P-Video-Replace
P-Video-Replace is a video transformation model that swaps the on-camera character in an existing video with the character from a reference image. It is built to preserve the original motion, timing, camera behavior, lighting, and background while changing who appears in the clip, making it useful for UGC ad variations, content localization, avatar or mascot insertion, and other scalable character-replacement workflows.
Complete technical specification for integration
Ready-to-use code snippets for common workflows
Step-by-step tutorials for advanced use cases
← All GuidesRecasting iconic film scenes
How to recreate recognizable film shot compositions with Seedance and recast them with any character through Pruna P-Video-Replace. Covers shot-composition prompting, scene-source generation, and casting yourself, your avatar, or your mascot into recognizable cinematic moments.
Introduction
Memes and creator content live and die on shared visual references. The dance and the apartment interrogation from Pulp Fiction, the desperate raft call from Cast Away, the bullpen panic run from The Office, the moonlit bike from E.T., the war cry from 300, the phone-call yell from Jerry Maguire, the slow-motion suit walk from Reservoir Dogs, these are shots a viewer reads in half a second, before the joke even arrives. The viral form is putting yourself or your character inside that recognizable moment.
The two-step pattern that makes this possible is straightforward. Generate the iconic shot composition with Bytedance Seedance 2.0 , prompting precisely enough that the composition, lighting, and motion all land. Recast the protagonist with P-Video-Replace , sending a portrait of the character you want in the role.
The source videos in this guide are generated by Seedance so the shipped sample assets don't depend on third-party clips. The replace step is the same regardless of where your source comes from. Use any source video you have rights to: a clip you've generated, recorded, or are remixing under fair use.
The seven recreations below (suit walk, twist dance, raft call, phone-call yell, war cry, moonlit bike, bullpen run) each include the Seedance prompt that produced the source video and the replace call that recast it. Replace is the load-bearing step. The Seedance prompts are there as ready-made shot recipes to lift or adapt.
Request shape
Each recast call takes the source video, one to three reference images, and an optional positivePrompt that names the swap target and what to preserve. The example below uses the suit walk's position-mapping prompt, the load-bearing pattern every multi-character scene in this guide leans on.
[
{
"taskType": "videoInference",
"taskUUID": "a1b2c3d4-e5f6-7890-abcd-ef1234567890",
"model": "prunaai:p-video@replace",
"deliveryMethod": "async",
"inputs": {
"video": "https://example.com/scene-reservoir-suits.mp4",
"referenceImages": ["https://example.com/ref-cast.jpg"]
},
"positivePrompt": "Replace the man in the centre of the group in the source video, the one adjusting his sunglasses, with the man from the reference image. Preserve the source video motion, audio, camera, lighting, and the other four men in the group exactly as they appear in the source.",
"resolution": "720p"
}
][
{
"taskType": "videoInference",
"taskUUID": "a1b2c3d4-e5f6-7890-abcd-ef1234567890",
"videoUUID": "f1e2d3c4-b5a6-7890-1234-567890abcdef",
"videoURL": "https://vm.runware.ai/video/os/a14d18/ws/2/vi/f1e2d3c4-b5a6-7890-1234-567890abcdef.mp4",
"seed": 837412938
}
]Each scene below exposes its own request body in the prompt={...} panel of the output video. The single-character scenes either omit the positivePrompt (Sparta) or add a short disambiguating prompt that preserves a held object (raft call, phone-call yell).
Scene 1: The suit walk
A line of five matching suits walking abreast in slow motion toward a backward-tracking camera, golden hour side light, long shadows. The recast targets the centre man of the group, leaving the other four in their original form. The position-mapping prompt language ("the one adjusting his sunglasses") gives the model a positional anchor when the references on a line of similar-looking figures could go anywhere.
