Family Sitcom · MiniMax H3

MiniMax H3Short DramaComedyVerticalLip SyncMacroFilm GrainGame Footage15s
A 15-second family sitcom with the male and female leads locked in via two reference images, featuring quick dialogue cuts with subtitles. Dialogue-driven short dramas demand the most in lip-sync and expression continuity, making this a test of MiniMax H3's limits.
PROMPT · 视频提示词
Create with AI
Duration: 15 seconds Aspect ratio: 9:16 vertical Genre: ultra-realistic family sitcom, live-action short drama, fast-paced dialogue editing Characters: two leads Reference Image 1: fixed as the male lead Reference Image 2: fixed as the female lead The characters' face shape, features, hairstyle, body type, clothing and sense of age must strictly follow their respective reference images throughout; do not redesign the costumes, do not swap character identities. Subtitles: no dialogue subtitles, lower-third captions, titles, or stickers of any kind. Language: natural English dialogue, accurate lip-sync. Overall Style: A highly realistic live-action family short drama, set in an ordinary home where the living room connects to the dining room, with a touch of old-school home-video texture. Warm yellow tungsten lighting, natural skin texture, slight film grain, realistic indoor shadows — colors are warm but not overly polished. Shots rely mainly on fixed camera positions, fast shot/reverse-shot cutting, and partial close-ups; the pacing is tight, and the performances carry a restrained, deadpan comedic absurdity — avoid exaggerated, cartoonish acting. Scene Continuity: The male lead sits at a wooden computer desk, on which sits a bulky old-school CRT monitor, a beige mechanical keyboard, and a mouse. The female lead remains standing behind him or just off to the side-rear of the desk throughout. In the distant dining room, a few out-of-focus family members are waiting to eat; the dining table holds an ordinary dinner and place settings. Background characters make only slight, natural movements and never steal focus. The following key props must remain locked/consistent: 1. The old-school CRT monitor must remain firmly on the computer desk from start to finish — it must never fall off, fly up, or cover the male lead. 2. What the female lead holds is a lightweight hardcover AI reference booklet — not a monitor, blueprint paper, a photo frame, or an ordinary folder. 3. Whenever the booklet's cover faces the camera, the bold headline text "AI PROMPTS" must be clearly visible, with the smaller line "ACTORS AND THE FUTURE OF FILM" beneath it, accompanied by three vertically arranged film concept images. 4. The female lead uses this same "AI PROMPTS" booklet throughout: she first bops the male lead on the head with it from behind, and at the end drops it down to cover his face. SHOT 1 | Family Argument Escalates (00:00-07:60) Subject: The male lead sits at the computer desk frantically typing AI prompts; the female lead stands behind him, urging him to come to dinner. Action: 00:00-00:40: A fixed medium shot establishes the scene. The male lead leans forward, both hands rapidly typing on the beige keyboard, eyes locked on the CRT monitor, completely absorbed in the computer. The female lead stands behind him with arms crossed, watching him while suppressing her anger. In the distance, family members are already seated at the dining table, waiting. 00:40-01:20: The female lead leans slightly toward the male lead, brows furrowed, asking in a questioning, faintly irritated tone: Female lead: "AI again?" The male lead doesn't turn around, fingers still typing, answering in a matter-of-fact tone: Male lead: "I'm prompting." 01:20-02:50: Quick cut to a close-up of the female lead's shoulder. She tilts her head slightly, trying to stay patient, gesturing toward the dining table: Female lead: "Come on, it's dinner time." Cut to a close-up of the male lead. He glances back at her only briefly, his expression showing intense irritation at being interrupted, then quickly turns back to the computer: Male lead: "I'm prompting!" 02:50-04:60: Cut to a frontal close-up of the female lead. Her eyes widen as she stares at him in disbelief, her voice gradually rising: Female lead: "You've been sitting there for twelve hours straight." After saying this, she shakes her head slightly, lips pressed tight, her anger beginning to build. 04:60-07:60: Rapid cross-cutting between the male lead, the female lead, and the keyboard. The male lead grows increasingly hyped, his back tensing, his body leaning even closer to the CRT monitor, hands pounding the keyboard, shouting three times in a row, each time faster and more excited: Male lead: "I'm prompting!" Male lead: "I'm prompting!" Male lead: "I'm prompting!" Each shout cuts to a closer shot: the first is a half-body shot of the male lead, the second a close-up of his face, and the third an extreme close-up of his fingers typing at high speed. The female lead shifts from crossed arms to resting one hand on her forehead, closing her eyes and slowly inhaling. When she opens her eyes again, she is no longer arguing — she simply fixes him with a cold, icy stare. Performance: