AI Dance Filter

GLOBAL LOCK: A photoreal vertical dance-transfer demo video using a fixed left-side instructional strip labeled “WAN 2.2 Swap.” Keep the composition consistent across all frames: a narrow left panel showing two stacked reference images with a yellow arrow and the text “WAN 2.2 Swap,” plus the main dance area on the right taking most of the frame. Keep the dancer consistent: young East Asian woman, fair skin, slim fit build, long dark hair down, round glasses, calm playful expression, full black fitted unitard or tight black one-piece outfit, barefoot. Keep the environment locked: simple empty indoor room with beige walls, light floor, soft natural light, minimal clutter. Motion is a copied viral dance with side steps, cross-steps, arm flicks, small hip shifts, and playful bounce timing. The face should remain stable even during body movement. No dialogue, no extra subtitles beyond the built-in left-side demo strip.

[00:00-00:03] Open with the dancer already stepping lightly across the floor while the WAN 2.2 Swap reference strip is visible on the left. She performs a smooth cross-step and small hand flick, making it clear this is a dance-transfer proof clip, not a cinematic scene.

[00:03-00:06] The dance gains confidence with a relaxed smile and more readable footwork. She shifts weight from one leg to the other, bringing one arm up in a playful gesture. Keep the room empty and visually quiet so the motion stays easy to read.

[00:06-00:09] She rotates her torso slightly and steps wider, adding a soft bounce and shoulder rhythm. Hair should move naturally without breaking facial identity. The black one-piece outfit must remain clean and form-fitting.

[00:09-00:12] The choreography becomes a little more expressive, with arms lifting and a side sway. The clip should still feel like a casual dance test generated from a reference rather than a polished music video.

[00:12-00:15] Final beat settles into a forward-facing pose after a last cross-step. End with the dancer centered and readable, proving that the identity swap or motion-transfer held through the full dance phrase.

NEGATIVE PROMPT: missing left reference strip, unreadable WAN 2.2 Swap text, duplicated limbs, broken feet, mutated hands, face drift, outfit color change, shoes appearing, dramatic camera zooms, cluttered room, subtitles, logos, watermarks beyond the intended strip, low-detail hair, unnatural dance timing, robotic stiffness, background changes.

SHOT PROMPTS:
SHOT 1 DELTA: establish WAN 2.2 Swap demo layout with dancer entering a cross-step pattern.
SHOT 2 DELTA: playful hand flick and relaxed smile, barefoot dance readability emphasized.
SHOT 3 DELTA: torso turn and wider side-step, hair moves naturally while face stays stable.
SHOT 4 DELTA: more expressive arm lift and bounce rhythm in the empty room.
SHOT 5 DELTA: final forward-facing pose after last cross-step, clean motion-transfer payoff.

SPEECH PACK:
Timecoded transcript: no spoken dialogue is present in the reference clip.
TAKE_A [00:00-00:15]: silent dance-transfer demo, no speech.
TAKE_B [00:00-00:15]: no spoken words, motion-copy showcase only.
TAKE_C [00:00-00:15]: quiet WAN 2.2 Swap demonstration of a viral dance in a plain room.
Closest audible version: no intelligible dialogue detected.
Safe paraphrase version: a woman in a black fitted outfit performs a copied viral dance while a left-side WAN 2.2 Swap reference strip shows the source setup.

GLOBAL LOCK: A vertical 4:5 AI dance-swap demo layout. Left side is a dark teal instructional sidebar showing two stacked reference images connected by a yellow curved arrow, with white/yellow text reading “WAN 2.2 swap.” Right/main side shows the generated output: a young woman AI influencer standing outdoors on a rocky riverbed/field edge with green grass and tall trees behind her. Keep the woman’s identity consistent: long black hair, glasses, hoop earrings, light skin, slim build, fitted sleeveless black romper, soft smile, and casual dance energy. The clip demonstrates motion transfer from a dance reference onto a static AI influencer image.

[00:00-00:02] Start with the woman standing front-facing in the outdoor location, body mostly still, arms relaxed near her sides. She looks into camera with a calm pleasant expression. The left tutorial sidebar remains visible with the two input images and the yellow curved arrow pointing down toward the “WAN 2.2 swap” label.

[00:02-00:04] The dance begins subtly. She lifts one arm outward and starts a small side-to-side upper-body sway. Her head tilts slightly, glasses remain aligned, and long hair stays smooth over the shoulders and back. The outdoor background stays bright and slightly soft, emphasizing the character rather than the scenery.

[00:04-00:06] The motion transfer becomes clearer: her shoulders and elbows move in a simple rhythmic dance, and one knee or hip angle shifts lightly as if following a reference choreography. The movement stays close to camera and mostly upper-body dominant, which helps preserve facial consistency. The expression brightens into a wider smile.

[00:06-00:08] Continue the playful dance with hand gestures closer to the torso and slight alternating arm positions. Her body remains mostly centered, with only small weight shifts. Keep the black romper fitted and stable, avoid fabric glitches, and preserve the clean face identity and glasses.

[00:08-00:11] End on the clearest dance-swap payoff: she smiles directly at camera while doing small finger-heart or pinched-finger style gestures with both hands near chest height, hips slightly angled. The result should feel charming and social-media friendly rather than technically perfect, with emphasis on identity preservation during simple choreography. The left-side instructional column and “WAN 2.2 swap” label remain on screen to underline the workflow.

NEGATIVE PROMPT: broken fingers, warped elbows, melted face, drifting glasses, identity swap, floating feet, broken knees, impossible hip twist, random camera zoom, missing sidebar, unreadable text, extra people, messy hair deformation, outfit flicker, body wobble, low-res landscape, overblown highlights, dance motion too large, face losing consistency.

SHOT PROMPT DELTA: tutorial demo layout, left reference sidebar, right generated influencer dancing outdoors, simple social dance, soft smile, black sleeveless romper, glasses and long hair stable, motion transfer test for WAN 2.2.