An empty industrial loading area at sunrise outside a brick warehouse in a 1990s American independent crime film, recreating the iconic suit-walk sequence. A group of FIVE MEN walk abreast in slow motion toward the camera across the cracked asphalt, all wearing matching slim-fit black single-breasted suits, crisp white dress shirts, narrow black neckties, polished black dress shoes, and dark wayfarer-style sunglasses. Their black suit jackets flutter slightly in a soft breeze, their footsteps land in unison on the asphalt. The MAN at the centre of the group adjusts the bridge of his sunglasses with his index finger; the other four stare straight ahead with stoic neutral expressions. Steady slow tracking shot moving backward in front of the group at chest height, holding the same distance from them the entire time. Hard golden-hour side light raking from the left, long blue shadows trailing right behind them across the asphalt. Warm muted 1990s indie colour grade with deep golden highlights and slightly faded blacks, slight 35mm film grain. Photorealistic. Cool surf-rock instrumental electric guitar music plays as the soundtrack, no spoken dialogue.
Replace the man in the centre of the group in the source video, the one adjusting his sunglasses, with the man from the reference image. Preserve the source video motion, audio, camera, lighting, and the other four men in the group exactly as they appear in the source.
Scene 2: The twist dance
Two dancers face-to-face, checkered floor, V-shaped hand sliding down across the woman's eyes. The shot is static and centred, head to knees. The recognition lives in the floor, the pose, and the V-fingers.
The replace step uses multi-character handling: the source has two on-camera characters, but only one of them needs recasting. A position-mapping prompt tells the model to swap the left dancer for the reference character while preserving the right dancer untouched.
A 1950s American themed diner interior set in a 1990s American crime film, recreating the iconic Jack Rabbit Slim's twist dance. The setting: red vinyl booth seats line the back wall, a checkered black-and-white tile dance floor stretches forward into the foreground, soft neon signage glows in the background, a small wooden stage area sits to one side. Two people dance the twist face-to-face on the checkered floor in the centre of the frame. On the LEFT, a young MAN in a slim black suit with a white shirt unbuttoned at the collar, dark hair slicked back. On the RIGHT, a young WOMAN with a black bob haircut and blunt fringe, wearing a crisp white button-down shirt tucked into high-waisted black trousers. Both dance barefoot. The woman holds her right hand in front of her face with two fingers split into a V-shape that slides down across her eyes in time with the music while the man sways his shoulders and snaps his fingers low at his hips. Static medium-wide shot framing both dancers head to knees, straight-on, eye level. Warm amber-tinted interior diner lighting, slight smoke haze in the air, classic 1990s film grain, slightly desaturated muted colour grade. Photorealistic. Upbeat instrumental twist rock-and-roll music plays as the soundtrack, no spoken dialogue.
Replace the man on the left in the source video, the one in the black suit dancing the twist, with the man from the reference image. Preserve the source video motion, audio, camera, lighting, and the woman on the right exactly as they appear in the source.
Scene 3: The raft call
A bearded shipwrecked man kneels on a small driftwood raft on the open ocean, arms outstretched toward a volleyball with a painted face that's drifting away on the swells, an anguished single-word cry breaking out of him. The recognition lives in the volleyball drifting just out of reach of the man on the raft.
Even with one character on camera, the replace step needs a short disambiguating prompt here. The volleyball with the painted face is a small object the model can read as the recast slot, dropping the reference into the volleyball position instead of swapping the man on the raft. The prompt names the man as the swap target and lists the volleyball as a preserved element.
A wide open daytime ocean horizon with calm pale-blue water and gentle rolling swells, in a photoreal survival drama film. A small ragged makeshift raft of bound weathered grey driftwood logs lashed together with thin coarse rope drifts in the centre of the frame on the water. A BEARDED MAN in his late thirties kneels on the raft, gaunt and thin from prolonged isolation, with deeply tanned weathered skin, long matted dark brown hair past his shoulders, a full dark unkempt beard, and ragged remnants of olive-khaki cargo shorts and a torn loose tan canvas shirt with the sleeves rolled up. Several meters away from the raft on the open water to camera right, a small white volleyball with a roughly hand-painted reddish-brown smiling face on its side drifts slowly on the swells, slowly drifting further from the raft. The MAN leans forward over the front edge of the raft, reaching both arms outstretched toward the drifting volleyball in desperate anguish, his face contorted with despair, his eyes wide and welling up. He yells with all his strength a single anguished cry to the heavens in a hoarse desperate MALE voice: "WILSON!" Around the raft, the open sea stretches in every direction to the flat horizon, scattered thin distant clouds in the pale sky, the sun glinting off the small waves. Static medium-wide shot framing the man on his raft from waist up at a slight low angle, straight-on, eye level with the water, the drifting volleyball just inside the right edge of the frame. Warm afternoon sunlight from above, soft natural shadows, slightly desaturated film colour grade with cool sea-blue and warm-tan tones, slight 35mm film grain. Photorealistic. His hoarse desperate single-word cry central in the audio with the slow lapping of small ocean waves against the raft and a low distant ocean wind in the background, no other dialogue.
