External-tool reference examples from the source experiments, not a claim of current FlashBoards model availability or installed automation. Use your own supported tools and authorized media; settings, timings and comparisons describe the tested runs.
Also called: AI clip, music video with an AI character, AI-generated music video, lyric video with a character
A music video that holds a whole song is a short film on the beat: one world, a story in three acts, the same face in every shot, the mouth on every word. Below, the method for your own track and character — the track, the story, every face replaced, the shots — and, as the worked example, «The Welder» made end to end on frames from Titanic: all 54 shots, every prompt, and the six tries it took to get there.
Watch first: the whole guide in 1:47, hosted by YaroVideo script
Hi, I'm Yaro. Today: a whole music video, made with AI. One world, one face in every shot, and lips on every word. Mine are warming up.
A music video is a short film on the beat. Ours is called The Welder: Mara boards the Titanic, cuts the pipes in the boiler room, and rows away blowing a kiss. Sorry, Jack.
Step one: the track. Scribe writes down every word with its time, and the beat is measured from the audio. The model said one fifty BPM. The audio said one thirty-two and a half. Never trust a model's tempo.
Step two: one film, three acts. We tried Titanic, The Matrix and The Wolf of Wall Street at once. Gorgeous frames. Zero story.
Step three: GPT Image 2.5 puts your character into every frame and invents every other face. But it keeps famous ones. Jack survived two passes. First time ever. So crop the head, swap it, paste it back.
Costumes are pictures too: a look sheet. And give it five sides of the face, or the head never turns.
Step four: MiniMax H3 Max. Eight shots a part, each written in full: camera, light, acting, lip-sync, and a transition. And never write the word phone. The model will draw one.
Seven parts run in parallel, in about thirty-five seconds. Want it even better? The source preferred Seedance 2.5 for quality; the measured comparison was a product commercial.
Upscale once, at the very end, with Topaz Starlight. Never the takes: every redo pays twice.
Every prompt and every failed try is right below. Scroll down. I'll be at the bow, arms wide.
What it does
A full-length music video in which your AI character lives a three-act story on the song’s beat, lip-synced.
Use
frames from one film with every face replaced (GPT Image 2.5), MiniMax H3 Max reference-to-video, 8 shots per 15-second part.
Keep
one film, one story, a five-angle face board as Image 1.
✓ Source experiment pick
One film, three acts, every face replaced
Your character as the lead of one film’s world, every other face invented, eight shots a part on H3 Max with constant lip-sync. The worked example, «The Welder»: Mara boards the Titanic and has the time of her life, goes down to the boiler room as a welder and cuts the pipes, and rows away in a lifeboat as the ship goes down — blowing a kiss to the camera. 88.8 seconds, 54 shots, constant lip-sync.
World
one film from the corpus — in the example Titanic (1997) — and a linear story in three acts
Faces
the lead → your character, every other face invented (GPT Image 2.5 Flare edit)
Video
MiniMax H3 Max reference-to-video · 7 parts × 8 shots · 16:9 · 768P
For your track and character, this external reference guide covers how to cut the track on the beat, pick the film and the frames with you, ask who plays any second role, replace every face, write eight shots a part fresh for your story — light, acting, transitions — and join the clip: the method with prompt structures, settings and times; «The Welder» on this page is the worked example.
A music video that holds for a minute and a half is a short film on the beat: one world, one story in three acts, the same face in every shot. Take that world from one film, put your character in its frames and take every actor’s face out, then let MiniMax H3 Max shoot eight shots per 15-second part — with the light, the acting and the lip-sync written shot by shot.
World✕ scenes from three films✓ one film, a linear story in three acts
Faces✕ “change every other face” in one pass✓ crop and edit every face the model keeps
Head✕ a frontal face as Image 1 — the head freezes✓ a five-angle face board
Edit✕ four shots a part, tied to the frames✓ eight shots, a new angle and transition on every cut
Engine✕ Turbo extend, text only — faces drift✓ H3 Max reference-to-video with the frames
Draft or quality: Seedance 2.5
The external method on this page uses MiniMax H3 for drafts and many takes. Seedance 2.5 reference-to-video accepts the same pictures and sound as references (@Image1, @Audio1), up to 30 seconds a call, with its own synced sound at 480p, 720p or 1080p. The source preferred Seedance for quality. Side by side on our product commercial, its reported cuts were approximately one second early and half a second late; it drew finer detail and added its own sound effects in sync. This side-by-side result is from a product commercial, not a measurement on every format.
1 · The track
Scribe v2 · audio-understanding · beat grid
Any track works — sung or rapped. Three passes on it before any picture: the lyrics with word times (ElevenLabs Scribe v2 on fal), the style and mood (fal audio-understanding) and the beat grid measured from the audio itself — never trust a model’s tempo. H3 takes up to 15 seconds, so the track is cut into parts of up to 15 seconds, each on a beat inside the longest pause between sung words; each part’s audio is its Audio 1.
Example run — «The Welder»: the track «3upd (Cover)», 88.8 seconds of female trap. The style model said 150 BPM; the audio measured 132.5 BPM. Seven parts:
Part
Act
Track
Generation
What happens
1
I · Joy
0.00–11.81 s
12 s
boarding at the dock, the cheering crowd, up the gangway
2
I · Joy
11.81–26.30 s
15 s
sunset on deck, “I’m flying” at the bow, the party below deck
3
II · Sabotage
26.30–41.24 s
15 s
the boiler room as a welder, the stokers, lighting the torch
4
II · Sabotage
41.24–52.56 s
12 s
the pipe gives way, the telegraph jerks to STOP, red steam
5
III · Flood
52.56–64.78 s
13 s
the hull bursts, the sea pours in, waist-deep
6
III · Flood
64.78–79.27 s
15 s
flooded corridors, the grand saloon under water, the doors
7
III · Ending
79.27–88.80 s
10 s
the ship breaks in two, the lifeboat, the kiss to camera
Reference guides for this step · Markdown instructions for Claude Code or any agent
The story comes from a film, not from thin air. Pick one film whose story fits your track, so the world never breaks, and write a linear story in three acts — joy → a turn → a catastrophe or an ending — with your character doing something in every beat. Search the corpus — 1,100 films, 120,000 scenes with their scripts and dialogue — for scenes that tell each beat, and take four frames per part.
Cast first. Your character is the lead. If the film has a second main role in your story, decide who plays it — another of your own characters, an invented face, or nobody (the person is removed). Never borrow a character from an example run.
Every frame is pulled from the film’s HD master over the network — the shot number maps to the HD time — and nothing is cached on disk.
Example run — «The Welder» on Titanic (1997) — one voyage, no other heroes:
Act I · joy — boarding at the dock, sunset at the bow, the party below deck.
Act II · sabotage — Mara in a welder’s apron goes down to the boiler room and cuts the pipes with a torch; the engines break.
Act III · the flood — the sea pours in, the ship sinks from the inside, and she rows away in a lifeboat as it breaks in two.
Reference guides for this step · Markdown instructions for Claude Code or any agent
Each frame goes through GPT Image 2.5 Flare edit. The lead becomes your character — the head, and the film’s costume may stay — a second main role goes to whoever you cast, and every other face becomes an invented one. The prompts are blind: they name the roles of the pictures and never describe the frame or your character’s looks. Write each one fresh from the structure below; {character} is yours, [brackets] are example values.
The method · replace prompt structure
Picture 1 is only the look of {character}: their face, hair and their whole [outfit name, e.g. "welder's outfit"]. Picture 2 is a frame from a film. Shoot picture 2 again exactly the same: the same camera, framing, set, light and atmosphere; but the main person of the frame is {character} from picture 1, in exactly their outfit from picture 1, in the same place, pose and movement. Every other person in the frame keeps their place, pose, clothes and hair colour but gets a completely different, invented face that belongs to nobody from any film or real life. Change nothing else.
