Creating believable AI video commercials requires solving two classic pain points: hardware hallucinations (where buttons and wheels warp randomly) and character drift (where the actor's face changes every cut).
In this tutorial, you will master the Zero-Gravity Commercial Workflow. By providing video diffusion models with orthographic multi-angle reference sheets, we lock every hardware contour and facial feature into place before directing dynamic camera motions.
1. How It Was Made: The 5-Step Pipeline
Rather than relying on one lucky generation, this commercial follows a structured 5-stage production process:
A tactile scroll wheel action disables gravity in sync. Everyday items, then room walls, then cars float into the sky.
Lock exact dimensions (128.2mm), MagSpeed wheel knurling, thumb wing, and haptic pad on a clean studio sheet.
Establish a photorealistic human anchor: 22yo graphic designer, wolf cut, natural skin texture, ARRI 50mm lighting.
Multi-panel front, profile, back, and macro tiles lock costume and facial morphology across dynamic camera turns.
Pass both sheets as dual references into Seedance 2.5 with locked lenses (16mm to 100mm macro) and foley sound rules.
2. Product Reference Sheet (Logitech MX Master 4)
The biggest challenge in commercial AI is hardware fidelity. Standard diffusion models often invent extra buttons or smooth out metal knurling. The fix is generating an orthographic multi-panel product sheet first.
- Exact Dimensions: Specify length (128.2 mm), width (88.4 mm), and height (50.8 mm) in prompt tokens to constrain aspect ratios.
- Neutral Background (#E4E5E7): Eliminates distracting shadows so the model isolates pure product geometry.
- Knurling & Materials: Explicitly mention solid brushed-silver knurled metal, matte graphite soft-touch body, and chrome ring accents.
Professional product reference sheet of the Logitech MX Master 4 wireless mouse in Graphite. Match the attached real product photos EXACTLY: shape, proportions, buttons, wheels, materials. Do not invent or simplify any part. Clean studio product photography on a seamless light grey background (#E4E5E7), soft diffused light, soft contact shadows, ultra-sharp detail, true-to-life materials. Wide 16:9 layout, two rows, thin white gutters, bold black uppercase sans-serif titles above every panel. PRODUCT: right-handed sculpted ergonomic mouse, 128.2 mm long, 88.4 mm wide, 50.8 mm tall. Matte graphite soft-touch body, slightly lighter satin grey main buttons with a thin light-grey front edge. One solid brushed-silver metal scroll wheel with uniform fine knurling across its whole width, sitting in a narrow slot at the front of the buttons. Behind the wheel: a small square mode-shift button and a tiny green LED. Small debossed "logi" on the left button. Left side: a wide wing-shaped thumb rest, a smooth haptic thumb pad with concentric ripple texture and a small ring of dots, a horizontal knurled metal thumb wheel with a polished chrome ring, two slim back/forward buttons below it. Front face: a USB-C port centered under the scroll wheel. Bottom: "MX Master 4" text, sensor, power switch, Easy-Switch 1-2-3 button, black glide feet. TOP ROW, five panels: 1. TOP VIEW: straight from above. Dimension lines "128.2 mm" and "88.4 mm". 2. FRONT VIEW: straight from the front at eye level: scroll wheel in the center, USB-C port below it, the thumb wing sticking out. Dimension line "50.8 mm". 3. SIDE VIEW (THUMB SIDE): pure left profile with thumb wheel, buttons and haptic pad. 4. BOTTOM VIEW. 5. THREE-QUARTER FRONT-LEFT: low camera at desk level, front and thumb side together. BOTTOM ROW, six tiles: - "MAGSPEED WHEEL": macro of the single solid knurled metal wheel, no split, no smooth middle band. - "THUMB WHEEL": macro of the knurled thumb wheel with chrome ring. - "HAPTIC PAD": macro of the ripple texture and ring of dots. - "FRONT & USB-C": macro of the front face with the USB-C port and wheel above it. - "IN HAND, FRONT-LEFT": a real right hand holding the mouse naturally, index finger on the scroll wheel, realistic hand size, five fingers. - "COLOR PALETTE": graphite, satin grey, light grey edge, brushed silver, chrome, matte black, LED green, knurled texture. Same product, identical design and proportions in every panel. No extra text, no watermark.
