Replace the selected person in <Picture 1> with the character in <Picture 2>. Keep the reference character's face, hair and clothing. Follow the source performance's pose, gaze, expression and movement. Preserve the scene and the other people.
subject_definitions:
<Subject 1> is a young Japanese woman. Her face is the face in <Picture 2>.
summary:
[reference generation] Faceswap: <Picture 1> becomes the video as it is, and her face is the face of <Subject 1> from <Picture 2> in every frame.
retention_analysis:
<Picture 1>: fully_preserved - everything appears as it is; only the face is different.
<Subject 1> (appears in [Shot 1]): fully_preserved - her face from <Picture 2> appears in every frame.
detailed_description:
[Shot 1] The shot is <Picture 1> itself. Her face is the face of <Subject 1> from <Picture 2>.
I decided to test my choreography skills with MiniMax H3. This model never ceases to impress me—it kept 30 background dancers consistent throughout the choreography. I’m even using it to bring multi-shot storyboards to life.
MiniMax delivered a magnificent dance MV: striking studio visuals and an epic performance. 🔥
Sound on—what do you think? 🎬🔊🔥
The face drift was a lesson in my workflow: I started with one image and only created a proper character sheet later. I wouldn’t put that entirely on the model.
The whole experiment cost around 6K credits.
🔎Prompts for the complete film
Shared reference and character lock
Use this with all six segments. Attach the canonical character sheet, the relevant scene/handoff image, and a matching audio excerpt where applicable.
subject_definitions:@Image1 supplies the appearance of ONE adult female lead. Its multiple views depict the same woman, not additional performers.
Preserve her recognizable face: dense fine freckles across the nose, cheeks and forehead; high cheekbones; narrow nasal bridge; almond-shaped eyes; full red lips; sharply defined winged eyeliner.
Hair: long blonde, center-parted loose waves extending to mid-back, flowing outside the visor and over both shoulders. Maintain the same length and texture through every camera angle.
Costume: translucent lime sports visor with visible eye direction; fitted black short-sleeved technical jersey with lime contoured piping; separate lime wrap athletic skirt with a black numeral “8” on the wearer’s right front panel; opaque black mid-thigh cycling shorts underneath; lime wristbands; flexible black high-top dance footwear with lime accents.
@Image2 supplies the scene composition, supporting dancers, starting pose and studio continuity. It does not override the canonical lead identity.
Ensemble: approximately thirty distinct adult dancers wearing black sleeveless athletic tops, black wide-leg trousers and dance footwear. Only the lead wears lime.
Studio: white cyclorama, black-and-white radial ceiling panels, glossy black floor with physically plausible reflections and clearly separated bodies. Hard high-key lighting, crisp shadows, commercial fashion-film contrast, restrained fine grain.
Movement: explosive, grounded, professional athletic choreography. Visible weight transfers, planted feet, controlled joints and safe spacing.
No singing or lip-sync. No face replacement, shortened hair, costume changes, duplicated lead, merged limbs or random text.
Clip A — Face hook and formation · 0–6s
References: opening portrait, costume sheet, ensemble-stage image.
summary:Six-second opening: a motionless face hook explodes into a synchronized ensemble reveal and advancing arm strikes.
retention_analysis:Preserve the lead’s original freckled face, long blonde waves, lime visor and technical dance costume. Retain the radial studio and mirrored floor.
detailed_description:[Shot 1 — 0–1.5s]Static extreme close-up. The lead stares directly into the lens, chin slightly lowered. Two male dancers lean close on either side without touching or concealing her face. Hold stillness while a ceiling stripe glints across the visor.
[Shot 2 — 1.5–2s]Her head snaps right. Blonde hair sweeps across the frame. Follow with one short whip pan, resolving quickly into sharp studio geometry.
[Shot 3 — 2–4s]Hard cut to a wide, slightly low frontal shot. The ensemble forms a symmetrical block behind her. Everyone lowers into a grounded squat, crosses fists at chest height, then snaps arms outward in clean synchronized hits.Large condensed uppercase “HIT” appears behind the formation, briefly readable before a lime registration edge snaps into alignment.
[Shot 4 — 4–6s]Low frontal tracking retreats as the lead advances two heavy steps, plants a stable wide stance and drives alternating diagonal arm strikes. Supporting dancers answer in a staggered front-to-back wave.End with the lead standing front-center and the ensemble upright behind her.
overall_soundscape:Dry stomps, clipped exhales, fabric snaps and short studio reverb.
non_diegetic_music:Add the opening section of the existing dance track in the final edit.