Variants (same structure):
· Face board only, the film’s costume stays: "Picture 1 is only the look of {character}: their face and hair", and keep the clothes of picture 2.
· A crowd frame: the same prompt without picture 1.
· A prop in hand: "{character} holds the [prop] from picture 1 [how and where]".
· Remove a person: "keep {character} exactly as they are... remove the [person] completely, show the [place] where they were".
· A face the model kept, on a ~350 px crop: "a completely different [man/woman] with a new invented face" — pasted back with a feathered mask.
GPT keeps famous faces. A face can survive both the first prompt and a second “change every other face” pass. What works: crop the head, edit only the crop — “a completely different man with a new invented face” — and paste it back with a soft edge.
A person you do not want next to your character is removed by a separate pass on the replaced frame.
A new costume is a picture first. Make a look sheet — the face and hair from the character sheet, never their clothes; a face close-up, the full body and the prop — and give it to every frame that needs it as picture 1.
Example run — «The Welder»: the lead became Mara — her head; in act I she keeps the film’s costumes. Cal and Jack survived both the first prompt and the second pass and were fixed on crops; the young man next to her was removed (“keep Mara exactly as she is… remove the young man next to her completely, show the room where he was”). The welder’s apron, gauntlets, goggles and the lit brass torch were made as a look sheet, and every frame of acts II and III got it as picture 1.
film✓ Mara, faces new
Boarding: the lead and every face around her
film✓ alone
The bow: the second person removed
Example run · look sheet · GPT Image 2.5 FlareExample run · face board · five sides · Image 1 of every part
film✓ the welder
The torch: costume and prop from the look sheet
film✓ the welder, alone
The stokehold: new faces, the lead removed
film✓ soaked
The flood
film✓ invented face
A stoker: no Mara, only a new face
Act I · every frame, film and replacedActs II–III · every frame, film and replaced
Example run — «The Welder» · prompt · the lead with a costume and a prop
Picture 1 is only the look of Mara: her face, hair and her whole welder's outfit (leather apron, work shirt, gauntlet gloves, goggles on her forehead). Picture 2 is a frame from a film. Shoot picture 2 again exactly the same: the same camera, framing, set, light and atmosphere; but the main person of the frame is Mara from picture 1, in exactly her outfit from picture 1, in the same place, pose and movement. She holds the lit brass welding torch from picture 1 against the pipes in front of her; a shower of bright sparks flies off them and lights her face. Every other person in the frame keeps their place, pose, clothes and hair colour but gets a completely different, invented face that belongs to nobody from any film or real life. Change nothing else.
Example run — «The Welder» · prompt · a crowd frame
Picture 1 is a frame from a film. Shoot picture 1 again exactly the same: the same camera, framing, set, light, people, poses, clothes and motion. Every other person in the frame keeps their place, pose, clothes and hair colour but gets a completely different, invented face that belongs to nobody from any film or real life. Change nothing else.
Example run — «The Welder» · prompt · remove a person
Picture 1 is only the look of Mara. Picture 2 is a film still. Keep picture 2 exactly the same, and keep Mara in it exactly as she is, with the same pose, outfit and movement. Remove the young man running next to her completely, with his whole body, and show the room, the steam and the people behind where he was. Every other person keeps their place but has a different, invented face. Change nothing else.
Example run — «The Welder» · prompt · a face the model kept (on the crop)
Picture 1 is a crop of a film still. Shoot it again exactly the same: the same framing, light, pose, expression, hat, hair colour and clothes; but the person in it is a completely different man with a new invented face (a different face shape, eyes, nose, mouth, brows and jaw) who looks like nobody from any film or real life. Change nothing else.
Reference guides for this step · Markdown instructions for Claude Code or any agent
Each part is one MiniMax H3 Max reference-to-video request: Image 1 your character’s five-angle face board, Image 2 the look sheet (when the story has a new costume), then four replaced frames, and Audio 1 the part of the track. Every frame gives two shots — the frame’s idea and another angle — so a part has eight shots, cut on beats. Prompt enhance is disabled; all parts run in parallel.
Every shot is written out in full: camera, light, action, acting (eyes, lips, breath, a gesture), head (where the character turns), performance (the character sings or raps — performs every word in sync with Audio 1 — and nods on every beat) and the transition from the H01 glossary — never the same one twice in a part. Write every prompt fresh for your track and story, in this structure; picture roles only, never your character’s looks. If you cast a second character, name who performs the words of Audio 1.
The method · part prompt structure (JSON)
{
"output": "A cinematic 16:9 music video told as a short film in three acts, part N of M ({act} and what happens). 8 shots with cuts and transitions on the beat at the given times, D seconds, then the video ends.",
"style": "Cinematic and natural: a moving camera, the real light of each place, natural skin and sweat, real acting. No text, no screens, no phones, no slow motion.",
"references": {
"Image 1": "only {character}'s face and hair, seen from five sides so they can turn their head freely; never on screen",
"Image 2": "only their outfit [and prop] — the look sheet, when the story has one",
"Image k": "the place of shots 2j+1 and 2j+2: place, light, atmosphere and the people around; never its framing",
"Audio 1": "the soundtrack… only {character} [raps/sings] the words"
},
"character": "{character} is never frozen: their head, eyes and face move all the time and their expression changes with every moment.",
"rules": [
"framing new every time",
"everyone around exactly like the images, no new crowds",
"each transition done in camera on the beat",
"no split screen"
],
"shots": [
{
"cut_at": "<on a beat, from beats.json>",
"transition": "<from the H01 glossary, never twice in a part>",
"camera": "<…>",
"light": "<the real sources of this place>",
"action": "<what {character} does>",
"acting": "<eyes, lips, breath, a gesture>",
"head": "<where they turn>",
"performance": "[raps/sings] every word in sync with Audio 1… nodding their head and shoulders on every beat",
"lyrics": "<the words of this shot, from the Scribe word times>"
}
]
}
Example run — «The Welder»: one shot from part 3 — read it, don’t reuse it.
Example run — «The Welder» · one shot · part 3
{
"shot": 7,
"images": "Image 6",
"cut_at": "00:11.320",
"transition": "a superimposition: her face briefly lies over the next place",
"camera": "a medium shot from the side",
"light": "cold grey machinery light cut by the blinding blue-white flame of the torch, sparks lighting her face from below",
"action": "Mara pulls the goggles down over her eyes and lights the brass torch at the pipes",
"acting": "a slow wicked smile, a deep breath in, then rapping through her teeth",
"head": "she turns her head toward frame-right to watch something, then looks back",
"performance": "Mara raps every word of this shot in sync with Audio 1, lips matching each syllable, nodding her head and shoulders on every beat",
"lyrics": "my name. Body is canvas,"
}
Transitions used in the example: a crash cut, a flash cut, a frame portal, a jump cut, a lap dissolve, a match motion, a match on action, a roll transition, a seamless transition, a shape transition, a smash cut on the strongest beat, a superimposition, a whip transition, a wipe, an arc transition, an object portal.
Example run — «The Welder» · every generation of the clip · video, inputs, prompt
Part 1 · I · Joy
boarding at the dock, the cheering crowd, up the gangway
{
"output": "A cinematic 16:9 music video told as a short film in three acts, part 1 of 7 (Act I, joy: Mara boards the great ship at the dock). 8 shots with cuts and transitions on the beat at the given times, 11.8 seconds, then the video ends.",
"style": "Cinematic and natural: a moving camera, varied angles and lenses, the real light of every place described shot by shot, shallow depth of field, natural skin and sweat, real physical acting. No text, no screens, no phones, no slow motion.",
"references": {
"Image 1": "only Mara's face and hair, seen from five sides so she can turn her head freely; it never appears on screen",
"Image 2": "the place of shots 1 and 2: its place, light, atmosphere and the people around; never its framing",
"Image 3": "the place of shots 3 and 4: its place, light, atmosphere and the people around; never its framing",
"Image 4": "the place of shots 5 and 6: its place, light, atmosphere and the people around; never its framing",
"Image 5": "the place of shots 7 and 8: its place, light, atmosphere and the people around; never its framing",
"Audio 1": "the soundtrack from the first frame to the last; every cut lands on its beat. Only Mara raps the words; nobody else sings."