3. Character Casting: The Human Anchor
Before asking a video engine to animate someone, establish their facial features in high resolution. Avoid generic keywords like "gorgeous model" or "hyperrealistic", which result in plastic, unconvincing AI faces. Instead, specify natural imperfections, tangible wardrobe details, and specific camera lighting.
- Micro-Anchors: A small mole on the left cheek, small silver nose stud, messy wolf cut with wispy bangs, and thin silver index ring guarantee consistency.
- Texture Defense: Demand "visible pores, fine peach fuzz, no retouching, no smoothing" to banish waxy diffusion artifacts.
- Lens Heritage: Specify ARRI Alexa with a 50mm lens and gentle morning sidelight for cinema-grade contrast.
Candid photorealistic portrait of a young adult woman, a graphic designer, sitting at a wooden desk by a large window in a bright New York apartment in the morning. 22 years old. East Asian features, warm light skin with a few small moles, dark brown eyes, soft straight brows, jet-black hair with a messy wolf-cut and wispy bangs, small silver nose stud. Relaxed, confident expression, slight smile. Wearing an oversized charcoal hoodie with a white T-shirt hem showing, wide black cargo trousers. Natural beauty with minimal neutral makeup: bare glowing skin, a touch of mascara, natural brows, soft tinted lip balm. Real skin texture: visible pores, fine peach fuzz, subtle natural skin variations, no retouching, no smoothing. Outfit: casual but stylish, like a real young creative. Short nails, a thin silver ring on her right index finger, small silver earrings. No logos on clothing. Light: bright warm morning sun from the side through the window, hard sunlight on one half of her face with a bright highlight on the cheekbone, the other half in soft shadow with gentle teal-tinted fill, cinematic contrast but bright and pleasant. Shot on ARRI Alexa with a 50mm lens, shallow depth of field, fine film grain, natural colors. Ultra-realistic photograph, looks like a frame from a feature film, not a render, not an illustration, no plastic skin, no beauty filter.
4. Character Turnaround Sheet (360° Continuity)
In a dynamic commercial, the camera circles the actor from multiple directions (16mm low angle, 35mm overhead, eye-level profile). A single portrait causes the model to guess what the back of her head or her outfit looks like. A Turnaround Sheet provides full 360° spatial coverage.
- 5 Full-Body Angles: Front, Profile, Back, Close-Up, and At-The-Desk poses prevent wardrobe switches during fast edits.
- Macro Detail Swatches: Detail crops of her cheek mole, eyes under fringe, ringed hand, and sneakers give Seedance high-frequency visual anchors.