Clip B — Kneeling wave and isolation · 6–11s
References: canonical sheet and Clip A’s handoff.
summary:Five-second continuation: a kneeling ensemble wave, a precise solo isolation and a controlled transition into floor choreography.
retention_analysis:Begin from the preceding stance. Preserve the lead’s exact face, long hair, separate jersey and skirt, supporting identities and studio lighting.
detailed_description:[Shot 1 — 0–2s]Low frontal medium-wide shot with a controlled upward tilt. The lead remains standing while the ensemble lowers onto one knee in staggered rows. Arms thrust diagonally toward the radial ceiling in a traveling wave, then sweep down.
[Shot 2 — 2–4s]Hard cut to a tilted medium shot with a slow push-in. The lead performs one ribcage isolation, transfers weight laterally and makes a controlled upper-back extension with knees flexed. Her arm traces downward; long hair follows the torso’s inertia. Return to a stable axis.Condensed uppercase “BREAK” appears behind her shoulder, briefly separating into aligned registration plates before rejoining.
[Shot 3 — 4–5s]Hard cut to an overhead view. The lead lowers through a bent knee, plants one palm and slides into a grounded floor pose. The ensemble folds down in an outward ripple, leaving visible gaps.End on a stable floor formation.
overall_soundscape:Staggered knee and boot contacts, controlled breaths, palm contacts and fabric slides.
non_diegetic_music:Continue the same master dance track in the final edit.
Clip C — Repaired floor finale · 11–16s
References: canonical sheet, corrected floor handoff, original face close-up.
summary:Five-second floor finale with restored facial identity and long blonde hair, ending in a floor-level hero close-up.
retention_analysis:The canonical sheet and original portrait control the lead’s identity. The corrected floor frame supplies pose, ensemble layout and camera geography. Keep the jersey and skirt separate.
detailed_description:[Shot 1 — 0–2s]High overhead view matching the corrected floor formation. Rotate slowly clockwise as the dancers execute one grounded contraction-and-extension ripple: knees fold, arms extend, torsos turn and settle. Maintain visible gaps. The lead remains central; her long hair follows the movement naturally.
[Shot 2 — 2–3.5s]Outer dancers push through hands and feet to stand, forming two banks around a clear central lane. The lead stays low and prepares a grounded athletic bear crawl.A brief camera roll ends in a hard cut to floor level—not a descent through bodies.
[Shot 3 — 3.5–5s]Floor-level medium close-up. The lead advances one deliberate grounded movement toward the lens, stops and lifts her chin into direct eye contact.Reproduce her original freckles, cheekbones, lips and eyeliner. Long blonde waves remain visible over both shoulders.“OWN IT” appears behind her, with a lime registration edge aligning on the final impact. Brief final hold.
overall_soundscape:Palm slaps, boot impacts, fabric movement and a collective exhale.
non_diegetic_music:Retain the existing master track.
Clip D — Rise and attack · 16–22s
References: canonical sheet, repaired hero frame, corresponding six-second music excerpt.
summary:Six-second continuation from the floor pose: an explosive rise, advancing stomps and sharp elbow accents.
retention_analysis:Preserve the canonical face, mid-back blonde waves, current costume and studio. @Audio1 supplies this segment’s musical rhythm; it does not require the lead to sing.
detailed_description:[Shot 1 — 0–2s]Low medium tracking begins from the crouched hero pose. The lead presses through grounded palms, draws one boot beneath her hip, pushes into a strong lunge and rises with visible weight transfer. The ensemble splits outward on kick accents.
[Shot 2 — 2–4.5s]Low frontal full-body tracking retreats as she advances two heavy stomps to the kick pulse. She drives right and left elbow strikes away from her torso, followed by precise shoulder hits on the backbeat. No contact with other dancers.The ensemble answers in a back-to-front wave.
[Shot 3 — 4.5–6s]Medium frontal shot eases closer. She resets into an upright wide stance, fists at chest height, lifts her chin and fixes the lens with a fierce gaze.Hold a clear facial identity checkpoint while her long hair settles.End in a stable stance.
overall_soundscape:Boot impacts, clothing snaps and controlled breath.
non_diegetic_music:Use @Audio1 as the rhythmic guide. Final assembly uses the corresponding original master-track section.
Clip E — Formation battle · 22–28s
References: canonical sheet, Clip D handoff, corresponding music excerpt.
summary:Six-second formation battle: diagonal strikes, opposing ensemble waves and a decisive synchronized freeze.
retention_analysis:Preserve the preceding stance, exact lead identity and long hair. Supporting dancers remain distinct, with clear travel lanes.
detailed_description:[Shot 1 — 0–2.5s]Lateral full-body tracking. The lead drives alternating diagonal arm strikes with grounded step-cross-step footwork. Each boot plants before the next directional change.The left ensemble bank answers first; the right bank follows on the next accent.