GLOBAL LOCK: A photoreal vertical split-layout demo video showing AI motion-transfer from a reference dance clip onto a consistent influencer character. Preserve the full format across all frames: a narrow left-side instructional panel with two small stacked reference images and bold text reading “WAN 2.2 Animate”, plus the main right-side performance area filling most of the frame. Keep the dancing subject consistent: young East Asian woman, fair skin, slim athletic build, long black hair in a high ponytail, expressive face, natural makeup, energetic but controlled smile. Wardrobe is locked: shiny red satin camisole or corset-style top with thin straps, fitted black high-waisted shorts. Environment is locked: bright minimal apartment or empty room with gray floor, white walls, open doorway, and a freestanding mirror in the back. Lighting is soft natural daylight from the front-left, realistic indoor brightness, no nightclub effects. Motion should clearly resemble a copied viral dance routine, with hands crossing, pointing, shoulder pops, and a gradual turn toward profile and back view. Keep the face identity stable even during arm motion. No dialogue, no subtitles beyond the built-in left-side label, no logos except the visible “WAN 2.2 Animate” text panel already present in the composition.

[00:00-00:02] Open with the dancer facing camera in the room while the left-side reference panel is already visible. She starts the dance in a relaxed stance, hips shifting lightly, one hand low and the other beginning to rise, establishing that this is a motion-copy demonstration rather than a cinematic music video.

[00:02-00:04] She brings both hands into the choreography with playful upper-body rhythm. The red satin top should catch soft daylight and stay glossy. Preserve the clean room, doorway, and mirror in the background without changing furniture or layout.

[00:04-00:06] The dance becomes more readable as she crosses one arm over the torso and points or sweeps the other hand outward. Her expression turns brighter and slightly cheeky, as if following a popular social-media dance challenge.

[00:06-00:08] She rotates into a three-quarter profile while continuing the same routine. Keep the ponytail swinging naturally but do not let the face or outfit mutate. The left-side panel with the source/reference images must remain fixed and legible throughout.

[00:08-00:10] Final beat transitions toward a back-facing pose with one hand lifting toward the hair. End like a tutorial proof-of-concept: the viewer should understand that the AI successfully transferred a reference dance onto the character while holding identity and outfit consistency.

NEGATIVE PROMPT: missing left panel, random UI overlays, broken text, mutated hands, duplicated arms, face drift, age changes, different outfit color, missing shorts, warped hips, extra dancers, crowded studio, nightclub lighting, dramatic cinematic camera movement, zoom crashes, smeared ponytail, broken mirror, furniture appearing suddenly, lip-sync speech, subtitles, watermarks beyond the intended layout, low-detail anatomy, jerky stop-motion motion.

SHOT PROMPTS:
SHOT 1 DELTA: front-facing dance start with visible WAN 2.2 Animate reference strip on the left.
SHOT 2 DELTA: playful hand choreography, red satin top catching daylight.
SHOT 3 DELTA: cross-body dance move, smile brightens, tutorial-demo energy.
SHOT 4 DELTA: rotate to three-quarter profile while preserving face consistency.
SHOT 5 DELTA: finish toward back pose with hair touch, clear motion-transfer payoff.

SPEECH PACK:
Timecoded transcript: no spoken dialogue is present in the reference clip.
TAKE_A [00:00-00:10]: silent dance-demo clip, no speech.
TAKE_B [00:00-00:10]: no spoken words, movement-transfer showcase only.
TAKE_C [00:00-00:10]: silent tutorial-style proof clip with visual dance performance.
Closest audible version: no intelligible dialogue detected.
Safe paraphrase version: a woman in a red satin top performs a copied viral dance in a bright room while a left-side panel shows the reference and WAN 2.2 Animate label.

GLOBAL LOCK: The subject is a young woman of Hispanic descent, approximately 22 years old, with olive skin, dark brown eyes, and long, straight black hair styled in a sleek, high ponytail. She has an athletic, toned build. She is wearing a matching black ribbed sports bra and high-waisted black mini shorts, paired with classic black stiletto high heels. The environment is a spacious, modern dance studio with white walls, large industrial-style windows, and a light grey, slightly reflective professional dance floor. Wall-to-wall mirrors are visible in the background, along with wooden ballet barres. The lighting is bright, natural, and high-key, coming from the windows. The color grade is clean and neutral with high clarity. No speech is present; the video is synced to upbeat dance music.

[00:00–00:02]
The subject walks confidently toward the camera from the center of the dance studio. She has a slight smile and looks directly at the lens. The camera is at eye level, capturing a full-body shot. The movement is smooth and rhythmic.

[00:02–00:04]
The subject begins the dance routine. She performs a quick series of arm gestures, crossing her hands in front of her chest and then throwing them outward. She performs a small, energetic jump with both feet leaving the floor. Her ponytail swings dynamically with the movement.

[00:04–00:06]
The subject transitions into a deep side lunge to her right, extending her left leg. She reaches her arms out toward the floor. The camera maintains a wide shot to capture the full range of motion. Reflections of her movements are visible on the polished floor.

[00:06–00:08]
She jumps back to a standing position and immediately places both hands behind her head, elbows out. She performs a rhythmic bounce/hop in place. The ponytail continues to show realistic physics, whipping behind her.

[00:08–00:10]
The subject performs a series of alternating side lunges. She extends her arms wide to the sides with each step. Her expression is focused and energetic. The lighting remains consistent, highlighting the muscle definition in her legs.

[00:10–00:12]
The subject completes the dance sequence with a final rhythmic step and then turns to her right, walking toward the side of the frame in a profile view. The camera follows her movement slightly. The video ends as she maintains her posture and walks out of the primary dance area.