},
"character": "Mara: exactly the face and hair of Image 1 in every shot; a rapper living this story, never frozen: her head, eyes and face move all the time and her expression changes with every moment.",
"rules": [
"Each shot uses its image only for the place, light, atmosphere and the people around; the camera angle and framing are new every time.",
"Everyone around Mara looks exactly like the people in the images; no new crowds.",
"Each transition is done in camera exactly as written, on the beat.",
"The screen is never split."
],
"shots": [
{
"shot": 1,
"images": "Image 2",
"camera": "a handheld selfie-style close-up at arm's length",
"light": "soft overcast morning light on the dock, drifting steam, cool silver tones, the black hull towering behind",
"action": "Mara turns to the camera in the waving crowd, the huge hull behind her",
"acting": "eyes wide with excitement, a big open grin, eyebrows up; she bites her lip and laughs",
"head": "she turns her head toward frame-right to watch something, then looks back",
"performance": "Mara raps every word of this shot in sync with Audio 1, lips matching each syllable, nodding her head and shoulders on every beat",
"lyrics": "Aya, Aya, Aya, A-A-A-A-A."
},
{
"shot": 2,
"images": "Image 2",
"cut_at": "00:01.850",
"transition": "an arc transition: the camera orbits and the cut hides in the orbit",
"camera": "a low-angle tracking shot moving beside her",
"light": "soft overcast morning light on the dock, drifting steam, cool silver tones, the black hull towering behind",
"action": "she strides through the crowd holding the brim of her hat",
"acting": "chin up, smug and delighted; she shoots a wink toward frame-left",
"head": "she rolls her head and neck to the beat",
"performance": "Mara raps every word of this shot in sync with Audio 1, lips matching each syllable, nodding her head and shoulders on every beat",
"lyrics": "Aya, Aya,"
},
{
"shot": 3,
"images": "Image 3",
"cut_at": "00:03.660",
"transition": "a wipe: something passes right in front of the lens and the next place is revealed behind it",
"camera": "a wide shot over the crowd, the camera rising on a crane",
"light": "the same overcast morning, steam from the funnels, flags and hats in the air",
"action": "the crowd waves and cheers toward the ship; Mara pushes through them toward the camera",
"acting": "she shouts the words with the crowd, one arm up",
"head": "she looks down, then lifts her eyes past the camera",
"performance": "Mara raps every word of this shot in sync with Audio 1, lips matching each syllable, nodding her head and shoulders on every beat",
"lyrics": "Aya on shine. Way,"
},
{
"shot": 4,
"images": "Image 3",
"cut_at": "00:05.020",
"transition": "a crash cut: a fast crash zoom carries the cut",
"camera": "a whip pan across the cheering faces that lands on Mara",
"light": "the same overcast morning, steam from the funnels, flags and hats in the air",
"action": "Mara raps in the middle of the crowd",
"acting": "cocky, playful, shoulders bouncing on the beat",
"head": "she turns into profile and raps to the side",
"performance": "Mara raps every word of this shot in sync with Audio 1, lips matching each syllable, nodding her head and shoulders on every beat",
"lyrics": "way up, I'm seeing the"
},
{
"shot": 5,
"images": "Image 4",
"cut_at": "00:06.370",
"transition": "a shape transition: a shape in one shot becomes a matching shape in the next",
"camera": "a close-up from slightly below, the hat filling the top of the frame",
"light": "morning light filtered through the brim of her huge hat, a soft violet glow on her face",
"action": "Mara tilts her face up toward the ship",
"acting": "lips parting in awe, then a slow smile, her eyes shining",
"head": "she looks around as she moves, her head never still",
"performance": "Mara raps every word of this shot in sync with Audio 1, lips matching each syllable, nodding her head and shoulders on every beat",
"lyrics": "skyline. Climb, climb"
},
{
"shot": 6,
"images": "Image 4",
"cut_at": "00:07.730",
"transition": "a jump cut: the same framing, a moment later",
"camera": "an extreme close-up of her eyes and lips",
"light": "morning light filtered through the brim of her huge hat, a soft violet glow on her face",
"action": "she raps straight into the camera",
"acting": "playful and cocky, a little head bob, one eyebrow raised",
"head": "she throws her head back, then drops her chin on the beat",
"performance": "Mara raps every word of this shot in sync with Audio 1, lips matching each syllable, nodding her head and shoulders on every beat",
"lyrics": "up higher, no"
},
{
"shot": 7,
"images": "Image 5",
"cut_at": "00:09.090",
"transition": "a frame portal: a dark door frame fills the screen and opens onto the next place",
"camera": "a wide shot of the long gangway",
"light": "bright morning backlight on the white gangway, the water glinting below",
"action": "Mara climbs the gangway among the passengers, looking back over her shoulder",
"acting": "laughing, waving goodbye to the dock",
"head": "she tilts her head from side to side with the rhythm",
"performance": "Mara raps every word of this shot in sync with Audio 1, lips matching each syllable, nodding her head and shoulders on every beat",
"lyrics": "cap. Cut off the weight,"
},
{
"shot": 8,
"images": "Image 5",
"cut_at": "00:10.450",
"transition": "a roll transition: the image rolls around the lens axis into the next shot",
"camera": "a tracking shot behind her going up into the ship",
"light": "bright morning backlight on the white gangway, the water glinting below",
"action": "she reaches the door of the ship and turns once more",
"acting": "a last cheeky grin over her shoulder, rapping the last word",
"head": "she glances over her shoulder toward frame-left, then snaps back",
"performance": "Mara raps every word of this shot in sync with Audio 1, lips matching each syllable, nodding her head and shoulders on every beat",
"lyrics": "no snakes on my back."
}
]
}
Part 2 · I · Joy
sunset on deck, “I’m flying” at the bow, the party below deck
{
"output": "A cinematic 16:9 music video told as a short film in three acts, part 2 of 7 (Act I, joy: sunset on deck, then the wild party below deck). 8 shots with cuts and transitions on the beat at the given times, 14.5 seconds, then the video ends.",
"style": "Cinematic and natural: a moving camera, varied angles and lenses, the real light of every place described shot by shot, shallow depth of field, natural skin and sweat, real physical acting. No text, no screens, no phones, no slow motion.",
"references": {
"Image 1": "only Mara's face and hair, seen from five sides so she can turn her head freely; it never appears on screen",
"Image 2": "the place of shots 1 and 2: its place, light, atmosphere and the people around; never its framing",
"Image 3": "the place of shots 3 and 4: its place, light, atmosphere and the people around; never its framing",
"Image 4": "the place of shots 5 and 6: its place, light, atmosphere and the people around; never its framing",
"Image 5": "the place of shots 7 and 8: its place, light, atmosphere and the people around; never its framing",
"Audio 1": "the soundtrack from the first frame to the last; every cut lands on its beat. Only Mara raps the words; nobody else sings."
},
"character": "Mara: exactly the face and hair of Image 1 in every shot; a rapper living this story, never frozen: her head, eyes and face move all the time and her expression changes with every moment.",
"rules": [
"Each shot uses its image only for the place, light, atmosphere and the people around; the camera angle and framing are new every time.",
"Everyone around Mara looks exactly like the people in the images; no new crowds.",
"Each transition is done in camera exactly as written, on the beat.",
"The screen is never split."