Professional character reference sheet of the young woman from the attached photo. Keep her exact face, facial structure, eyes, brows, nose, lips, skin tone, the small mole on her left cheek, hair and accessories. She is around 22 years old. Clean studio photography on a seamless light grey background (#D9D9D9), soft even diffused light, neutral white balance, ultra-sharp detail. Wide horizontal layout split into two rows separated by thin white gutters, bold black uppercase sans-serif titles above every panel. APPEARANCE (identical in every panel): natural matte-satin skin with visible pores, fine peach fuzz, subtle natural unevenness, only a very soft natural sheen on the nose and cheekbones, NOT oily, NOT glossy, no retouching. Almost no makeup, natural brows, bare natural lips. Dark brown-black hair in a loose messy high bun with a soft wispy uneven fringe over the forehead and a few loose strands framing the face, real individual strands and flyaways. Small silver stud in her nostril, silver pearl-cluster flower stud earrings. A thin polished silver band ring on her right index finger. Neat short rounded nails with a clean solid milky-nude polish. OUTFIT: plain oversized charcoal grey hoodie in soft brushed cotton with drawstrings, NO text, NO print, NO logos; a white T-shirt hem slightly visible at the bottom; relaxed black cargo trousers; white leather sneakers. TOP ROW, five panels: 1. FRONT: full body standing relaxed, arms down, facing camera. 2. SIDE: full body in pure left profile. 3. BACK: full body from behind, showing the bun and the hoodie hood. 4. CLOSE-UP: head and shoulders portrait, calm neutral expression with a faint half-smile, looking into the lens. 5. AT THE DESK: medium shot sitting at a light wooden desk, three-quarter angle, her right hand resting naturally on a dark graphite computer mouse. BOTTOM ROW, six labeled detail tiles: - "SKIN TEXTURE": macro of her cheek with pores, peach fuzz and the small mole. - "EYES": macro of her dark brown eyes and natural lashes under the wispy fringe. - "HAIR": macro of the messy bun and the fringe, showing individual strands. - "HAND & RING": macro of her right hand with neat milky-nude nails and the thin silver ring on the index finger. - "SNEAKERS": macro of the white leather sneakers, showing the leather texture and stitching. - "COLOR PALETTE": 8 swatches in a grid: charcoal grey hoodie, white T-shirt, black cargo, white leather, silver, dark brown-black hair, warm light skin tone, dark brown eyes. Same person, same face and proportions in every panel, realistic human anatomy, five fingers on each hand, true-to-life colors, looks like real photographs from a casting session, not a render, no text except the panel titles, no watermark.
5. 8-Shot Video Choreography (Seedance 2.5)
Now assemble both references into Seedance 2.5. Rather than giving a vague description, break the 15-second commercial into 8 precise director cuts with locked lens values, shutter rules, and physics triggers.
Shot 1: The Discovery
Low desk angle in a sunny Manhattan loft. As she scrolls lazily, her ceramic coffee mug lifts into mid-air. She smiles in playful surprise and whispers: "Wait... what?"
Shot 2: Tactile Focus
Extreme close-up from the front at desk level. Her index finger rolls smoothly across the knurled metal MagSpeed wheel, turning in sync.
Shot 3: Gravity Switches Off (The Hook)
She scrolls deliberately. Every object in the loft peels off the floor and floats up—plants, vinyl records, lamps, cushions. The walls and giant window lift off whole, revealing a sunny NYC crosswalk around her desk.
Shot 4: The Crosswalk From Above
Overhead view from a 4th-floor window looking down. Her desk stands centered in the zebra crosswalk as pedestrian traffic steps around her in shock.
Shot 5: Diving Into The Free Spin
Camera flies at high speed toward the mouse, yaws 15° to the right, and dives directly into the center of the scroll wheel as she flicks it into infinite free-spin mode.
Shot 6: The Entire Street Lifts Off
Looking up from the asphalt. Solid slabs of street lift into the sky along with real yellow NYC taxis and black sedans, turning slowly with heavy physical weight.
Shot 7: Center of the Storm
Frontal push-in. She sits laughing at her desk in the street center as shadows of soaring cars sweep across her face, wind lifting her hair upward.
Shot 8: The Haptic Stop
Rapid crash zoom onto the spinning metal wheel. Her index finger drops straight down onto the knurling, halting the wheel instantly with a tactile haptic thud.