NEGATIVE PROMPT: visual artifacts, flickering, distorted limbs, extra fingers, blurry face, inconsistent hair length, floating clothing, jittery background, robotic movement, unnatural joint angles, low resolution, watermarks, text overlays on the subject, mismatched reflections.

SPEECH PACK:
speech_present: false
music_style: Upbeat pop/dance, female vocals, high energy.
sync_notes: All major jumps and arm extensions must align with the rhythmic beats of the background track.

GLOBAL LOCK: Create a vertical tutorial-style AI motion-control reel that demonstrates how a Frida Kahlo-inspired woman and Diego Rivera-inspired man can be animated into a realistic couple dance. Preserve the recognizable art-inspired styling of both characters: Frida with floral hair adornments, traditional dress, and bold folk-art color accents; Diego with a fuller build, blue shirt, dark trousers, and painterly portrait realism. Structure the video as a workflow reel with three layers: final generated output, software interface walkthrough, and live-action dance reference. Keep motion fidelity, couple synchronization, and art-character identity stable throughout. No dialogue or lip sync.

[00:00-00:06.5] Open on the final generated result: a Frida-and-Diego-inspired painted couple dancing together in a warm indoor room. They step side to side, raise arms, and move in sync while a small reference dancer inset appears near the lower edge of frame. Keep the art style stable and the choreography readable.

[00:06.5-00:19.0] Cut to a dark software interface screen recording showing the motion-control workflow inside an AI tool. Display generation settings, control panels, and progress elements that explain how the dance is created from source footage. Keep the interface legible and clearly framed as a process demonstration.

[00:19.0-00:24.5] Show the live-action reference pair in a phone-like vertical frame performing the original couple dance. Their steps, arm lifts, and body timing should match the generated output logic. Keep this section straightforward and tutorial-oriented.

[00:24.5-00:35.2] Return to the final generated Frida-and-Diego dance result, now letting the viewer compare it mentally against the reference. Preserve the pair's stable identities, coordinated body movement, and painterly cultural styling while they continue dancing side by side in the warm interior.

Create a vertical AI video test that demonstrates copying a viral dance performance from a reference clip onto a static AI influencer image using WAN 2.2 Animate. The subject is a young brunette woman with her hair tied up in a casual bun, wearing thin glasses, a light blue floral mini dress with ruffled hem, and white knee-high boots. Place her outdoors on a tiled patio at dusk, framed by tall hedges, a wooden railing, and a large planter glowing with warm light.

Keep the camera locked in a full-body medium-wide view so the dance motion is easy to judge. The performance should feel like a social dance test rather than a polished music video: quick arm swings, side-to-side hip movement, small foot pivots, one pose with both arms extended, one with a hand touching her head, one with a hand on her hip, and one energetic bounce that lifts her hair upward from motion. Preserve the same face, glasses, dress pattern, and body proportions across every move. Prioritize consistency in facial identity while translating the reference choreography.

Visually present it like a creator demo reel. Add a slim vertical strip at the left that shows the two source images used for the transfer, connected by a plus sign, and place a small "WAN 2.2 Animate" label near the bottom so viewers understand which model generated the motion. The final effect should communicate that close-to-camera dance references can be copied onto a static AI character with decent consistency, while still feeling like a real benchmark of motion fidelity.

GLOBAL LOCK: A consistent young woman of Hispanic/Latina descent, mid-20s, with long dark hair, wearing black-framed glasses and a black beanie. She wears an oversized black-and-white graphic hoodie with street-art style prints, olive green cargo pants, and chunky white sneakers. The environment is a dimly lit industrial warehouse with exposed brick walls, colorful graffiti, and large factory windows. Lighting is a mix of warm overhead industrial lamps and cool natural light. Cinematic color grade, high contrast, sharp textures.

[00:00–00:03]
The subject stands in the center of the warehouse, facing the camera. She begins a rhythmic, low-energy bounce, swaying her hips slightly. The camera is a static medium-full shot. Lighting emphasizes the folds in her oversized hoodie.

[00:03–00:06]
The subject performs a fluid arm "wave" motion, crossing her arms in front of her chest and then extending them outward. She has a slight, confident smile. The motion is smooth and perfectly timed to a rhythmic beat.

[00:06–00:09]
The subject transitions into footwork, shifting her weight from side to side in a "shuffle" style. Her hands move rhythmically near her waist. The graffiti background remains sharp and stable.

[00:09–00:11]
The subject performs a chest-pop and a quick arm flourish, pointing towards the camera. Her glasses and beanie remain perfectly in place. The lighting creates a rim-light effect on her shoulders.

[00:11–00:13]
The subject finishes the dance with a final energetic pose, looking directly into the lens with a friendly expression. The video ends on a high-energy beat.

NEGATIVE PROMPT: Texture flickering, boiling clothes, face warping, extra limbs, blurry graffiti, robotic motion, sliding feet, inconsistent lighting, low resolution, watermark, text overlays on character, distorted glasses, hair clipping through beanie.

SPEECH PACK:
(No speech present in this video. The focus is entirely on rhythmic motion and music synchronization.)
TAKE_A: [Rhythmic breathing sounds synced to dance movements]
TAKE_B: [Silence, focus on ambient warehouse room tone]
TAKE_C: [Slight fabric rustle sounds during arm movements]

A vertical comparison reel showing AI video generation results from a single cinematic reference image. Each segment uses a split-screen stack: the top frame is labeled “REFERENCE IMAGE,” while the bottom frame shows the output from a specific model such as “SEEDANCE 2.0 OMNI” or “HEYGEN.” The example scenes are grounded, live-action-style dramatic setups rather than flashy VFX. One sequence shows a young man in a beanie talking with a woman on a wooden pier at dusk, with string lights and a lighthouse in the background. Another shows two men standing in front of a decaying Victorian haunted house under a grey overcast sky. A later scene places a smiling couple seated together in a subway car with cinematic teal-orange grading. The reel keeps the composition nearly identical between source and generated result to highlight motion fidelity, facial consistency, and realism across tools.