],
"shots": [
{
"shot": 1,
"images": "Image 2",
"camera": "a slow dolly-in to a medium close-up",
"light": "warm low golden sunset light from frame-right, long shadows, the white superstructure glowing",
"action": "Mara leans on the rail, the wind in her hair",
"acting": "dreamy half-closed eyes, a soft smile; she sways her head to the beat while rapping softly",
"head": "she rolls her head and neck to the beat",
"performance": "Mara raps every word of this shot in sync with Audio 1, lips matching each syllable, nodding her head and shoulders on every beat",
"lyrics": "Aya, Aya, Aya,"
},
{
"shot": 2,
"images": "Image 2",
"cut_at": "00:01.810",
"transition": "a crash cut: a fast crash zoom carries the cut",
"camera": "a profile tracking shot at her side",
"light": "warm low golden sunset light from frame-right, long shadows, the white superstructure glowing",
"action": "she turns her face into the wind and raps",
"acting": "free and ecstatic, eyes closed for a second, then open and bright",
"head": "she looks down, then lifts her eyes past the camera",
"performance": "Mara raps every word of this shot in sync with Audio 1, lips matching each syllable, nodding her head and shoulders on every beat",
"lyrics": "big wave. Whole room shaking,"
},
{
"shot": 3,
"images": "Image 3",
"cut_at": "00:03.620",
"transition": "a frame portal: a dark door frame fills the screen and opens onto the next place",
"camera": "a wide low-angle shot from the front of the bow",
"light": "a deep orange sunset behind her, rim light on her hair and outstretched arms, the sea burning gold",
"action": "Mara stands at the very bow with her arms spread wide like wings",
"acting": "eyes closed, face lifted to the wind, mouthing the words",
"head": "she turns into profile and raps to the side",
"performance": "Mara raps every word of this shot in sync with Audio 1, lips matching each syllable, nodding her head and shoulders on every beat",
"lyrics": "it cannot behave. Ice in"
},
{
"shot": 4,
"images": "Image 3",
"cut_at": "00:05.430",
"transition": "a jump cut: the same framing, a moment later",
"camera": "a high crane shot circling around her",
"light": "a deep orange sunset behind her, rim light on her hair and outstretched arms, the sea burning gold",
"action": "she opens her eyes and laughs into the sky",
"acting": "overwhelmed, euphoric, head thrown back",
"head": "she looks around as she moves, her head never still",
"performance": "Mara raps every word of this shot in sync with Audio 1, lips matching each syllable, nodding her head and shoulders on every beat",
"lyrics": "my veins, I'm sliding on tracks."
},
{
"shot": 5,
"images": "Image 4",
"cut_at": "00:07.240",
"transition": "a lap dissolve: one second where both places overlap",
"camera": "a handheld close-up in the middle of the crowd",
"light": "warm smoky tungsten light below deck, sweat glowing on faces, beer foam and dust in the air",
"action": "Mara laughs her head off as dancers spin around her",
"acting": "head thrown back, eyes squeezed shut with laughter, then rapping between laughs",
"head": "she throws her head back, then drops her chin on the beat",
"performance": "Mara raps every word of this shot in sync with Audio 1, lips matching each syllable, nodding her head and shoulders on every beat",
"lyrics": "They try to grab, but I'm breaking"
},
{
"shot": 6,
"images": "Image 4",
"cut_at": "00:09.060",
"transition": "a roll transition: the image rolls around the lens axis into the next shot",
"camera": "a push-in through the crowd toward her",
"light": "warm smoky tungsten light below deck, sweat glowing on faces, beer foam and dust in the air",
"action": "she raps into the camera, shoulders bouncing",
"acting": "cheeky, sweaty, alive, nodding hard on every beat",
"head": "she tilts her head from side to side with the rhythm",
"performance": "Mara raps every word of this shot in sync with Audio 1, lips matching each syllable, nodding her head and shoulders on every beat",
"lyrics": "the lash. Aya, Aya,"
},
{
"shot": 7,
"images": "Image 5",
"cut_at": "00:10.870",
"transition": "a flash cut: a white flash on the beat",
"camera": "a low angle from the floor",
"light": "the same smoky party light, fiddles flashing, lamps swinging",
"action": "Mara throws both arms up and shouts with the crowd",
"acting": "wild joy, mouth wide open, hair flying",
"head": "she glances over her shoulder toward frame-left, then snaps back",
"performance": "Mara raps every word of this shot in sync with Audio 1, lips matching each syllable, nodding her head and shoulders on every beat",
"lyrics": "Aya, I'm queen. Circle"
},
{
"shot": 8,
"images": "Image 5",
"cut_at": "00:12.680",
"transition": "a match motion: the movement continues through the cut in the same direction",
"camera": "the camera circling her fast",
"light": "the same smoky party light, fiddles flashing, lamps swinging",
"action": "she spins with her arms up, rapping",
"acting": "dizzy with happiness, head rolling with the rhythm",
"head": "she turns her head toward frame-right to watch something, then looks back",
"performance": "Mara raps every word of this shot in sync with Audio 1, lips matching each syllable, nodding her head and shoulders on every beat",
"lyrics": "is tight, I don't need that ring."
}
]
}
Part 3 · II · Sabotage
the boiler room as a welder, the stokers, lighting the torch
{
"output": "A cinematic 16:9 music video told as a short film in three acts, part 3 of 7 (Act II, sabotage: Mara, now a welder, goes down into the boiler room and starts cutting the pipes). 8 shots with cuts and transitions on the beat at the given times, 14.9 seconds, then the video ends.",
"style": "Cinematic and natural: a moving camera, varied angles and lenses, the real light of every place described shot by shot, shallow depth of field, natural skin and sweat, real physical acting. No text, no screens, no phones, no slow motion.",
"references": {
"Image 1": "only Mara's face and hair, seen from five sides so she can turn her head freely; it never appears on screen",
"Image 2": "only Mara's welder's outfit and brass torch (the face is the same as Image 1); a reference, never on screen; she wears this outfit in every shot of this part",
"Image 3": "the place of shots 1 and 2: its place, light, atmosphere and the people around; never its framing",
"Image 4": "the place of shots 3 and 4: its place, light, atmosphere and the people around; never its framing",
"Image 5": "the place of shots 5 and 6: its place, light, atmosphere and the people around; never its framing",
"Image 6": "the place of shots 7 and 8: its place, light, atmosphere and the people around; never its framing",
"Audio 1": "the soundtrack from the first frame to the last; every cut lands on its beat. Only Mara raps the words; nobody else sings."
},
"character": "Mara: exactly the face and hair of Image 1 in every shot; a rapper living this story, never frozen: her head, eyes and face move all the time and her expression changes with every moment.",
"rules": [
"Each shot uses its image only for the place, light, atmosphere and the people around; the camera angle and framing are new every time.",
"Everyone around Mara looks exactly like the people in the images; no new crowds.",
"Each transition is done in camera exactly as written, on the beat.",
"The screen is never split."
],
"shots": [
{
"shot": 1,
"images": "Image 3",
"camera": "a tracking shot backing away in front of her",
"light": "hellish orange furnace light, smoke and coal dust hanging in the air, hard black shadows",
"action": "Mara, in her welder's outfit, strides through the boiler room between the stokers",
"acting": "focused, jaw set, eyes narrowed, a hint of a dangerous smile, rapping under her breath",
"head": "she looks down, then lifts her eyes past the camera",
"performance": "Mara raps every word of this shot in sync with Audio 1, lips matching each syllable, nodding her head and shoulders on every beat",
"lyrics": "Jungle around, but I'm hunting that king."
