and reuse the seed.
08. Cut and deliver: Beat sync, sound design, grade, captions and delivery in 9:16, 16:9 and 1:1. The cut is the performance.
Receipt for every take: model, seconds, cost. The track is the clock; no shot survives if it misses the beat.
WORLD
YOUR WORLD - paste the setting for each scene and keep the description identical in every image and video prompt. The world is ordinary and specific: a kitchen, a salon, a night bus. The transformation happens in the same world, not in a fantasy set.
Palette: YOUR PALETTE - one 3D-animated-feature look held across every frame. Lock the style sentence once and reuse it verbatim. The hero and the betrayer must be separable by colour alone in a thumbnail.
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)
- HERO (Identity slot): HERO: [fixed age], [fixed face], [fixed hair], [signature garment], front / three-quarter / full-body / close-up sheets; new sheet after the transformation.
- MENTOR (Identity slot): MENTOR: [fixed age], [fixed face], [fixed hair], [signature garment] - the credible complete version of the hero, appears before the product.
- BETRAYER (Identity slot): BETRAYER: [fixed age], [fixed face], [fixed hair], [signature garment] - the closest person to the hero.
LOCKS (location, props, wardrobe, axis)
- location: One card per scene: the room, the light, the one object the song returns to. The vindication shot must reuse the humiliation location exactly, same lens, same light.
- props: The product is a prop with a card: size, label, how it is held. It does not appear before the halfway mark and it never changes shape after it.
- wardrobe: The hero has two states: before and after. Both are written on the cast sheet. The betrayer keeps one garment so the reversal reads instantly.
- voice: One sung voice, one register. The mentor may double the hero an octave down; nobody else sings.
- axis: The humiliation and the vindication are the same setup, so the axis is fixed once and reused. Do not re-stage the room for the payoff.
- syllables: Count syllables per line and keep them inside the range the song can carry. A line that reads well but cannot be sung is a script bug, not a performance problem.
- colour: The hero's transformation is carried by one colour shift on one garment. If the whole palette changes, the audience reads a new character.
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. [0.0-3.5] 35mm, slow push, Ordinary world: [HERO] in the ordinary world before the insult. The room is the room the payoff returns to.
Prompt: [HERO LOCK] in [LOCATION CARD], ordinary morning, one window light, 35mm, slow push, 3D-animated feature look, vertical 9:16, no text
Foreground: rain-beaded glass edge across the left third, out of focus, between the lens and the subject
Audio: Song intro, room tone underneath.
Line: [first sung line: name the ordinary day]
2. [3.5-7.5] 50mm, locked, Humiliation: [BETRAYER] delivers the line; [HERO] freezes. The insult is heard, not narrated.
Prompt: [BETRAYER LOCK] speaking to [HERO LOCK], two-shot, same room, 50mm, locked, 3D-animated feature look, vertical 9:16
Foreground: a curtain fold, out of focus across the right third, out of focus, between the lens and the subject
Audio: Song lifts on the insult.
Line: [sung: the insult, word for word]
3. [7.5-11.5] 85mm, slow push, Wound: [HERO] plays the wound as body detail: a hand, a swallow, a step back.
Prompt: [HERO LOCK] close-up, held expression, no tears, 85mm, slow push, 3D-animated feature look, vertical 9:16
Foreground: the soft edge of a shoulder across the right third, out of focus, between the lens and the subject
Audio: Song drops to one instrument.
Line: [sung: what the insult cost]
4. [11.5-15.0] 24mm, handheld, Failed solution: A product that does not work, shown honestly, with the receipt in frame.
Prompt: [HERO LOCK] trying [FAILED PRODUCT], cluttered room, 24mm, handheld, 3D-animated feature look, vertical 9:16
Foreground: the out-of-focus shoulder of a chair back across the right third, out of focus, between the lens and the subject
Audio: Comic beat on the failure, then silence.
Line: [sung: the money and time it cost]
5. [15.0-18.5] 50mm, locked, Mentor: [MENTOR] enters, calm and credible, before any product is named.
Prompt: [MENTOR LOCK] entering [LOCATION CARD], warm key, 50mm, locked, 3D-animated feature look, vertical 9:16
Foreground: a passing guard rail, smeared by motion across the right third, out of focus, between the lens and the subject
Audio: Song opens up, one new instrument.
Line: [sung: the truth the mentor gives]
6. [18.5-22.0] 35mm, lateral track, Doubt: [HERO] almost refuses the gift. The pause is the turn.
Prompt: [HERO LOCK] looking at [PRODUCT CARD] in [MENTOR LOCK] hand, 35mm, lateral track, 3D-animated feature look, vertical 9:16
Foreground: the soft edge of a shoulder across the right third, out of focus, between the lens and the subject
Audio: Song thins to voice.
Line: [sung: the doubt]
7. [22.0-26.0] 35mm, montage, The climb: Three quick states of change, same room, same light, wardrobe shifts with the hero.