GLOBAL LOCK: A vertical social-media tutorial demo video showing a motion-control result for AI creator workflow. The main subject is one young adult woman with light-to-medium skin tone, long dark hair, slim build, large clear eyeglasses, hoop earrings, and a fitted black sleeveless mini dress, dancing barefoot in a simple beige concrete space with plain walls and open light from the side. The layout must keep a static split-style composition: on the left, a dark teal vertical sidebar containing two rounded-rectangle reference panels stacked top and bottom, a plus symbol between them, a curved arrow, and bold text reading “KLING 3.0 Motion Control”; on the right, the live motion result occupies most of the frame. Keep the camera locked-off, 4:5 vertical framing, soft natural daylight, low-production tutorial aesthetic, no scene cuts, and rhythmic side-to-side dance motion with arm gestures and stepping footwork.

[00:00-00:03] The dancer stands wide-legged facing the camera in the open beige room, smiling while beginning a simple side-to-side groove. Her black mini dress stays body-hugging and stable, and her glasses and hoop earrings remain visible. The left sidebar shows the top input pose image and the lower generated-dress result image, separated by a plus sign and arrow. Maintain a static tutorial composition with no camera movement.

[00:03-00:06] She continues the dance with small hip shifts, alternating arm swings and light shoulder bounces while staying centered in frame. One hand rises briefly near the head as the legs step outward and inward in rhythm. The plain room, concrete floor, and side light remain unchanged, reinforcing the raw test-video feeling.

[00:06-00:09] The movement becomes slightly more animated as she raises one arm higher, smiles more broadly, and shifts weight from one leg to the other. Her dress moves minimally with the steps, and the barefoot grounding remains visible. The left-side visual instructional stack stays fixed, with the “KLING 3.0 Motion Control” label continuously readable.

[00:09-00:12] She keeps the same dance phrase, adding a playful upper-body sway and a higher hand flick near the head while stepping laterally. The framing remains locked, with the main right-side result panel dominating the screen and the left sidebar functioning as a visual explanation of source pose plus result target.

[00:12-00:15] She transitions toward a finishing pose while still dancing lightly, crossing one leg forward and softening into a smaller, playful hand gesture near the face. The split-layout tutorial structure, beige practice room, black mini dress, glasses, and motion-control branding remain consistent through the end. Finish without cuts, without zooms, and with the same creator-education demo aesthetic.

GLOBAL LOCK: Vertical AI-tool explainer video with a single male creator speaking directly to camera in a warm, softly lit studio while dark product-demo panels appear above him. The host is a light-skinned man in his 20s or early 30s with side-parted brown hair, blue eyes, clean-shaven face, slim build, and animated pointing gestures. He wears an off-white knit sweater and speaks into a black podcast microphone centered in front of him. The product visuals belong to an AI aggregator environment featuring Seadream 5.0 and related generation tools, with clean black UI, rounded cards, and large preview images. Example outputs emphasize prompt adherence, object transformation, and character consistency: off-road car shots, consistent face close-ups, full-body portrait variations of the same man, and other stylized image/video results. The tone is informative, creator-focused, and positioned around all-in-one workflow convenience.

[00:00-00:06] Open on a dark product panel labeled like a quick tutorial, showing an input video of a vehicle tire on a sandy road and a processed result beneath it. The host appears below, pointing upward to emphasize how fast the workflow is. Vertical composition, clean UI edges, no camera motion.

[00:06-00:12] Continue with vehicle and object-generation examples above the host, comparing source and result imagery. The presenter speaks quickly and confidently, explaining that the model produces artistic outputs with strong consistency and accuracy.

[00:12-00:20] Shift into the Seadream 5.0 interface. A close crop of human eyes and facial detail appears, followed by smaller facial-region thumbnails. The host remains below, stressing prompt adherence and image quality. Keep the UI dark, minimal, and product-demo realistic.

[00:20-00:28] Show character consistency examples: a young man on a plain background, then multiple versions and close-up crops that preserve face structure across outputs. The host continues in a persuasive tutorial tone, pointing and nodding to the examples above.

[00:28-00:36] More interface views and stacked result cards appear, suggesting the tool can keep the same character stable while varying framing and scene treatment. The host stays centered, speaking into the mic with direct eye contact and crisp lip sync.

[00:36-00:43] Transition to comparison content or adjacent tools inside the same aggregator, including additional preview cards and motion-oriented samples. The presenter frames Artlist as an all-in-one place to access multiple AI tools without switching apps.

[00:43-00:48] End on another polished sample panel, including a dynamic performance shot labeled with a related model name, while the host delivers the CTA to comment for the link. Keep the final feeling practical: this is a creator workflow demo centered on convenience, consistency, and prompt fidelity.

GLOBAL LOCK: cinematic 1980s-style street-dance performance inspired by a fedora-wearing pop icon; central male dancer in black hat, black sequined or sharp black jacket, white shirt, black tie, dark trousers, and white gloves; moody industrial stage or subway-like set with backup dancers in dark suits; synchronized footwork, sharp arm hits, spins, and confrontational dance staging; cool blue-gray lighting with warm practical highlights; no text overlays, no logos, no fantasy elements, no modern casual outfits.

00:00-00:04
Open on the central fedora-wearing male dancer commanding the frame while backup dancers form a loose semicircle behind him. The performance space feels industrial and theatrical, with dramatic overhead lighting and strong contrast.

00:04-00:08
The choreography tightens into iconic pop-dance gestures: hat-brim emphasis, crisp upper-body hits, quick pivots, and face-forward attitude. Supporting dancers mirror and challenge the lead, creating a confrontational performance rhythm.

00:08-00:12
The scene expands into a larger group formation. The lead dancer drives the center while surrounding performers move in synchronized bursts, with kicks, slides, and sharp directional changes across the floor.