},
{
"shot": 2,
"images": "Image 3",
"cut_at": "00:02.260",
"transition": "a jump cut: the same framing, a moment later",
"camera": "a low-angle shot past a stoker's shovel",
"light": "hellish orange furnace light, smoke and coal dust hanging in the air, hard black shadows",
"action": "the stokers turn to look at her; she does not slow down",
"acting": "cold and determined, nodding to the beat, eyes flicking from side to side",
"head": "she turns into profile and raps to the side",
"performance": "Mara raps every word of this shot in sync with Audio 1, lips matching each syllable, nodding her head and shoulders on every beat",
"lyrics": "Roar on the beat, take your head for"
},
{
"shot": 3,
"images": "Image 4",
"cut_at": "00:04.070",
"transition": "a flash cut: a white flash on the beat",
"camera": "a wide shot of the furnaces, the camera drifting sideways",
"light": "white-hot furnace mouths, orange flicker on sweaty skin, sparks of coal",
"action": "stokers shovel coal into the roaring furnaces; Mara passes behind them",
"acting": "her goggles flash in the firelight as she turns her head",
"head": "she looks around as she moves, her head never still",
"performance": "Mara raps every word of this shot in sync with Audio 1, lips matching each syllable, nodding her head and shoulders on every beat",
"lyrics": "a ring. If they want scars, then"
},
{
"shot": 4,
"images": "Image 4",
"cut_at": "00:05.890",
"transition": "a roll transition: the image rolls around the lens axis into the next shot",
"camera": "an extreme close-up of the fire in a furnace mouth, then up to Mara's face",
"light": "white-hot furnace mouths, orange flicker on sweaty skin, sparks of coal",
"action": "Mara smiles at the flames and raps",
"acting": "hungry, mischievous, the fire dancing in her eyes",
"head": "she throws her head back, then drops her chin on the beat",
"performance": "Mara raps every word of this shot in sync with Audio 1, lips matching each syllable, nodding her head and shoulders on every beat",
"lyrics": "I give 'em a tag. Sign on"
},
{
"shot": 5,
"images": "Image 5",
"cut_at": "00:07.700",
"transition": "a seamless transition: the camera passes behind a pillar or a cloud of steam into the next place",
"camera": "a medium shot of a stoker",
"light": "orange rim light, black coal dust floating",
"action": "a sweating stoker stops shovelling and stares at Mara as she passes",
"acting": "Mara gives him a mocking little salute with two fingers",
"head": "she tilts her head from side to side with the rhythm",
"performance": "Mara raps every word of this shot in sync with Audio 1, lips matching each syllable, nodding her head and shoulders on every beat",
"lyrics": "their skin like I'm tagging a bag."
},
{
"shot": 6,
"images": "Image 5",
"cut_at": "00:09.510",
"transition": "a match motion: the movement continues through the cut in the same direction",
"camera": "an over-the-shoulder shot from behind her",
"light": "orange rim light, black coal dust floating",
"action": "she looks back at him and raps a line at him",
"acting": "teasing and fearless, a crooked smile",
"head": "she glances over her shoulder toward frame-left, then snaps back",
"performance": "Mara raps every word of this shot in sync with Audio 1, lips matching each syllable, nodding her head and shoulders on every beat",
"lyrics": "Ink in the air when I'm dropping"
},
{
"shot": 7,
"images": "Image 6",
"cut_at": "00:11.320",
"transition": "a superimposition: her face briefly lies over the next place",
"camera": "a medium shot from the side",
"light": "cold grey machinery light cut by the blinding blue-white flame of the torch, sparks lighting her face from below",
"action": "Mara pulls the goggles down over her eyes and lights the brass torch at the pipes",
"acting": "a slow wicked smile, a deep breath in, then rapping through her teeth",
"head": "she turns her head toward frame-right to watch something, then looks back",
"performance": "Mara raps every word of this shot in sync with Audio 1, lips matching each syllable, nodding her head and shoulders on every beat",
"lyrics": "my name. Body is canvas,"
},
{
"shot": 8,
"images": "Image 6",
"cut_at": "00:13.130",
"transition": "a match on action: the cut lands in the middle of her movement and continues it closer",
"camera": "a close-up of the torch biting into the pipe in a shower of sparks, tilting up to her face",
"light": "cold grey machinery light cut by the blinding blue-white flame of the torch, sparks lighting her face from below",
"action": "she cuts into the pipe",
"acting": "fierce concentration, sweat on her temple, lips moving with every word",
"head": "she rolls her head and neck to the beat",
"performance": "Mara raps every word of this shot in sync with Audio 1, lips matching each syllable, nodding her head and shoulders on every beat",
"lyrics": "I'm painting the flame."
}
]
}
Part 4 · II · Sabotage
the pipe gives way, the telegraph jerks to STOP, red steam
{
"output": "A cinematic 16:9 music video told as a short film in three acts, part 4 of 7 (Act II, sabotage: the pipe gives way, the engines break, the boiler room fills with steam and she walks away). 8 shots with cuts and transitions on the beat at the given times, 11.3 seconds, then the video ends.",
"style": "Cinematic and natural: a moving camera, varied angles and lenses, the real light of every place described shot by shot, shallow depth of field, natural skin and sweat, real physical acting. No text, no screens, no phones, no slow motion.",
"references": {
"Image 1": "only Mara's face and hair, seen from five sides so she can turn her head freely; it never appears on screen",
"Image 2": "only Mara's welder's outfit and brass torch (the face is the same as Image 1); a reference, never on screen; she wears this outfit in every shot of this part",
"Image 3": "the place of shots 1 and 2: its place, light, atmosphere and the people around; never its framing",
"Image 4": "the place of shots 3 and 4: its place, light, atmosphere and the people around; never its framing",
"Image 5": "the place of shots 5 and 6: its place, light, atmosphere and the people around; never its framing",
"Image 6": "the place of shots 7 and 8: its place, light, atmosphere and the people around; never its framing",
"Audio 1": "the soundtrack from the first frame to the last; every cut lands on its beat. Only Mara raps the words; nobody else sings."
},
"character": "Mara: exactly the face and hair of Image 1 in every shot; a rapper living this story, never frozen: her head, eyes and face move all the time and her expression changes with every moment.",
"rules": [
"Each shot uses its image only for the place, light, atmosphere and the people around; the camera angle and framing are new every time.",
"Everyone around Mara looks exactly like the people in the images; no new crowds.",
"Each transition is done in camera exactly as written, on the beat.",
"The screen is never split."
],
"shots": [
{
"shot": 1,
"images": "Image 3",
"camera": "an extreme close-up of the pipe splitting open under the torch, steam spitting out",
"light": "the blue-white torch flame and a burst of white steam, red warning glow",
"action": "the pipe gives way",
"acting": "Mara flinches back, then grins",
"head": "she turns into profile and raps to the side",
"performance": "Mara raps every word of this shot in sync with Audio 1, lips matching each syllable, nodding her head and shoulders on every beat",
"lyrics": "When they all stare at you like what the"
},
{
"shot": 2,
"images": "Image 3",
"cut_at": "00:01.810",
"transition": "a roll transition: the image rolls around the lens axis into the next shot",
"camera": "a crash zoom onto her goggled face lit by sparks",
"light": "the blue-white torch flame and a burst of white steam, red warning glow",
"action": "Mara pushes the goggles up and laughs",
"acting": "triumphant, laughing, rapping the words",
"head": "she looks around as she moves, her head never still",
"performance": "Mara raps every word of this shot in sync with Audio 1, lips matching each syllable, nodding her head and shoulders on every beat",
"lyrics": "fuck? When they all"
},
{
"shot": 3,
"images": "Image 4",
"cut_at": "00:03.170",
"transition": "a superimposition: her face briefly lies over the next place",
"camera": "an extreme close-up of the engine telegraph handle",
"light": "dim brass and glass, one lamp, deep shadows",
"action": "the handle jerks hard to STOP on its own, the dial needle swinging",
"acting": "no people",
"lyrics": "stare at me like not enough."