Replace the bearded man kneeling on the raft in the source video, the one reaching his arms outstretched toward the floating volleyball and yelling, with the man from the reference image. Preserve the source video motion, audio, camera, lighting, the wooden raft, the open ocean, and the small white volleyball with the painted face drifting on the water exactly as they appear in the source.
The voice in the output is the voice from the source's audio track. The reference image controls the recast's face and build, not the voice, so the puppet keeps the source's hoarse desperate cry exactly as carried by the source.
Scene 4: The phone-call yell
A sports agent in a white dress shirt with the sleeves rolled and a loosened navy tie stands at his desk, a black phone receiver pressed hard to his right ear, his left fist raised in celebration, yelling "Show me the money!" with the energy of a deal coming together on the line. The corner-office setup and the over-the-top phone-call delivery carry the recognition before the line even lands.
Even with one character on camera, the replace step needs a short disambiguating prompt here too, the same pattern as the raft call. The black phone receiver pressed to his ear is a small held object the model can read as the recast slot, dropping the reference into the phone position instead of swapping the man. The prompt names the man as the swap target and lists the phone receiver as a preserved element.
A bright modern 1990s American sports agent corner office in a 1996 American romantic drama film, recreating the iconic "Show me the money" phone call moment. A MAN in his late thirties stands centred in the frame in front of a polished wood desk, dressed in a clean white dress shirt with the sleeves rolled to the elbows, a loosened dark navy necktie hanging undone at his throat, and tailored grey suit trousers, his short brown hair slightly tousled. He has a clean-shaven face and an open enthusiastic expression. He holds a black push-button office desk phone receiver tightly pressed to his right ear with his right hand, his left hand raised in a celebratory clenched fist at shoulder height. He paces a half-step in place, smiling broadly, his eyes wide with excitement, and yells enthusiastically in a sharp clear MALE voice: "Show me the money!" Behind him, the warm interior of a 1990s sports agent office with floor-to-ceiling windows showing a softly defocused city skyline, a leather couch to one side, framed sports memorabilia on the cream-coloured walls. Static medium shot framing him from waist up at eye level, straight-on. Warm late-afternoon golden sunlight from the windows behind him, soft natural shadows, slightly warm 1990s film colour grade, slight 35mm film grain. Photorealistic. The clean enthusiastic spoken line central in the audio with quiet office ambience underneath, no other dialogue.
Replace the man standing at the desk in the source video, the one in the white dress shirt and loose navy tie holding the phone receiver to his ear and yelling, with the man from the reference image, replacing his face, head, build, and identity completely. Preserve the source video motion, audio, camera, lighting, the office setting, the white dress shirt with rolled sleeves and the loose navy tie, and the black phone receiver pressed to his ear exactly as they appear in the source.
Scene 5: The war cry
A bare-chested warrior king in a deep crimson cape stands planted at the edge of a great stone well, framed by polished pillars and bronze braziers, yelling a three-syllable battle cry to the heavens. The recognition lives in the crimson cape against the bronze braziers plus the booming three-beat yell from the source's audio.
The replace step is the simplest case: one character on camera, one reference image, no prompt needed.
A grand stone Spartan throne chamber in a 2006 stylised historical action film, recreating the iconic battle-cry moment at the edge of the great well. A MUSCULAR MAN in his late thirties stands in the centre of the frame at the edge of a large dark circular stone well sunk into the polished stone floor, dressed in a deep crimson woollen cape that billows behind him in a constant artificial wind, polished bronze greaves on his shins, a leather kilt of bronze studded straps at his waist, a thick dark full beard and shoulder-length dark wavy hair. His muscular bare chest is exposed, his right arm raised mid-gesture, his face contorted into a furious open-mouthed warrior shout, eyes wide. He plants his feet firm and yells full-throated to the heavens in a deep booming MALE voice: "THIS! IS! SPARTA!" Behind him, towering polished stone pillars and a wide stone arch frame the chamber, with bronze braziers casting warm orange firelight up the walls. Static medium-wide shot framing him from knees up at a slight low angle, straight-on, eye level. Heavily saturated bronze-and-teal colour grade pushed to extreme contrast, deep shadows, hot golden highlights, heavy 35mm film grain. Photorealistic. Deep low drum hits and the strong echo of his shout in the audio, no other dialogue.