00:12-00:15
The routine resolves on the lead figure reclaiming center stage, framed by fallen or staggered dancers and strong pose-based finishing beats that preserve the music-video intensity.

NEGATIVE PROMPT:
bright daylight, empty studio, casual hoodies, neon cyberpunk effects, fantasy powers, readable text, UI panels, broken anatomy, low-energy movement, cartoon rendering, soft pastel palette, extra props, random crowd spectators

GLOBAL LOCK: A vertical 9:16 creator-education reel explaining Seedance Omni as a reference-driven AI video workflow. The piece should feel like an advanced AI creator breakdown rather than a generic ad. Maintain a fast-cut social-video rhythm with bold white kinetic captions placed near the lower third, each section introducing a new reference-media test. The entire reel is built as a visual proof montage showing what happens when multiple reference images, video clips, and audio cues are fed into a multimodal video model. Keep the tone analytical, slightly excited, and highly demonstrative.

The visual structure is a sequence of short proof-of-concept mini-scenes. Each test should look like an AI-generated or AI-remixed output with strong consistency to its reference set. Use clear on-screen labels such as REFERENCE IMAGE, REFERENCE VIDEO, REFERENCE AUDIO, or REFERENCE MULTISHOT where needed. The labels should feel like creator-tutorial overlays, not cinematic subtitles.

Speech and semantic lock: the narration should communicate that Seedance 2.0 includes a feature that may be even more transformative than the base model itself, namely Seedance Omni. The voiceover should explain that, similar to Kling Omni, Luma Modify, and Runway Aleph, the model can accept multiple references and supports combinations such as up to 9 reference images, 3 video clips, and 3 audio clips. The narration should frame the reel as a practical stress test across several very different scenarios. Delivery should be confident, conversational, and creator-native, as if an experienced AI educator is showing rapid experiments to other creators.

[00:00-00:07] Open on an extreme close-up of a human blue eye filling most of the vertical frame. The eyelid, skin pores, eyelashes, and iris should be crisp and realistic. Over a few beats, the iris mutates from a normal blue human iris into a reptilian or serpent-like yellow slit pupil while preserving the same eye shape, eyelid geometry, and camera angle. The transformation should feel smooth and uncanny rather than gory. Use tight macro framing, shallow depth of field, soft natural facial lighting, and highly detailed iris textures. Overlay creator-style white labels that indicate this is an image-reference identity preservation test.

[00:07-00:14] Continue the eye test with alternating states: blue human iris, reptile iris, then back toward a stable identity-preserved close-up. Maintain subject consistency, same eyelid folds, same skin tone, same eyebrow edge, same framing. The point is that the model is modifying a reference rather than generating a new person from scratch.

[00:14-00:22] Cut to a clean blue-sky and beach-horizon scene that initially appears minimal, then resolves into a POV vacation shot from a seated person looking toward the ocean. Visible in the foreground are tanned bare legs stretched out on a beach towel or lounge setup, with summer accessories nearby. A straw hat appears near the lower left or lower center. The shot should feel like a calm, sunny, lifestyle travel clip, captured handheld or gently stabilized. White text labels should imply that this is another reference-driven style or scene transfer test.

[00:22-00:30] Stay with the beach POV while weather and atmosphere subtly change. One beat should show a darker, overcast coastal sky while preserving the same composition and seated point of view. Another beat should return to bright daylight. The central idea is consistency of composition across different conditions. Keep the ocean horizon level, sand texture visible, and the viewer’s legs anchored in the same place.

[00:30-00:38] Transition to an urban street scene in a European-looking brick neighborhood. The camera faces down a quiet city block lined with red brick buildings, bare trees, parked cars, and pedestrians on the sidewalk. A person in a strong orange coat or jacket walks through the middle distance. Keep this sequence realistic, documentary-like, and gently stabilized, as if using reference video to maintain a consistent place while changing motion or timing. White labels should continue indicating the experiment mode.

[00:38-00:45] Hold on the same street layout for multiple beats, maintaining the same block geometry, winter trees, and brick façades while the walker’s position changes slightly. The reference-driven consistency matters more than dramatic action. It should read as an urban reference clip that is being preserved across variations.

[00:45-00:50] Cut hard to a green alpine meadow under soft daylight. A realistic white-and-brown cow stands in the foreground with rolling hills behind it. A person appears farther back in the field. The composition should feel like a pastoral documentary frame. Emphasize texture in the cow’s fur, the wet grass, and the cool mountain-air atmosphere.

[00:50-00:55] Shift within the same mountain meadow world to a large brown bear occupying a similar compositional role. Preserve the same field, same hills, same cloudy daylight, and same general camera position, as if one reference animal has been swapped or remixed while the environment remains locked. The contrast between cow and bear is part of the test.

[00:55-01:02] Cut to a dramatic open-ocean action sequence. A slim missile or rocket skims above the water surface from a distant background position toward the camera. As it advances, the rear exhaust glows orange and throws reflections across the dark blue sea. The shot escalates into a fiery low-altitude pass with explosive energy. Keep the ocean horizon broad and cinematic, with long-lens compression or stabilized action-footage framing.

[01:02-01:08] Intensify the missile sequence: the projectile is now closer, flames brighter, wake or reflected light streaking across the water. Motion should be fast but readable. Use cinematic contrast, orange fire against cool blue water, and a controlled action aesthetic. Overlay labels that suggest multi-reference video guidance or action transfer.

[01:08-01:15] Transition into a dusty automotive destruction scene on land. A vintage or older pickup truck drives through a battlefield-like environment with explosions erupting around it. Dirt plumes and debris rise behind and beside the vehicle. The camera angle stays low and frontal or front-three-quarter, preserving vehicle identity through multiple cuts. The shot should feel like a reference-conditioned action test rather than a polished Hollywood trailer.