},
{
"shot": 4,
"images": "Image 4",
"cut_at": "00:04.530",
"transition": "a match motion: the movement continues through the cut in the same direction",
"camera": "a Dutch-angle close-up as the whole telegraph shakes",
"light": "dim brass and glass, one lamp, deep shadows",
"action": "the glass of the dial cracks",
"acting": "no people",
"lyrics": "I don't see demons,"
},
{
"shot": 5,
"images": "Image 5",
"cut_at": "00:05.890",
"transition": "a smash cut on the strongest beat",
"camera": "a close-up of a stoker",
"light": "red emergency glow through orange steam",
"action": "a stoker shouts in alarm as the machinery screams and shudders",
"acting": "panic in his eyes",
"lyrics": "I'm leveling up. I put"
},
{
"shot": 6,
"images": "Image 5",
"cut_at": "00:07.240",
"transition": "a match on action: the cut lands in the middle of her movement and continues it closer",
"camera": "a handheld shot swinging past him",
"light": "red emergency glow through orange steam",
"action": "stokers run past through the steam",
"acting": "chaos all around",
"lyrics": "a wolf on each baby"
},
{
"shot": 7,
"images": "Image 6",
"cut_at": "00:08.600",
"transition": "a whip transition: a fast whip pan blurs into the next place",
"camera": "a slow push-in",
"light": "thick orange steam backlit by the furnaces, silhouettes in the haze",
"action": "Mara walks out of the steam toward the camera, goggles up on her forehead",
"acting": "calm, satisfied smile, rapping the words, head nodding on the beat",
"head": "she throws her head back, then drops her chin on the beat",
"performance": "Mara raps every word of this shot in sync with Audio 1, lips matching each syllable, nodding her head and shoulders on every beat",
"lyrics": "wolf pup. Aya,"
},
{
"shot": 8,
"images": "Image 6",
"cut_at": "00:09.960",
"transition": "an arc transition: the camera orbits and the cut hides in the orbit",
"camera": "a profile tracking shot as she passes",
"light": "thick orange steam backlit by the furnaces, silhouettes in the haze",
"action": "she glances back at the chaos behind her",
"acting": "a smug little wink, then eyes forward",
"head": "she tilts her head from side to side with the rhythm",
"performance": "Mara raps every word of this shot in sync with Audio 1, lips matching each syllable, nodding her head and shoulders on every beat",
"lyrics": "Aya, Aya, moonshine."
}
]
}
{
"output": "A cinematic 16:9 music video told as a short film in three acts, part 5 of 7 (Act III, the flood: the hull bursts and the sea pours into the boiler room). 8 shots with cuts and transitions on the beat at the given times, 12.2 seconds, then the video ends.",
"style": "Cinematic and natural: a moving camera, varied angles and lenses, the real light of every place described shot by shot, shallow depth of field, natural skin and sweat, real physical acting. No text, no screens, no phones, no slow motion.",
"references": {
"Image 1": "only Mara's face and hair, seen from five sides so she can turn her head freely; it never appears on screen",
"Image 2": "only Mara's welder's outfit and brass torch (the face is the same as Image 1); a reference, never on screen; she wears this outfit in every shot of this part",
"Image 3": "the place of shots 1 and 2: its place, light, atmosphere and the people around; never its framing",
"Image 4": "the place of shots 3 and 4: its place, light, atmosphere and the people around; never its framing",
"Image 5": "the place of shots 5 and 6: its place, light, atmosphere and the people around; never its framing",
"Image 6": "the place of shots 7 and 8: its place, light, atmosphere and the people around; never its framing",
"Audio 1": "the soundtrack from the first frame to the last; every cut lands on its beat. Only Mara raps the words; nobody else sings."
},
"character": "Mara: exactly the face and hair of Image 1 in every shot; a rapper living this story, never frozen: her head, eyes and face move all the time and her expression changes with every moment.",
"rules": [
"Each shot uses its image only for the place, light, atmosphere and the people around; the camera angle and framing are new every time.",
"Everyone around Mara looks exactly like the people in the images; no new crowds.",
"Each transition is done in camera exactly as written, on the beat.",
"The screen is never split."
],
"shots": [
{
"shot": 1,
"images": "Image 3",
"camera": "a wide shot",
"light": "dirty golden lamp light, spray everywhere, the lamps swinging",
"action": "the hull plates burst and a wall of sea water smashes into the boiler room",
"acting": "stokers stagger back",
"head": "she looks around as she moves, her head never still",
"performance": "Mara raps every word of this shot in sync with Audio 1, lips matching each syllable, nodding her head and shoulders on every beat",
"lyrics": "Way, way up, I'm seeing the"
},
{
"shot": 2,
"images": "Image 3",
"cut_at": "00:01.810",
"transition": "a match motion: the movement continues through the cut in the same direction",
"camera": "a low angle at the water line",
"light": "dirty golden lamp light, spray everywhere, the lamps swinging",
"action": "the torrent knocks men off their feet",
"acting": "chaos",
"lyrics": "skyline. Climb, climb up"
},
{
"shot": 3,
"images": "Image 4",
"cut_at": "00:03.620",
"transition": "a whip transition: a fast whip pan blurs into the next place",
"camera": "a close shot of the jet roaring through the split seam",
"light": "the same golden light through a white jet of water",
"action": "the water jet hammers the floor plates",
"acting": "no people in focus",
"head": "she throws her head back, then drops her chin on the beat",
"performance": "Mara raps every word of this shot in sync with Audio 1, lips matching each syllable, nodding her head and shoulders on every beat",
"lyrics": "higher, no cap. Cut"
},
{
"shot": 4,
"images": "Image 4",
"cut_at": "00:05.430",
"transition": "a match on action: the cut lands in the middle of her movement and continues it closer",
"camera": "a whip pan to Mara's face as she sees it",
"light": "the same golden light through a white jet of water",
"action": "Mara stares at the water",
"acting": "shock, then a crazy breathless laugh, rapping",
"head": "she tilts her head from side to side with the rhythm",
"performance": "Mara raps every word of this shot in sync with Audio 1, lips matching each syllable, nodding her head and shoulders on every beat",
"lyrics": "off the weight, no snakes on my"
},
{
"shot": 5,
"images": "Image 5",
"cut_at": "00:06.790",
"transition": "an object portal: the camera pushes into a round shape (a porthole, a lamp, a furnace mouth) and comes out in the next place",
"camera": "a handheld shot running behind a fleeing stoker",
"light": "spray hitting the lens, golden light, shaking camera",
"action": "the stoker runs through the falling water",
"acting": "panic",
"lyrics": "back. Aya,"
},
{
"shot": 6,
"images": "Image 5",
"cut_at": "00:08.150",
"transition": "an arc transition: the camera orbits and the cut hides in the orbit",
"camera": "the camera ducking under the spray",
"light": "spray hitting the lens, golden light, shaking camera",
"action": "Mara runs past, splashing",
"acting": "eyes wide, mouth moving with the words",
"head": "she glances over her shoulder toward frame-left, then snaps back",
"performance": "Mara raps every word of this shot in sync with Audio 1, lips matching each syllable, nodding her head and shoulders on every beat",
"lyrics": "Aya, Aya, cold wave."
