PDMD T2VA：8条小样本对比

• 两人先越障后跳过巷隙，人数不变
• 落地翻滚与鸽群飞散衔接清楚
• 喊声、罐响和落地声对应可见动作

integrated_multimodal_description: [Shot 1] Live-action with anamorphic flares and wet nighttime textures. A handheld tracking shot follows two thieves across a rain-slick rooftop, their satchels of stolen bearer bonds bouncing as they run. The wiry East Asian woman in a black windbreaker (S1), speaking in a sharp, breathless, quick-paced voice, clears an air duct and shouts: <d>[English] Don't slow down — jump!</d> Beside her, a stocky, bearded Black man in his forties hurdles a vent pipe; his leg clips stacked paint cans, which tumble and clatter behind the pair. [Shot 2] At 00:03.000, the camera cuts to a low angle at the roof edge. It pushes in with small amplitude at fast speed as the same two thieves leap across the alley, their coats spreading against sodium-lit fog. They reach the far rooftop, land heavily and roll across its gravel, sending pigeons flapping away from their landing path.

overall_soundscape: Rain patters beneath the thieves' rapid footsteps and ragged breathing, with police sirens sounding in the distance. The struck paint cans clatter behind them; two grunts accompany the landings, followed by scraping gravel and a burst of pigeon wingbeats.

non_diegetic_music: A fast staccato string ostinato plays over taiko-style drums. Both rise in volume toward a single accented hit as the thieves land.

• 传球先于起跳扣球
• 切镜前后球与球网空间关系保持
• 击球声只在手球接触时出现

integrated_multimodal_description: [Shot 1] 2D anime opening-title imagery with strong graphic poses, backlighting flashes and directional motion blur. On a sunset-lit volleyball court, six players in matching jerseys hurry into position across polished wooden flooring. The setter jumps and taps the ball upward against an orange sky rendered in bold flat gradients. The camera pedestals up with large amplitude at fast speed along the ball's rising path as the tallest spiker springs vertically with his hitting arm drawn back. [Shot 2] At 00:02.600, the camera cuts to a side view at net height. It pans left quickly as the spiker's open palm meets the descending ball and sends it down beyond two blocking silhouettes, with radiating impact lines and sunset flare marking the strike. An unseen vocalist (S1), an energetic bright tenor, sings: <d>[English] One more point, we're never backing down!</d> All six players' lips stay closed during the play and lyric.

overall_soundscape: Sneakers squeak as the players run and pivot on the wood. Separate ball contacts make short thuds, followed by feet landing and the net snapping taut.

non_diegetic_music: N/A

• 拉杆释放后泡沫弹才飞出
• 弹击倒塔楼不是塔楼提前消失
• 两段广告词画外完成并保留原品牌字卡

integrated_multimodal_description: [Shot 1] An oversaturated 1980s toy commercial with mild videotape wobble. In a backyard sandbox, a freckled boy kneels beside a gray plastic fortress. He presses and releases a spring lever, launching one soft foam shell into a plastic sentry tower. The tower falls into the sand as a medium, moderate-speed zoom isolates the impact. An unseen announcer (S1), in a booming, carefully articulated baritone, says in an off-screen voiceover: <d>[English] Iron Ridge Fortress — under siege and never surrenders!</d> The boy's lips remain closed as he grins and raises a fist. [Shot 2] At 00:06.000, the camera cuts to the boxed fortress turning slowly on a studio turntable against a starburst. Yellow lettering reads "IRON RIDGE FORTRESS — $14.99!" The same announcer (S1) says in an off-screen voiceover: <d>[English] Batteries not included, glory is.</d> The announcer has no on-screen counterpart; no visible person mouths his words. The product shot contains no human figure.

overall_soundscape: The spring clicks and releases before the soft projectile impact and plastic pieces tapping the sand. Faint studio room tone alone accompanies the turntable's product view.

non_diegetic_music: A brisk marching snare and short brass fanfares accompany the demonstration, with a cymbal accent on the tower's fall.