[01:15-01:22] Continue the truck sequence with repeated passes of the same vehicle through the same dusty environment while explosion timing changes around it. Maintain truck consistency, same body shape, same color family, same old-metal texture, same framing logic. The action is intense but the main point is that the vehicle identity remains coherent through multiple high-energy beats.

[01:22-01:26] Close on a calmer atmospheric ocean-and-cloud frame or a visual reset that gives the reel a final exhale after the action montage. Let the ending feel like the creator has completed a string of experiments and proven the system across categories: facial modification, lifestyle scene consistency, city references, animal/environment swaps, maritime action, and explosion-heavy vehicle sequences.

Camera and edit language: every mini-scene should be concise and creator-friendly, with hard cuts every few seconds, no ornamental transitions, and persistent overlay text that contextualizes each experiment. The framing changes radically between scenes, but within each scene the composition should feel locked to the reference. Keep a social-reel cadence, as if each example is there to prove a single capability quickly.

Lighting and grade: use realistic lighting that matches the scene category. Macro eye shots should be soft and detailed. Beach shots should be sunlit and airy, with one overcast variation. City shots should feel naturally cold and muted. Meadow animal shots should be overcast alpine daylight. Ocean missile shots should be cinematic with cool blues and bright orange flame accents. Truck explosion scenes should use dusty golden-brown grading with high contrast and particulate haze. Overall, the grade should feel credible and reference-bound, not oversaturated AI slop.

Audio direction: use one primary narrator with clear, studio-clean voiceover. Pace should be brisk and informative, roughly 145 to 165 words per minute. Mic perspective should feel close and modern, with minimal room echo. The narration should hit phrase boundaries close to the visual transitions so each new capability lands with a new example. If subtle background music exists, keep it supportive and low, allowing the tutorial value to dominate. Important phrases include Seedance Omni, reference images, reference video, reference audio, and the claim that the feature may matter more than the base model itself.

Invariants to lock: the reel must remain a creator-analysis montage about multimodal references, not a random compilation. Keep white tutorial labels, fast social pacing, proof-oriented structure, and consistent reference preservation within each mini-scene. The eye must remain the same eye across mutation beats; the beach POV must remain the same seated composition; the city block must remain the same location; the alpine field must remain the same environment across animal swaps; the missile shot must preserve ocean horizon and missile approach logic; the truck must remain the same truck across explosions.

Variables allowed to drift: exact overlay wording, exact pedestrian positions in the city, cloud shapes, splash detail on the ocean, size and timing of explosions, and fine-grain movement intensity. Voice pitch can vary slightly, but the explanatory meaning and creator-native cadence should remain locked.

NEGATIVE PROMPT: avoid random unrelated scenes, generic montage aesthetics, meme editing, giant captions that block the frame, unrealistic face changes that alter the person entirely, beach shots without the seated POV legs, city shots without the orange-coated walker or brick-neighborhood feel, mountain scenes that look tropical, missile shots that read like space combat, and truck action that changes the vehicle identity every cut. Avoid excessive glitch effects, fantasy color grading, unreadable UI text, or ad-style polish that removes the practical testing vibe.

GLOBAL LOCK: 
The video must maintain a high-fidelity cinematic aesthetic throughout, characterized by photorealistic textures, complex physics, and consistent lighting. 
Subjects: Diverse range of characters including a young Asian woman in tactical gear, high-fantasy warriors, gritty survivors, historical figures (Isaac Newton, Stephen Hawking), and a realistic Will Smith. 
Environment: Varies from futuristic skylines and high-seas battles to post-apocalyptic tunnels and bright TV studios. 
Camera: Dynamic movements including high-angle tracking, handheld shaky-cam, and smooth pans. 
Lighting: Motivated by the scene (cool city lights, warm firelight, high-key studio lights). 
Color Grade: High dynamic range with specific palettes for each genre (teal/orange for fantasy, desaturated for grit). 
Speech: Enthusiastic male narrator (Kallaway), warm tone, crisp articulation, medium-fast pace.

[00:00–00:03]
Subject: A young Asian woman with dark hair, wearing a grey crop top and dark tactical pants.
Environment: A futuristic, sprawling city with sharp, glass-covered skyscrapers under a hazy, overcast sky.
Action: The woman jumps from a high ledge, falling toward the city streets with realistic hair and clothing physics.
Camera: A wide, high-angle tracking shot following her descent.
Lighting: Flat, cool daylight with soft shadows.
Speech: "This might be the Hollywood killer." (Kallaway, high energy, on-camera reaction below).

[00:04–00:13]
Subject: Dragons flying over a fleet of wooden ships; a hand in the foreground casting a water spell.
Environment: A vast ocean with turbulent waves and multiple ships on fire.
Action: Two massive water spouts rise from the sea in an 'X' shape, controlled by the foreground hand. Dragons fly through the smoke.
Camera: Dynamic wide shot, transitioning to a POV-style shot of the hand.
Lighting: Dramatic sunset lighting with bright blue magical glows and orange fire reflections.
Speech: "This is Seedance 2. It’s a brand new video model from ByteDance in China."

[00:14–00:22]
Subject: A rugged man with a beard and a young girl (resembling Joel and Ellie from The Last of Us).
Environment: A dark, overgrown concrete tunnel with light streaming from a distant opening.
Action: The man fights off a fungal-infected zombie, pushing it against a wall with intense, realistic struggle and fluid motion.
Camera: Close-up, handheld shaky-cam to emphasize the grit and tension.
Lighting: Low-key, moody lighting with high contrast and naturalistic highlights.
Speech: "And it will break the internet when it launches publicly in the next couple weeks."