},
{
"shot": 7,
"images": "Image 6",
"cut_at": "00:09.510",
"transition": "a wipe: something passes right in front of the lens and the next place is revealed behind it",
"camera": "a medium shot as she fights through the waist-deep surge",
"light": "foaming water lit gold by the lamps, spray flying across the frame",
"action": "Mara pushes through the rushing water",
"acting": "gasping, teeth clenched, rapping the words through the spray",
"head": "she turns her head toward frame-right to watch something, then looks back",
"performance": "Mara raps every word of this shot in sync with Audio 1, lips matching each syllable, nodding her head and shoulders on every beat",
"lyrics": "Whole room shaking,"
},
{
"shot": 8,
"images": "Image 6",
"cut_at": "00:10.870",
"transition": "a crash cut: a fast crash zoom carries the cut",
"camera": "a close-up as a wave slams into her",
"light": "foaming water lit gold by the lamps, spray flying across the frame",
"action": "she shakes the water from her face and keeps going",
"acting": "fierce, unbroken, head snapping toward frame-right to find a way out",
"head": "she rolls her head and neck to the beat",
"performance": "Mara raps every word of this shot in sync with Audio 1, lips matching each syllable, nodding her head and shoulders on every beat",
"lyrics": "it cannot behave."
}
]
}
Part 6 · III · Flood
flooded corridors, the grand saloon under water, the doors
{
"output": "A cinematic 16:9 music video told as a short film in three acts, part 6 of 7 (Act III, the flood: the ship sinks from the inside; Mara fights her way up). 8 shots with cuts and transitions on the beat at the given times, 14.5 seconds, then the video ends.",
"style": "Cinematic and natural: a moving camera, varied angles and lenses, the real light of every place described shot by shot, shallow depth of field, natural skin and sweat, real physical acting. No text, no screens, no phones, no slow motion.",
"references": {
"Image 1": "only Mara's face and hair, seen from five sides so she can turn her head freely; it never appears on screen",
"Image 2": "only Mara's welder's outfit and brass torch (the face is the same as Image 1); a reference, never on screen; she wears this outfit in every shot of this part",
"Image 3": "the place of shots 1 and 2: its place, light, atmosphere and the people around; never its framing",
"Image 4": "the place of shots 3 and 4: its place, light, atmosphere and the people around; never its framing",
"Image 5": "the place of shots 5 and 6: its place, light, atmosphere and the people around; never its framing",
"Image 6": "the place of shots 7 and 8: its place, light, atmosphere and the people around; never its framing",
"Audio 1": "the soundtrack from the first frame to the last; every cut lands on its beat. Only Mara raps the words; nobody else sings."
},
"character": "Mara: exactly the face and hair of Image 1 in every shot; a rapper living this story, never frozen: her head, eyes and face move all the time and her expression changes with every moment.",
"rules": [
"Each shot uses its image only for the place, light, atmosphere and the people around; the camera angle and framing are new every time.",
"Everyone around Mara looks exactly like the people in the images; no new crowds.",
"Each transition is done in camera exactly as written, on the beat.",
"The screen is never split."
],
"shots": [
{
"shot": 1,
"images": "Image 3",
"camera": "a backward tracking shot in front of her",
"light": "cold flickering white corridor light, dark water rising",
"action": "Mara wades fast up the corridor, soaked",
"acting": "panting, determined, mouthing the words, turning her head to listen to the groaning ship",
"head": "she throws her head back, then drops her chin on the beat",
"performance": "Mara raps every word of this shot in sync with Audio 1, lips matching each syllable, nodding her head and shoulders on every beat",
"lyrics": "Mud in the water gets washed when it floods."
},
{
"shot": 2,
"images": "Image 3",
"cut_at": "00:02.260",
"transition": "a match on action: the cut lands in the middle of her movement and continues it closer",
"camera": "an over-the-shoulder shot from behind as she heads for the stairs",
"light": "cold flickering white corridor light, dark water rising",
"action": "she glances back at the rising water",
"acting": "a flash of fear, then a defiant smirk",
"head": "she tilts her head from side to side with the rhythm",
"performance": "Mara raps every word of this shot in sync with Audio 1, lips matching each syllable, nodding her head and shoulders on every beat",
"lyrics": "All of the dirt turning back into"
},
{
"shot": 3,
"images": "Image 4",
"cut_at": "00:04.070",
"transition": "a wipe: something passes right in front of the lens and the next place is revealed behind it",
"camera": "a slow high crane shot over the flooded saloon",
"light": "an eerie green underwater glow, floating chairs and wreckage, broken chandeliers",
"action": "the grand saloon slowly sinks, furniture floating",
"acting": "Mara swims across far below",
"head": "she glances over her shoulder toward frame-left, then snaps back",
"performance": "Mara raps every word of this shot in sync with Audio 1, lips matching each syllable, nodding her head and shoulders on every beat",
"lyrics": "dust. Aya, Aya,"
},
{
"shot": 4,
"images": "Image 4",
"cut_at": "00:05.890",
"transition": "an arc transition: the camera orbits and the cut hides in the orbit",
"camera": "a low tracking shot at the water line",
"light": "an eerie green underwater glow, floating chairs and wreckage, broken chandeliers",
"action": "Mara swims past floating chairs",
"acting": "awed and scared, breathing hard, still mouthing the words",
"head": "she turns her head toward frame-right to watch something, then looks back",
"performance": "Mara raps every word of this shot in sync with Audio 1, lips matching each syllable, nodding her head and shoulders on every beat",
"lyrics": "up, up, up."
},
{
"shot": 5,
"images": "Image 5",
"cut_at": "00:07.700",
"transition": "a shape transition: a shape in one shot becomes a matching shape in the next",
"camera": "a wide shot",
"light": "pale green light through glass doors, water gushing, lamps flickering",
"action": "water pours in through the doors; the lights flicker",
"acting": "no one in focus",
"lyrics": "And you, and you down, down, down."
},
{
"shot": 6,
"images": "Image 5",
"cut_at": "00:09.510",
"transition": "a crash cut: a fast crash zoom carries the cut",
"camera": "a push-in to Mara forcing the doors open against the water",
"light": "pale green light through glass doors, water gushing, lamps flickering",
"action": "she shoves the doors with her shoulder",
"acting": "furious effort, teeth bared",
"head": "she rolls her head and neck to the beat",
"performance": "Mara raps every word of this shot in sync with Audio 1, lips matching each syllable, nodding her head and shoulders on every beat",
"lyrics": "Reload, reload my"
},
{
"shot": 7,
"images": "Image 6",
"cut_at": "00:11.320",
"transition": "a frame portal: a dark door frame fills the screen and opens onto the next place",
"camera": "a medium shot",
"light": "teal glow of the flooded corridor, lamps shining under the water",
"action": "Mara, chest-deep, pulls herself along the pipes overhead",
"acting": "gasping, eyes wide, still rapping",
"head": "she looks down, then lifts her eyes past the camera",
"performance": "Mara raps every word of this shot in sync with Audio 1, lips matching each syllable, nodding her head and shoulders on every beat",
"lyrics": "swag. How I'm catching myself,"
},
{
"shot": 8,
"images": "Image 6",
"cut_at": "00:13.130",
"transition": "a jump cut: the same framing, a moment later",
"camera": "a close-up from above",
"light": "teal glow of the flooded corridor, lamps shining under the water",
"action": "she takes a deep breath and ducks under the water",
"acting": "a last fierce look up before she goes under",
"head": "she turns into profile and raps to the side",
"performance": "Mara raps every word of this shot in sync with Audio 1, lips matching each syllable, nodding her head and shoulders on every beat",
"lyrics": "haha, ha."