Scene 6: The moonlit bike
A boy pedals hard down a suburban road on a clear autumn night, a small wrinkly alien wrapped in a blanket in the front basket of his BMX, the wheels just barely lifting off the asphalt as a huge full moon fills the sky behind. The recognition lives in the silhouette against the full moon with the alien's pale forehead poking out of the basket.
The replace step uses multi-character handling again: the boy gets the recast, the alien in the basket stays. A position-mapping prompt names the boy as the target and the alien as a preserved element.
A quiet suburban hillside road on a clear autumn night in a 1982 American family science-fiction film, recreating the iconic moonlit bike moment just before takeoff. A BOY of about ten years old with messy dark-blonde hair, wearing a red zip-up hooded sweatshirt and faded jeans, rides a small chrome BMX bicycle hard down the road in the centre of the frame. In the front basket of the bike sits a small wrinkly alien figure wrapped in a faded brown blanket, only his round head with large dark eyes and an elongated pale-brown forehead visible above the blanket. The boy stands on the pedals, leaning forward, pumping the pedals hard, his expression fixed and determined. The bicycle wheels just barely begin to lift from the asphalt at the end of the shot, the front wheel rising an inch off the ground. Behind them, an enormous bright full moon hangs low in the cobalt-blue night sky, framing the silhouettes of the boy and alien against the moon. Distant suburban houses and the silhouettes of pine trees on either side of the road. Steady tracking shot moving alongside the boy and bike at chest height, holding the same pace. Cool moonlit blue night colour grade with warm amber pools of light from distant porch lamps, slight 35mm film grain. Photorealistic. The whoosh of the bike wheels on asphalt and a rising soaring orchestral horn-and-strings melody in the audio, no spoken dialogue.
Replace the boy riding the bike in the source video, the one pedaling hard with the alien in the front basket, with the man from the reference image. Preserve the source video motion, audio, camera, lighting, the bike, and the alien in the basket exactly as they appear in the source.
Scene 7: The bullpen run
A middle-aged office worker runs through a beige open-plan bullpen toward the camera with both arms raised high and flailing, yelling "It's happening! Everybody stay calm!" as seated coworkers' heads turn to track him from their cubicles. The striped shirt with the tie tucked between the buttons and the cubicle geography do most of the recognition work before the line even lands.
The replace step uses multi-character handling: the running man gets the recast, the seated workers stay at their desks. A position-mapping prompt names the runner as the target and the seated workers as preserved elements.
The interior of a generic mid-2000s American beige open-plan corporate office bullpen at midday, in a photoreal workplace comedy film. The bullpen has rows of beige fabric-panelled cubicle dividers, a sea of desks with old beige desktop computer monitors and dual stacked paper trays, fluorescent overhead lighting reflecting off the beige industrial carpet, a small office kitchenette with a tall blue water cooler visible at the far back, and grey-paneled office walls. A MAN in his mid-forties runs from the back of the bullpen toward the camera through the centre aisle of the frame, dressed in a beige-and-tan striped dress shirt tucked into pleated grey-brown dress pants, a neatly knotted maroon-and-grey striped necktie tucked between the buttons of his shirt at sternum level, and plain black dress shoes. He has slightly thinning dark brown hair side-parted, soft pale skin, and a panicked-thrilled facial expression with his eyes wide and his mouth wide open. Both his arms are raised high above his head and flail in big alternating circles as he runs. As he runs forward, he yells loudly and excitedly in an enthusiastic MALE voice: "IT'S HAPPENING! EVERYBODY STAY CALM!" Around him in soft focus, several seated office WORKERS at their desks look up alarmed from their work to track him as he passes their cubicles, their heads turning to follow him, but they otherwise stay seated. Static medium-wide shot framing him from knees up at eye level, straight-on, the centre aisle of the bullpen stretching back behind him with the kitchenette far in the distance. Slightly cool fluorescent overhead lighting, faint shadows, classic mid-2000s digital video colour grade with slightly desaturated beige tones, slight digital grain. Photorealistic. The clean panicked-excited spoken line central in the audio with quiet office ambience underneath including the soft hum of computers and the distant click of keyboards, no other foreground dialogue.