[00:23–00:28]
Subject: Isaac Newton (with powdered wig) and Stephen Hawking (in his motorized wheelchair).
Environment: A classic library/study with bookshelves and a globe, overlaid with a fighting game UI.
Action: Newton rings a large bell that falls toward Hawking; Hawking deflects it with a rainbow-colored energy beam.
Camera: Medium shots alternating between the two characters, mimicking a 2.5D fighting game.
Lighting: Warm, theatrical library lighting.
Speech: "Or say you want to create an imaginary video game sequence between Stephen Hawking and Isaac Newton in Mortal Kombat."

[00:29–00:36]
Subject: A photorealistic Will Smith with a goatee, wearing a black t-shirt.
Environment: A modern, clean kitchen with a stainless steel refrigerator and a glass of red wine.
Action: Will Smith picks up a large forkful of spaghetti and eats it with realistic mouth movements, sauce staining his lips, and a satisfied expression.
Camera: Close-up macro shot focusing on the face and the food.
Lighting: Bright, natural indoor lighting.
Speech: "And of course we can't forget the coveted Will Smith spaghetti test that this model absolutely destroys. Mmm, that's the good stuff."

[00:37–00:46]
Subject: A young woman in a light blue athletic outfit.
Environment: A brightly lit TV game show set with a large audience in the background and a pool of water below.
Action: She runs across a narrow balance beam while being hit by a rotating padded obstacle, maintaining realistic balance and physics.
Camera: Medium tracking shot following her movement across the beam.
Lighting: High-key, multi-source studio lighting.
Speech: "This is so realistic I would never know the difference if this was shown on a feed."

[00:47–00:53]
Subject: A man resembling Donnie Yen in a black martial arts suit and an anime girl with pink hair in a school uniform.
Environment: A traditional Chinese courtyard with a pagoda in the background.
Action: The two engage in a stylized martial arts spar, with the anime character emitting pink energy from her fists.
Camera: Profile medium shot, fast-paced editing.
Lighting: Soft, cinematic daylight.
Speech: "When it does, I'll make another video fully experimenting with everything you can do."

[00:54–01:04]
Subject: Creator (Kallaway) in a black cap and sweatshirt.
Environment: A professional studio with a computer screen showing the Artlist website.
Action: The creator points to the screen and gestures toward the camera, concluding the review.
Camera: Medium shot, static.
Lighting: Warm studio lighting with a purple/blue accent light in the background.
Speech: "Until then, on Artlist you can already use the Seedream image model... stay tuned because when this comes out, it is going to be a huge leap forward."

NEGATIVE PROMPT: 
Visual: Morphing limbs, flickering backgrounds, distorted faces, unnatural liquid physics (except where intended), blurry textures, low-resolution artifacts, robotic movement, inconsistent lighting between frames.
Speech: Robotic cadence, flat delivery, muffled audio, background hiss, lip-sync delay, unnatural pauses, mispronunciation of "Seedance" or "ByteDance".

SPEECH PACK:
[00:04-00:06] "This might be the Hollywood killer."
TAKE_A: (Breathless, excited) "This... might be the Hollywood killer!"
TAKE_B: (Serious, authoritative) "This might be the Hollywood killer."
TAKE_C: (Whispered, conspiratorial) "This... might be the Hollywood killer."

[00:29-00:33] "And of course we can't forget the coveted Will Smith spaghetti test."
TAKE_A: (Laughing) "And of course, we can't forget the *coveted* Will Smith spaghetti test!"
TAKE_B: (Impressive) "And of course, we can't forget the coveted Will Smith spaghetti test."
TAKE_C: (Nostalgic) "And of course, we can't forget that famous Will Smith spaghetti test."

[01:00-01:04] "It is going to be a huge leap forward."
TAKE_A: (Punchy) "It is going to be a HUGE leap forward."
TAKE_B: (Confident) "It's going to be a massive leap forward for all of us."
TAKE_C: (Finality) "This is the leap forward we've been waiting for."

GLOBAL LOCK: vertical 9:16 comparison reel split into two stacked halves. Top half labeled `AI:` shows the transformed cinematic version of the same man. Bottom half labeled `Original:` shows the raw talking-head recording of the creator performing hand gestures against a plain indoor backdrop. The subject identity must remain the same across both halves: young adult male, short brown hair, light skin, expressive face, medium build.

MASTER INTENT: create a short before-and-after AI transformation reel where each gesture in the original footage is mirrored by a stylized cinematic conversion in the top half. The AI version should progressively change wardrobe, mood, and environment while preserving timing and body movement from the original clip. End with a comment CTA for the guide.

00:00:00-00:00:02
Open with the creator pointing to his head in the original lower half while the upper AI half presents a cleaned-up enhanced version of the same pose. Simple gray studio background below; upgraded styling above.

00:00:02-00:00:04
Shift the AI half into more dramatic wardrobe changes: open shirt styling, then black leather jacket and sunglasses, while the original lower half remains plain and casual. Keep the gesture timing synchronized between top and bottom.

00:00:04-00:00:06
Move into higher-intensity transformations. The AI half places the same man in a warmer dramatic environment, including moody background lighting and a stronger cinematic grade. The original lower half still shows the untouched performer moving through the same gesture.

00:00:06-00:00:08
Push the transformation further with a fiery action-style background in the AI half. Flames or bright orange effects appear behind the subject while he continues reacting with raised hands and animated facial expression. Overlay a bold CTA near the lower section of the AI frame: `comment "AI" for guide`.

00:00:08-00:00:09.3
End on the strongest before-and-after comparison, holding the transformed fiery look on top and the original gesture on the bottom long enough for the CTA to be read clearly.

CAMERA: static front-facing camera, medium framing, same timing across original and transformed versions.

LIGHTING: original footage uses flat indoor creator lighting; AI version upgrades this into fashion-commercial and action-movie lighting depending on the segment, including clean studio light, editorial contrast, and warm flame reflections.

GRADE: original remains natural and unpolished; AI side becomes crisp, cinematic, contrasty, and style-forward.

MOTION: gesture-synced transformation reel, no camera shake, fast jump cuts between AI styling states.