}
]
}
Part 7 · III · Ending
the ship breaks in two, the lifeboat, the kiss to camera
{
"output": "A cinematic 16:9 music video told as a short film in three acts, part 7 of 7 (Act III, the end: the Titanic breaks and sinks; Mara rows away in a lifeboat and blows a kiss to the camera). 6 shots with cuts and transitions on the beat at the given times, 9.5 seconds, then the video ends.",
"style": "Cinematic and natural: a moving camera, varied angles and lenses, the real light of every place described shot by shot, shallow depth of field, natural skin and sweat, real physical acting. No text, no screens, no phones, no slow motion.",
"references": {
"Image 1": "only Mara's face and hair, seen from five sides so she can turn her head freely; it never appears on screen",
"Image 2": "only Mara's welder's outfit and brass torch (the face is the same as Image 1); a reference, never on screen; she wears this outfit in every shot of this part",
"Image 3": "the place of shots 1 and 2: its place, light, atmosphere and the people around; never its framing",
"Image 4": "the place of shots 3 and 4: its place, light, atmosphere and the people around; never its framing",
"Image 5": "the place of shots 5 and 6: its place, light, atmosphere and the people around; never its framing",
"Audio 1": "the soundtrack from the first frame to the last; every cut lands on its beat. Only Mara raps the words; nobody else sings."
},
"character": "Mara: exactly the face and hair of Image 1 in every shot; a rapper living this story, never frozen: her head, eyes and face move all the time and her expression changes with every moment.",
"rules": [
"Each shot uses its image only for the place, light, atmosphere and the people around; the camera angle and framing are new every time.",
"Everyone around Mara looks exactly like the people in the images; no new crowds.",
"Each transition is done in camera exactly as written, on the beat.",
"The screen is never split."
],
"shots": [
{
"shot": 1,
"images": "Image 3",
"camera": "an extreme wide shot from the sea",
"light": "night, moonlight and starlight on the black sea, the ship's lights failing",
"action": "the giant ship breaks in two and the stern rises into the night sky",
"acting": "no faces",
"lyrics": "Aya, Aya, up, up,"
},
{
"shot": 2,
"images": "Image 3",
"cut_at": "00:02.260",
"transition": "an arc transition: the camera orbits and the cut hides in the orbit",
"camera": "a slow push-in toward the breaking ship",
"light": "night, moonlight and starlight on the black sea, the ship's lights failing",
"action": "the stern tilts higher, lights going out",
"acting": "no faces",
"lyrics": "up. And you, and you down,"
},
{
"shot": 3,
"images": "Image 4",
"cut_at": "00:04.080",
"transition": "a frame portal: a dark door frame fills the screen and opens onto the next place",
"camera": "a wide shot from the water",
"light": "deep blue night, the ship glowing on the horizon, its lights reflected on the black water",
"action": "a lone lifeboat rows away from the sinking ship",
"acting": "tiny figures at the oars",
"lyrics": "down, down. Reload,"
},
{
"shot": 4,
"images": "Image 4",
"cut_at": "00:05.430",
"transition": "a crash cut: a fast crash zoom carries the cut",
"camera": "a tracking shot gliding alongside the lifeboat",
"light": "deep blue night, the ship glowing on the horizon, its lights reflected on the black water",
"action": "Mara rows hard, alone at the oars",
"acting": "exhausted, breathing in rhythm with the strokes, rapping",
"head": "she tilts her head from side to side with the rhythm",
"performance": "Mara raps every word of this shot in sync with Audio 1, lips matching each syllable, nodding her head and shoulders on every beat",
"lyrics": "reload my swag."
},
{
"shot": 5,
"images": "Image 5",
"cut_at": "00:06.790",
"transition": "a lap dissolve: one second where both places overlap",
"camera": "a medium shot in the lifeboat",
"light": "cold moonlight on her face, warm points of the ship's lights behind her on the water",
"action": "Mara, soaked, still in her welder's apron, rows away, the dying ship behind her",
"acting": "relieved, a slow smile spreading, rapping the last words, head nodding",
"head": "she glances over her shoulder toward frame-left, then snaps back",
"performance": "Mara raps every word of this shot in sync with Audio 1, lips matching each syllable, nodding her head and shoulders on every beat",
"lyrics": "How I'm catching"
},
{
"shot": 6,
"images": "Image 5",
"cut_at": "00:08.150",
"transition": "a jump cut: the same framing, a moment later",
"camera": "a slow dolly-in to a close-up of her face",
"light": "cold moonlight on her face, warm points of the ship's lights behind her on the water",
"action": "Mara looks straight into the camera and blows a kiss: mwah",
"acting": "cheeky and triumphant, a wink after the kiss",
"head": "she turns her head toward frame-right to watch something, then looks back",
"performance": "Mara raps every word of this shot in sync with Audio 1, lips matching each syllable, nodding her head and shoulders on every beat",
"lyrics": "myself, haha, ha."
}
]
}
Reference guides for this step · Markdown instructions for Claude Code or any agent
Each part is trimmed to its exact length, the seven are joined at 1344 × 768, 24 fps, and the original track goes underneath — the generated audio is discarded. The cuts land 0.1–0.5 s early; the edit absorbs it.
Then one upscale for the finished cut: Topaz Starlight Precise 2.6 through Magnific’s video upscaler — 1344 × 768 → 1918 × 1096, 24 fps, focus and sharpen at 0. The 88.8-second clip took one final Magnific upscale and came back silent, so the track is laid back on without re-encoding the picture. Starlight Fast 3 is not in the fal or Magnific APIs — only in the Topaz Video app. Measured on the previous cut of the clip, the selfie-reel version:
768P✓ Starlight 2.6
Boarding · a close-up
768P✓ Starlight 2.6
The grand staircase · a medium shot
768P✓ Starlight 2.6
The corridor · a close-up
H3 Max · 768PStarlight 2.6
0:00
The same 15 seconds playing at once: left of the line the joined H3 Max clip at 768P (enlarged to the same size), right of it after Topaz Starlight Precise 2.6. Drag across the video.
Skin, hair and fabric get redrawn: freckles, lace and embroidery come out sharp; the face itself stays the same.
Upscale once, at the very end — never the takes: upscale the assembled final clip once, rather than every part you redo.
Reference guides for this step · Markdown instructions for Claude Code or any agent
Your character as the lead of one film’s world, every other face invented, eight shots a part on H3 Max with constant lip-sync. The worked example, «The Welder»: Mara boards the Titanic and has the time of her life, goes down to the boiler room as a welder and cuts the pipes, and rows away in a lifeboat as the ship goes down — blowing a kiss to the camera. 88.8 seconds, 54 shots, constant lip-sync.
World
one film from the corpus — in the example Titanic (1997) — and a linear story in three acts
Faces
the lead → your character, every other face invented (GPT Image 2.5 Flare edit)
Video
MiniMax H3 Max reference-to-video · 7 parts × 8 shots · 16:9 · 768P
Cut the track into parts of up to 15 seconds on the beat, write a three-act story, take the frames from one film and put your character in them, then generate each part on MiniMax H3 Max reference-to-video with the part of the track as Audio 1 and eight shots written out. Join the parts and lay the original track underneath.
How do I keep the same face in every shot?
Give H3 a face board with five sides of the face as Image 1, put the character into every film frame first, and describe the character by reference (“exactly the face and hair of Image 1”), never in words.
Can I use frames from a real film?
Our pipeline replaces every face in them — the lead with your character and everyone else with invented faces — so no actor appears on screen. Check each face with a crop: the model keeps famous faces unless you edit the crop.
How do I get lip-sync to the song?
Give the part of the track as Audio 1 and write in every shot that the character performs — sings or raps — every word in sync with Audio 1, with the words of that shot. H3 Max plays the reference audio back and animates the mouth to it.
What settings were used for the music-video example?
The example is an 88.8-second clip generated with MiniMax H3 Max at 768P, with edited reference frames and a final Magnific upscale to 1080p.
Is MiniMax H3 Turbo extend good for music videos?
Not on its own: extend-video takes no images, so a text-only chain lets the faces drift. A frame per shot works, but each shot only animates its frame.
How do I add transitions in an AI video?
Ask for them in the prompt, shot by shot, on the beat: a whip pan, a crash zoom, a camera passing through a porthole or a door frame, a wipe by someone crossing the lens. Use a different one on every cut; some still come out as plain cuts.
FlashBoards Learn