Prompt: [HERO LOCK] three states in [LOCATION CARD], same light, wardrobe changes per state, 35mm, montage frames, 3D-animated feature look, vertical 9:16
Foreground: a curtain fold, out of focus across the right third, out of focus, between the lens and the subject
Audio: Song builds, drums enter.
Line: [sung: week by week]
8. [26.0-30.0] 50mm, locked, Vindication: The opening insult reverses in the same framing, same room, same lens as shot 02.
Prompt: [HERO LOCK] and [BETRAYER LOCK] same two-shot as shot 02, roles reversed, 50mm, locked, 3D-animated feature look, vertical 9:16
Foreground: a curtain fold, out of focus across the right third, out of focus, between the lens and the subject
Audio: Song peaks on the reversal.
Line: [sung: the insult returned word for word]
9. [30.0-34.0] 24mm, slow push, Product hero: Product handed over after the halfway mark, in the mentor's hand, then the hero's.
Prompt: [PRODUCT CARD] passed from [MENTOR LOCK] to [HERO LOCK], 24mm, slow push, 3D-animated feature look, vertical 9:16
Foreground: a blurred strand of hair across the left third, out of focus, between the lens and the subject
Audio: Song holds, no new element.
Line: [sung: the product named once]
10. [34.0-38.0] 50mm, locked, Return: The hero in the transformed room, same window light, one new object that proves the change.
Prompt: [HERO LOCK] in [LOCATION CARD] after transformation, same window light, one new object, 50mm, locked, 3D-animated feature look, vertical 9:16
Foreground: the blurred rim of a shop window across the right third, out of focus, between the lens and the subject
Audio: Song resolves.
Line: [sung: the elixir]
11. [38.0-42.0] 85mm, slow push, Offer: Product on a clean surface, label readable, price typeset by the editor.
Prompt: [PRODUCT CARD] on clean surface, label readable, soft key, 85mm, slow push, 3D-animated feature look, vertical 9:16, text added in edit
Foreground: an out-of-focus fingertip entering frame across the bottom third, out of focus, between the lens and the subject
Audio: Song tail, one button.
On screen: Offer line, typeset in the edit
12. [42.0-45.0] 50mm, locked, End card: The hero and the product in one frame, no new information.
Prompt: [HERO LOCK] and [PRODUCT CARD] end card, same palette, 50mm, locked, 3D-animated feature look, vertical 9:16
Foreground: a passing guard rail, smeared by motion across the right third, out of focus, between the lens and the subject
Audio: Song ends on the chorus, no outro.
On screen: Product name and CTA, typeset
MOODBOARD PROMPTS
- Cast sheet (3:4): Four views per character: face close-up, full body front, two three-quarter views. Same light, same background, description pasted word for word. New sheet after the transformation.
- Shot sheet (16:9): One image holding four frames: the setting once, then camera and action per frame. The model sheet is a storyboard and must never appear in the video.
- Style lock (16:9): One frame that fixes the 3D-animated-feature look: material response, light wrap, depth of field. Reuse the sentence verbatim in every prompt.
- Reversal pair (9:16): The humiliation frame and the vindication frame side by side, same lens, same room, roles reversed.
- End card (9:16): Product, hero and offer space in one composition. Leave the lower third empty for typeset text.
- Lyric sheet (16:9): The lyric as a page with line numbers and syllable counts in the margin. This is the timing document the edit is built from.
- Colour shift (3:4): The hero before and after, same pose, one garment colour changed. The transformation must read at thumbnail size.
SOUND AND MUSIC
The song is the only audio: no dialogue stem, no sound effects, no room tone in the final mix. Generate two takes from one style prompt, keep the chorus free of an outro and the vocal free of an intro, and crossfade on the chorus. Cut every clip to the word timing map, trim the first and last flash frame, and add captions last. Loudness -14 LUFS for music-led vertical, true peak -1 dB. Keep the lyric sheet beside the timeline: every clip is cut to a line, and every line has a start and an end. When a line is re-sung, the map changes and the clips that sit on it are re-cut, not stretched. Never time-stretch a clip to fit a new take; the mouth will drift and no amount of grading hides it.