GLOBAL STYLE 15-second 9:16 commercial for the Logitech MX Master 4 mouse. Live-action look shot on 35mm film (Kodak Vision3), vintage spherical lenses, 180-degree shutter, natural motion blur. Real materials, real cars, real weight. NOT CGI, NOT plastic, NO beauty retouch. Energetic young camera: handheld, fast push-ins, crash zooms, varied angles and lenses. The horizon stays level: the camera never rolls or flips. Hard cuts. No on-screen text. NO MUSIC, only sound effects. REFERENCES The FIRST reference image is the girl: exact face, hair and clothes in every shot: petite, about 22, wispy fringe, messy bun, charcoal hoodie, nose stud, pearl earrings, silver ring on right index finger, nude nails. The SECOND reference image is the mouse: Logitech MX Master 4 in graphite, exact shape: low, wide sculpted body, one solid brushed-metal knurled scroll wheel between the two buttons, a USB-C port on the front face under the wheel, metal thumb wheel on the left side, haptic thumb pad, green LED. Same shape and size in every shot. RULES - Real scrolling: only her index finger moves. The fingertip rolls over the metal wheel and the wheel turns WITH the finger. The wheel never turns under a still finger. - The world reacts to her scrolling like gravity switching off: every object lifts off separately and flies up on its own, tumbling at its own speed. Objects stay whole, no cracking, no shattering. - Her desk always holds a slim black monitor, a white keyboard, a ceramic coffee mug, a dark felt desk mat and the mouse. She and her desk never float. - Light and wind on her always match her surroundings. SHOT 1 (0.0-2.5s) - She notices - Lens: 16mm, very low camera at the desk surface at the near left corner, looking diagonally across the desk at her. The mouse under her hand is the main object in the lower left foreground, sharp. Her face is the second focus. The keyboard is in the middle; the coffee mug stands FURTHER AWAY, next to the monitor on the far side of the desk, small in frame, never in the foreground. Handheld, slow push-in toward the mouse. - Location: a cozy high-floor Manhattan loft in the morning. Behind and beside her a HUGE floor-to-ceiling window that fills most of the background, with a wide view of Manhattan: rooftops, water towers, the Empire State Building. Warm sun through the window, hard light on one side of her face. - Action: she scrolls lazily, bored, eyes on the monitor. With her scroll, the coffee mug on the far side of the desk lifts a few centimeters into the air and hangs there, coffee level. Her eyes move to it. She scrolls once more, carefully: the mug rises higher, in sync. A half-smile of amused surprise spreads across her face, one corner of her mouth lifting in a playful smirk, and she says softly, with a little disbelieving laugh: "Wait... what?" Still smirking, she looks down at the mouse. SHOT 2 (2.5-4.0s) - One slow scroll, from the front - Lens: 100mm macro. The camera looks at the mouse from the FRONT, low at desk level: the front face with the USB-C port, the metal wheel in the center above it, her fingertip on top of the wheel. Slow push-in. - Action: ONE slow, long, continuous scroll: her fingertip starts on the front top of the wheel and slowly rolls back over it, away from the camera toward her palm, the knurled metal turning together with the fingertip. No button clicks. SHOT 3 (4.0-6.5s) - Gravity switches off (HOOK) - Lens: 16mm, same low angle as shot 1. Handheld, energetic push-in, then a quick tilt up following the flying objects and back down to her. - Action: she scrolls on purpose, grinning. Gravity in the room turns off. Object by object, everything around her lifts off the floor and walls and flies upward, each piece separately, tumbling and spinning at its own speed with real weight: the plants and their pots, soil floating out of them, the floor lamp, books opening mid-air, vinyl records sliding out of their sleeves, the record player, framed prints, cushions, the rug peeling off the floor, a chair. Then the walls and the big window, frame and glass intact, lift off and fly upward in large whole pieces, never dissolving or fading. - Revealed behind: a real sunny Manhattan street around her: a zebra crosswalk under her desk, dark asphalt, brick buildings with fire escapes, autumn trees with yellow leaves, glass towers, clear blue sky. - As the window flies up and away with the wall, the warm window light on her is replaced by crisp autumn daylight from the side, a breeze moves her fringe. She looks around, amazed, laughs. SHOT 4 (6.5-8.0s) - The crosswalk from above - Lens: 35mm, high angle from a fourth-floor window, looking steeply down, slow drift. - A straight street with a zebra crosswalk across it from sidewalk to sidewalk, its white stripes parallel to the street. Her desk stands in the middle of the crosswalk, she sits scrolling, long shadow over the stripes. Pedestrians cross the street on the crosswalk from one sidewalk to the other, stepping over the stripes. A man with a coffee steps around her desk and looks back, confused. A yellow taxi waits before the crosswalk. SHOT 5 (8.0-10.5s) - Diving into the free spin - Starts: a straight frontal medium shot of her hand on the mouse on the desk mat, camera level, the sunny street blurred behind. - Camera move: the camera flies fast toward the mouse. While flying, it yaws about 15 degrees to the RIGHT, so it arrives at an angle and dives straight into the CENTER of the scroll wheel, ending in an extreme macro where the knurled metal fills the frame. One continuous, smooth, fast move. The horizon stays level, no roll. - Action: as the camera flies in, her index finger makes ONE strong, fast stroke and lifts completely off the wheel. The wheel spins freely on its own, faster and faster, the knurling blurring, sunlight flashing on the metal edges. SHOT 6 (10.5-12.0s) - The street lifts off - Lens: 18mm, very low angle from the asphalt, looking up at a red-brick building with fire escapes, an autumn tree and a glass tower against the blue sky. Handheld, slight push-in. - Action: the whole street lifts. The asphalt itself rises in big whole slabs together with the cars on it. Two real yellow NYC taxis and a black sedan lift off and rise diagonally upward past the building, turning slowly with real weight, their wheels hanging. Clouds of street dust, newspapers and autumn leaves fly up around them. The brick buildings grow taller, stretching upward floor by floor. People stop and look up. SHOT 7 (12.0-13.5s) - She in the center - Lens: 24mm, eye level, frontal, she sits in the center at her desk on the street, orange traffic cones around. Fast push-in toward her. - Action: on both sides and behind her, taxis and cars lift off the asphalt and float up, leaves and newspapers rising. Wind lifts her fringe and loose strands upward, shadows of rising cars sweep over her face. She looks up, delighted, laughing, her hand on the mouse. SHOT 8 (13.5-15s) - The stop - Lens: 100mm macro, low three-quarter angle on the wheel. A fast crash zoom in, then the camera freezes completely. - Action: the metal wheel is spinning fast by itself. Her index finger hovers above it, then lowers straight down onto it. The wheel stops dead under the fingertip, the mouse gives a tiny haptic jolt. She only presses and holds, no scrolling. LIGHTING AND GRADE (very important) A graded feature-film look, like a scanned film negative, NOT a phone video: no HDR, no over-sharpening, no clipped highlights anywhere, no blown-out sky or windows. Highlights roll off softly, blacks slightly lifted and milky, gentle halation around bright edges, visible film grain, soft lens with a light vignette. Colors look painted and even: rich but controlled, smooth tones across surfaces, not glossy. Shots 1-2: warm morning sun through the big window, the window view stays visible, not white. Shots 3-8: clear autumn day, sun low from the side, long soft-edged shadows, deep teal-blue sky, warm saturated yellow taxis and golden trees, brown brick, grey asphalt with warm sunlit patches. NOT orange, NOT golden hour glow. SOUND (no music) Quiet room tone. A soft rolling ratchet sound of the wheel whenever she scrolls, no button clicks. In shot 1 she says softly, amused and surprised, with a smirk in her voice and a tiny breathy laugh: "Wait... what?" (only these two words). Whooshes of objects flying up. City ambience, footsteps, wind. A sharp flick and a rising metallic whir on the free spin. Deep rumble of the lifting street. At the stop: total silence and one deep haptic "thunk". Heavy thuds as the cars land. A soft rolling wheel sound on the packshot, then quiet room tone and silence.
6. Foley Sound & Film Emulation
Two subtle choices give this commercial its grounded, premium feel:
- Foley Over Music: Banning background music makes the video hit harder. The soundscape relies on mechanical tactile cues: the rolling ratchet of the wheel, the rising metallic whir of free spin, and the heavy physical thud of landing cars.
- 35mm Film Grade (Kodak Vision3): Soft highlight roll-off, lifted milky blacks, and gentle halation eliminate the tell-tale hyper-glossy look of standard AI renders.