[Shot 2 — 2.5–4.5s]Hard cut to overhead. Two diagonal banks travel into a staggered symmetrical wedge. Each dancer takes two short marked steps, then plants safely.Arms sweep upward into a sharp chest-height stop. The entire formation freezes on a strong musical accent.
[Shot 3 — 4.5–6s]Cut to a frontal medium close-up. The lead gives one final shoulder accent, then holds still. Her eyes remain focused through the visor; long hair continues moving briefly after her body stops.End front-center, ready for the next phrase.
overall_soundscape:Alternating boot impacts, clipped breaths and fabric snaps.
non_diegetic_music:Use @Audio1 for movement timing; preserve the original master track in the edit. No lip-sync.
Clip F — Final takeover · 28–34s
References: canonical sheet, Clip E handoff, final music excerpt.
summary:Six-second final takeover: powerful lateral lunges, a grounded half-turn, a formation split and a planted hero finish.
retention_analysis:Maintain the lead’s original face, freckles, long center-parted waves, visor and costume throughout the turn and final close-up.
detailed_description:[Shot 1 — 0–2.5s]Medium-full frontal view from the preceding stance. During the musical swell, the lead sinks into a strong left lateral lunge, transfers right and delivers two sharp torso-and-shoulder hits.The ensemble contracts diagonally toward center while preserving clear gaps.
[Shot 2 — 2.5–4s]Hard cut overhead. On the track’s drop, the lead plants and executes a powerful grounded half-turn. Her long hair follows late.The ensemble opens into two banks, forming a clear central lane. Arms snap outward on kick accents, then stop.
[Shot 3 — 4–6s]Low frontal tracking retreats as the lead advances two heavy steps through the lane. She plants into a wide hero stance, finishes with a sharp elbow stop and lifts her chin toward the lens.End on a medium close-up of her canonical freckled face and long blonde waves.Large condensed “OWN IT” appears behind her; the lime registration edge aligns on the final impact. Hold the pose, then hard cut to black.
overall_soundscape:Stomps, clipped exhales, fabric snaps and a short room-reverb tail.
non_diegetic_music:Use the supplied music phrase for movement timing. Final assembly retains the continuous original song.
Prompt:
integrated_multimodal_description: [Shot 1] Cinematic, live-action, an extreme close-up frames the face of a young Japanese female idol. She has flawless, radiant skin, sparkling eyes, and soft pink lips with subtle glittery eyeshadow. The camera holds a static shot as she stares playfully into the lens with a bright, charming expression, while a soft light glint passes across her face. [Shot 2] At 00:03.000, the shot cuts to a wide, low-angle frontal shot in a professional dance studio featuring a white cyclorama, pastel-colored ceiling panels, and a glossy white floor with soft reflections. The lead, dressed in a cute pastel pink and white cropped idol top with a deep sweetheart neckline revealing her cleavage, and a matching mini skirt with ribbons and lace, revealing her midriff, stands before an ensemble of thirty young Japanese female idols in coordinated pastel blue and white cropped tops with low-cut necklines and mini skirts. The camera pans right with large amplitude at fast speed, resolving into the symmetrical formation. The ensemble lowers into a grounded squat and snaps their arms outward in synchronized, energetic hits. A large condensed uppercase "HIT" appears behind the formation, with a soft pink sparkling edge snapping into alignment. [Shot 3] At 00:07.000, the camera cuts to a low frontal tracking shot that retreats as the lead advances two rhythmic, bouncy steps. She drives alternating diagonal arm strikes with energetic footwork while the ensemble answers in a staggered, front-to-back wave. [Shot 4] At 00:11.000, the shot cuts to a floor-level medium close-up. The lead advances one deliberate, bouncy movement toward the lens, stops, and tilts her head to meet the camera with a bright, infectious smile. Her long, flowing hair, smooth midriff, and flawless, radiant skin are captured in high-key, bright lighting. A large condensed "OWN IT" appears behind her, with a soft pink sparkling edge aligning on the final impact. The camera holds the pose for a brief moment before a hard cut to black.
overall_soundscape: Light, rhythmic footsteps echo on the glossy floor, accompanied by cheerful, synchronized breathing and the soft rustle of layered stage costumes. A short, bright studio reverb tail follows the final movement.
non_diegetic_music: An upbeat, bubbly J-pop track featuring bright synthesizers, sparkling percussion, and a catchy, high-tempo melody that synchronizes precisely with the dancers' energetic movements and strikes.
introducing the all new Krea Video.
produce Hollywood-grade footage with the world's most powerful videos right in your browser.
cinematic camera controls, 100+ viral presets, UGC-ready templates, video editing, style transfer, sound generation.
try it now 👇