• 人物擦泪与拍手按先后完成
• 手机或电脑位置稳定且茶杯不被穿透
• 机位固定仅轻微光学变焦，不又向前移动

integrated_multimodal_description: [Shot 1] Warm, grainy webcam imagery in a small kitchen. An elderly Vietnamese woman with silver hair sits at a laptop beside a rice cooker, its turned-away screen reflecting light in her glasses without revealing readable content. She covers her mouth with both hands, gives a soft sob and laughs through it. Lowering one hand, she lifts her floral apron to dab her eyes, then lets the fabric fall back before clapping exactly twice. She leans toward the laptop with a broad tearful smile. A teacup remains untouched beside her. The camera never changes position; a very slight, slow optical zoom tightens the view on her face without moving the viewpoint forward. She speaks no words.

overall_soundscape: Soft sobbing breaths turn into a breathy laugh and two separate handclaps. Apron fabric rustles against the chair over a steady laptop-fan hum.

non_diegetic_music: N/A

• 线头确实从同一针眼穿过，不能溶入针杆
• 拉动后针眼两侧仍是同一连续线而非凭空分叉
• 固定微距下手、针、线的遮挡和焦平面稳定

integrated_multimodal_description: [Shot 1] A photoreal craft macro at a sunlit sewing table. Adult left fingers hold one rigid steel needle so its eye stays clearly exposed. The right hand brings a neatly cut blue thread tip level with the opening, pauses to align it and pushes it through the eye rather than against the shaft. The right thumb and index finger then catch the end protruding on the far side and pull a short length through. One continuous thread now hangs on both sides of the same eye. The locked camera keeps the eye and fingertips within a shallow but readable focus plane throughout; the needle neither bends nor changes shape. No speech or visible text.

overall_soundscape: Close fingertip friction and a faint brushing thread are audible against quiet room tone. There is no amplified metal impact.

non_diegetic_music: N/A

• 相邻两滴接触合并为一滴而不是凭空出现新滴
• 合并滴受重力沿玻璃向下，不穿出玻璃
• 滑动留下连贯清晰轨迹，背景散景不跟着移动

integrated_multimodal_description: [Shot 1] Photoreal extreme macro of a vertical window covered in condensation. A garden remains defocused into green circles beyond the glass under soft overcast daylight. Two neighboring droplets enlarge until their curved edges meet, then unite into one heavier drop with a continuous reflective highlight. The merged drop slowly slides downward along the pane, leaving a narrow clearer track above it. Smaller droplets on either side remain where they were. The camera and focus plane stay locked to the glass, so neither the stationary droplets nor the garden bokeh follows the falling water.

overall_soundscape: Outdoor rain is faint and muffled through the closed window. The tiny moving drop has no exaggerated close-up sound.

non_diegetic_music: N/A

• 相啮合齿轮接触处运动方向正确，不互穿
• 齿条与平台保持刚性连接并沿导轨上升
• 停止摇柄后平台停止，不能继续无动力漂浮

integrated_multimodal_description: [Shot 1] A clear, fixed three-quarter view of a wooden tabletop rack-and-pinion mechanism. A crank turns a small gear whose teeth engage directly with a vertical toothed rack. The rack is rigidly attached to a tiny platform guided by two straight rails. One hand slowly turns the crank exactly twice. Successive gear teeth push the rack upward, raising rack and platform together through a short, smooth travel while the rails stay still. At the end the hand holds the crank in place, and the linked platform stops at the same time. The entire linkage stays visible without labels, diagrams or cuts.

overall_soundscape: Light wooden clicks repeat as teeth engage, under faint guide-rail creaks. The mechanism falls quiet when the hand stops the crank.

non_diegetic_music: N/A

• 调配重前先抓停摆杆，调整后才重新摆动
• 配重向下后节拍更快而非更慢
• 每次摆到交替侧的机械释放点有一声click，无多余背景节奏

integrated_multimodal_description: [Shot 1] A locked frontal product close-up of a wooden mechanical metronome on a piano. Its pendulum begins with slow, even swings. One hand stops the rod at the center and holds it still while sliding the small metal weight downward along the rod. The hand then gives a slight starting push and releases. The pendulum resumes at a visibly faster regular rate. Each click belongs to an alternating mechanical release point, with complete silence from the mechanism during the held adjustment. The scale remains too small to read, and the camera does not move.

overall_soundscape: Dry ticks establish the slower rhythm, cease while the rod is held and return at a faster even interval. A faint finger-on-metal rub occurs during the weight adjustment.

non_diegetic_music: N/A