TEXT PACK: exact visible labels `AI:` and `Original:`; final CTA `comment "AI" for guide`.

NEGATIVE PROMPT: different person in AI half, broken hand sync, mismatched gestures, unstable split-screen, unreadable labels, warped sunglasses, cartoon flames, overprocessed skin, blurred original footage, extra subtitles, watermark, logo corruption.

GLOBAL LOCK: Preserve the exact vertical promo-edit structure of a dark combat showcase branded as SEEDANCE 2.0. Keep the persistent upper graphic strip containing multiple small preview thumbnails and the bold text label “SEEDANCE 2.0” across the top portion of the frame. In the main lower action panel, maintain two Mortal Kombat-style masked ninja fighters in a smoky, high-contrast gray-black environment, facing off in side-view and three-quarter combat compositions. One fighter should read as a gray-blue Sub-Zero-like character, while the opposing fighter reads as a darker black-clad Scorpion-like rival. The action should cycle through weapon clashes, dodges, lunges, jumps, and dramatic distance changes, all with a game-trailer feel rather than a naturalistic movie scene. Do not remove the promo UI overlay, and do not turn the characters into generic modern soldiers.

0.00-4.00s: Open with the SEEDANCE 2.0 title and thumbnail strip already visible at the top, while the main panel below shows the two masked ninjas in close-range confrontation. Their bodies are angled aggressively, weapons and arms raised, with fog and backlight shaping the silhouettes in a dark arena-like void.

4.00-8.00s: Expand into wider combat beats where the fighters separate, circle, and re-engage. Show quick lateral movement, one fighter springing or stepping backward while the other advances through the smoky space. The monochrome visual palette should stay gritty and game-cinematic.

8.00-12.50s: Continue the duel with sharper impact moments: side-on exchanges, airborne kicks or leaps, and strong silhouette poses that resemble a fighting-game special move showcase. The upper thumbnail bar and title remain static, reinforcing that this is a promo package rather than diegetic footage.

12.50-17.00s: Intercut more aggressive stances and dramatic pause moments. The Sub-Zero-like and Scorpion-like figures should feel evenly matched, with alternating offensive and defensive beats. Smoke plumes, energy-like blur, and heavy contrast keep the environment abstract.

17.00-20.92s: End with the strongest dramatic action frames, including near-collision poses and a final hero-like composition where the fighters are still locked in conflict beneath the SEEDANCE 2.0 branding. The closing mood should be intense, dark, stylized, and unmistakably game-promo driven.

NEGATIVE PROMPT: realistic soldiers, bright daylight battlefield, fantasy elves, historical warriors, indoor living room, removed UI overlay, no title text, cartoon cel shading, comedic fight, blood gore close-up, crowd spectators, colorful sci-fi lasers, sports arena, soft pastel palette, modern firearms, casual clothing, photoreal actor faces.

SHOT PROMPTS:
1. SEEDANCE 2.0 promo layout with top thumbnail strip and lower main battle panel featuring two masked ninja fighters.
2. Dark smoky fighting-game confrontation between gray-blue and black-clad rivals in side-view action poses.
3. Alternating lunges, dodges, jumps, and clash silhouettes under a monochrome cinematic game trailer grade.
4. Final branded combat hero frame that still feels like a mod or feature showcase rather than a film scene.

SPEECH PACK:
- This reads like a vertical gameplay promo more than a standalone cinematic.
- The top strip and title keep reminding you that the fight is being showcased as a feature set.
- The main appeal is the dark ninja-versus-ninja choreography underneath the branding.
- It feels like Mortal Kombat-style action repackaged into a short social trailer.

GLOBAL LOCK: Vertical premium poster-style AI commercial video with a split composition. The upper two-thirds show a cinematic ruined-city transformation sequence: a muscular adult man in a torn gray t-shirt crawls through rubble, reaches a black neon-green energy can, and transforms into a towering black armored winged entity with glowing toxic-green energy. The lower third remains a fixed product-card layout containing a clean can packshot, a silhouette or character reference, dark infographic panels, and small technical-text style blocks. Keep the environment cold, smoky, and post-apocalyptic, while the energy effects stay vivid green and the product section remains crisp, graphic, and consistently anchored at the bottom of every frame.

[00:00-00:03] Start with the ruined city street in the upper section: fires burn in the distance, debris covers the road, and the man crawls toward the glowing energy can. The lower panel already displays the product-card layout with the can image, a humanoid silhouette reference, and dark UI-style text blocks.

[00:03-00:06] The man gets closer to the can and reaches toward it. Keep the lower third static and readable like a futuristic ad board while the upper action remains cinematic and dusty. The contrast between moving narrative above and fixed promotional panel below should stay clean.

[00:06-00:08] As he grabs the can, green light erupts from his body. The upper scene becomes intensely illuminated with toxic neon energy, while the lower panel continues showing the can, character reference, and compact product-spec graphic treatment.

[00:08-00:11] The transformation escalates into black armor with green energy lines and expanding wing structures. Maintain the poster-like composition: the upper section carries the spectacle, the lower section continues functioning as a branded concept sheet with high contrast and dark minimal layout.

[00:11-00:13] The transformed creature stands fully revealed in the upper frame, dominant over the rubble with wings spread wide. Keep the bottom panel unchanged, reinforcing the idea that this is an ad or creative concept board tied to the transformation.

[00:13-00:15] End with the winged figure lifting or surging upward into the sky as a bright green beam or glow cuts through the clouds. The lower product-card section remains locked in place until the final frame, preserving the hybrid between cinematic trailer and infographic-style commercial poster.

NEGATIVE PROMPT: full-screen cinematic without lower panel, bright cheerful colors, cartoon style, blurry product section, unreadable can, modern clean city, extra characters, blue energy, broken wings, anime proportions, watermark, chaotic typography, low-detail infographic, humor tone, misplaced layout elements