QC GATE
- Identity: Four-view sheet matches in every shot; new sheet after the transformation; on fail: Any drift: regenerate from the sheet, never from the previous clip
- Style: The style sentence is pasted verbatim and the look does not change between scenes; on fail: A new look: discard the take
- Timing: Every clip length comes from the word timing map; on fail: A fixed-length clip: re-cut to the map
- Lip sync: Mouth closes on the sung syllable, not after it; on fail: Offset over two frames: re-generate or trim
- Reversal: Vindication uses the humiliation framing, lens and room; on fail: A new setup: re-board the shot
- Product timing: Product first appears after the halfway mark; on fail: An early appearance: cut the shot
- Captions: Word-level captions match the sung line, checked against the script; on fail: Drift: rebuild from the script, not from ASR
- Audio: Song only, no room tone, no clipping, no gap at the crossfade; on fail: A gap: re-crossfade on the chorus
- Runtime: Full cut and paid cut both land inside their platform limits; on fail: Over: cut a verse, never the vindication
- Disclosure: AI label and music disclosure present on the master; on fail: Missing: block delivery
- Syllables: Every line lands inside the syllable range the melody can carry; on fail: A rushed line: rewrite the lyric, do not speed the vocal
- Colour shift: The transformation is one garment colour and nothing else; on fail: A full palette change: re-grade to the sheet
FAILURE ATLAS
- Song and picture drift apart (Clips were cut to a fixed length instead of the timing map) -> Rebuild the map from the script and re-cut every clip
- Whisper mishears the lyric (ASR was run on sung audio) -> Use the script words for the map, never the transcript
- Face changes after the transformation (A new sheet was not generated) -> Make the second sheet before shooting the climb
- Mouth does not match (Clip starts before the sung line) -> Trim to the word start, then check two frames either side
- Vindication feels weak (The insult was not re-staged exactly) -> Reuse the humiliation setup and reverse the roles
- Product feels bolted on (It appeared before the mentor or before the halfway mark) -> Move the first appearance later and give the mentor the truth first
- Captions lag (They were built from ASR instead of the script) -> Rebuild from the script and the timing map
- Cost blows past the cap (Every shot generated at premium before the song was locked) -> Lock the song and the map first, then generate only the used frames
- Singer rushes the back half (Lines were written longer than the melody can carry) -> Cut syllables from the line and re-sing
- Transformation reads as a new character (The palette changed with the wardrobe) -> Hold the palette and change one garment colour only
STEPS
1. Write the lyric script in first person, short lines, 1,100-1,500 words. Read it aloud and cut any line that cannot be sung.
2. Check the script against the twelve-stage journey. Every stage must be present or deliberately cut, and the humiliation and vindication must use the same words.
3. Generate the song twice from one style prompt with the lyrics verbatim. Keep the chorus free of an outro and the vocal free of an intro.
4. Crossfade the two takes on the chorus and confirm the result is one continuous performance.
5. Build the word timing map from the script: start and end seconds for every line. Never use ASR on sung audio.
6. Draw the cast sheets: four views per character, description pasted word for word, new sheet after the transformation.
7. Build one shot sheet per scene as a four-frame image. Generate one sheet and inspect it before making the rest.
8. Generate silent clips in small batches, hardest shot first, each length taken from the timing map.
9. Trim the first and last flash or still frame from every clip before it enters the edit.
10. Stitch early with the song as the only audio and watch the whole cut before fixing any single shot.
11. Add captions last, from the script and the timing map, and check them at 390 px.
12. Run the QC table, export the master and the paid cut, and write the AI and music disclosures.
13. Count syllables per line and mark the ones that cannot be sung, then rewrite before generating the song.
14. Keep the lyric sheet open beside the timeline and re-cut clips when a line is re-sung.
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 - Drama Song Ad
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.
- avatar-research - brand voice, avatar and subavatar: Mine reviews, forums, comments, ad comments and transcripts for the buyer's own words, the 3am pains, the objections and the unique mechanism. Build the brand voice card and the subavatar before a line is written.
- 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
- Hero-journey script
- Song and timing map
- Four-view cast sheets
- Four-frame shot sheets
- Silent clips sized to the map
- Assembled cut with captions
- Compliance pass
CHECKS
- The lyric is first person and every line can be sung
- The timing map comes from the script, not from ASR
- Cast sheets exist before any clip is generated
- Every clip length is read from the timing map
- The product first appears after the halfway mark
- The vindication reuses the humiliation framing
- Captions are built from the script and checked at 390 px
- The final mix is song only with no room tone
- AI and music disclosures ship with the master
- Every line has a syllable count inside the melody's range
- The transformation is carried by one garment colour, not a new palette
REFUSE OR STOP WHEN
- No clip generated before the song and the timing map are locked
- No ASR transcript used as the lyric
- No fixed clip length
- No product appearance before the halfway mark
- No new character sheet after the transformation is shot
- No delivery without the disclosure lines
Render budget, today
45s on the board · Seedance 2.5 720p
12 shots, 45 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
| Role | Tool |
|---|---|
| Script and journey | Claude |
| Song | Suno |
| Cast sheets | GPT Image |
| Video | Runway |
| Timing and transcript | Whisper |
| Edit | CapCut |
Other packs
8 more
AI Drama Series
A recurring cast, an episode arc, and the same face in every shot.
10 shots / 60s episode · cast to scriptCartoon · 9:16 · 3-7s fragmentsAI Cartoon Short
A cartoon with model sheets that survive every cut.
8 shots / 40s short · cast to scriptUGC · 9:16 · 100 ads a dayUGC Ad Factory
One shoot day, a hundred ads, each written for a real buyer.
6 shots / 30s ad · cast to scriptProduct · 9:16 + 16:9 · styleframe-firstProduct Film
A product hero film built from real product pages, not a storyboard guess.
8 shots / 30s film · cast to script