Replace the man running through the centre aisle of the office bullpen in the source video, the one with his arms raised high and flailing as he runs toward the camera yelling, with the man from the reference image. Preserve the source video motion, audio, camera, lighting, and the seated office workers at their desks tracking him as he passes exactly as they appear in the source.
The character on the cast list
Every recast in this guide uses the same reference: Mo, a friendly muppet-style felt puppet. The mismatch is the point: dropping a felt puppet into the Reservoir Dogs suit walk, the Pulp Fiction twist dance, the Cast Away raft scream, the Jerry Maguire phone-call yell, the 300 Spartan war cry, the moonlit E.T. bike ride, the office bullpen panic run, and the apartment interrogation in the hero above is what makes the recast funny. The same pattern works for any reference image you want to cast: a real person, a creator avatar, a brand mascot, a stylized cartoon character. The Seedance source determines the costume the role wears (because the costume is part of the on-camera appearance the source video carries), so the reference image only needs the face, hair, build, and overall identity of who you want in the role.
A photoreal close-up product photograph of a friendly muppet-style felt hand puppet character named Mo, framed head-and-shoulders as a portrait. Mo has a round soft lavender-purple felt head with visible stitching seams running along the side and across the top. Two large round googly eyes with black pupils that wobble inside clear plastic casings. A small embroidered orange triangle nose. A wide stitched-on red felt smile across the lower face. A bright fluffy tuft of orange-red yarn hair sprouting from the top of his head in a small messy puff. He is wearing a small horizontally striped t-shirt in navy blue and cream that covers his felt shoulders. Plain pale grey seamless studio background, soft even three-point lighting, gentle soft shadow under him. He is positioned facing the camera directly, head and shoulders visible. Photorealistic close-up of a high-quality handmade felt puppet character, modern children's TV show production quality, every fibre and stitch sharp and visible.
The reference's striped shirt and grey backdrop don't transfer to the output. Only what makes Mo recognizable as Mo does. The wardrobe (the dance-floor black suit, the castaway's torn canvas shirt, the striped office shirt with tie, the red hooded sweatshirt, the crimson cape, the sports-agent dress shirt with loose tie, the apartment-scene suit and tie, the matching black suit jackets), the lighting, the setting, and the action all come from the source video.
Tips
- The source carries the costume, the reference carries the identity. The wardrobe, lighting, and setting in the output all come from the source video. Your reference only needs to be a clean portrait: face, hair, build, identity. Don't dress the reference up to match the source's scene. The source already handles that.
- Use a position-mapping prompt for multi-character or held-object scenes. When the iconic shot has two or more on-camera characters, or includes a small held or nearby object that could read as the recast slot, name the target by position ("the man on the left," "the bearded man on the raft") and list every other character or object as preserved. The model holds the listed elements untouched.
- One source, many casts. A clean source video for an iconic shot is reusable. Cast a real person, then a stylised avatar, then a brand mascot, against the same source video file each time. You pay for one Seedance run and as many replace runs as you want recasts.
- Mismatches are the joke. For the meme angle to land, choose a reference that's incongruous with the scene. A felt puppet in the Reservoir Dogs suit walk, or a corporate mascot doing the Pulp Fiction twist. The further the reference sits from what the role usually looks like, the funnier the recast.
- Bring your own source video if you have one. Replace doesn't care whether the source was generated, recorded, or licensed. If you have a clip you can use (your own footage, a clip you have rights to, or one you're remixing under fair use for parody or commentary), feed it directly into the replace step.
- For action-heavy or extreme-camera sources, expect identity drift. Whip pans and back-of-head framing give the model less of the character's face to anchor on. The cleaner and more static the source's framing, the tighter the recast's identity hold.