The male lead isn't simply angry — he's utterly obsessed with AI generation the way one gets addicted to a video game, his speech growing faster and faster; the female lead's emotional shift must be clearly shown as "restrained displeasure → disbelief → complete loss of patience." Camera: Fixed-position, fast shot/reverse-shot cutting; both the male and female leads are shot on a realistic 40–50mm portrait focal length. As the emotion escalates, the shots gradually push in from medium to close-up. No camera shake, no sudden changes to the room's layout. Audio: Dense keyboard clacking, faint distant sounds of cutlery and family chatter. No background music, no subtitles. SHOT 2 | Showing Off the So-Called "Super Ultra Real" AI Result (07:60-11:70) Subject: The keyboard, the AI-generated footage on the CRT monitor, the excited male lead, and the indifferent female lead. Action: 07:60-08:30: Extreme close-up on the beige keyboard. The male lead's hands type at high speed, the old-style keycaps continuously pressing down and springing back, the force of each keystroke growing stronger. Finger movement must look realistic, with a normal five-finger structure — no extra fingers. 08:30-10:30: Cut to a frontal close-up of the CRT monitor. On screen is a crude, early-generation AI video-generation interface: a grayish-white window, a prompt input area, a generation-progress area, and a centered video preview window. The preview window shows a somewhat rough AI-generated video clip: a long-haired woman wearing light-colored, slightly disheveled clothing with a bit of dust on it, running straight toward the camera in front of a manor building with tall windows and a wide entrance; her expression is tense, her arms bent, fists clenched, her hair swaying slightly as she runs. The footage has a cool-toned, low-saturation, slightly desaturated cinematic color grade, and the character's movement has a bit of the stiff, unnatural quality typical of early AI-generated video. The computer preview screen must show "a long-haired woman running straight toward the camera in front of a manor" — not three book-cover images, not an abstract pattern, not a page of text, and not modern game footage. 10:30-11:20: Cut back to a two-shot medium. The male lead points excitedly at the CRT monitor, eyes shining, eyebrows raised, showing off as if he'd just completed a major masterpiece: Male lead: "Look! Very super ultra real!" As he speaks, he points at the screen first, then slaps the keyboard once, a proud smile on his lips, completely oblivious to the female lead behind him. 11:20-11:70: The focus of the shot slowly shifts from the male lead to the female lead. She stares at him with a blank expression, her eyes cold, then reaches over to pick up a large, lightweight hardcover booklet from the side of the desk. The booklet's cover must show: a bold, eye-catching headline "AI PROMPTS" at the top; the smaller line "ACTORS AND THE FUTURE OF FILM" beneath it; three film concept images arranged on the cover; overall, it should look like a professional hardcover book about AI prompts and filmmaking. After picking up the booklet, the female lead slowly draws it back behind her shoulder, preparing to swing it. The male lead remains completely absorbed in the computer, oblivious. Camera: Macro close-up of the keyboard → frontal close-up of the CRT → two-shot medium → a subtle focus shift onto the female lead and the AI booklet. Do not use flashy camera moves — build the comedic beat through editing rhythm instead. Audio: The keyboard sound cuts out abruptly for half a second as the male lead speaks excitedly; a faint sound of cardboard friction as the female lead picks up the booklet. No score, no subtitles. SHOT 3 | The Booklet Payoff Gag (11:70-15:00) Subject: The female lead, the male lead, the same "AI PROMPTS" hardcover booklet, and the CRT monitor, which remains on the desk throughout. Action: 11:70-12:25: The two-shot medium continues. The female lead, standing behind and to the right of the male lead, draws the "AI PROMPTS" booklet back behind her right shoulder to build momentum, then swings it quickly along a clear downward arc toward the back of his head. This is an exaggerated but harmless live-action sitcom gag: the booklet is a lightweight, hollow prop; the first hit lands on the back of the male lead's head and upper neck; the moment of contact is obscured by the booklet, his head, and motion blur together — no injury detail is shown. A short, crisp, slightly exaggerated "smack" sounds at the same instant. The female lead must not hit the desk; the booklet's trajectory must clearly aim at the back of the male lead's head — not the keyboard or the desktop. 12:25-12:85: Startled, the male lead's head jerks forward and down, his upper body lunging onto the keyboard, his chest briefly resting against the edge of the desk. His body then goes limp, his hips lift off the chair, and he slides off the side of the chair down under the desk, ultimately falling out of frame. The chair slides back slightly but must not tip over. The motion is a continuous, comedic slump — not the character suddenly vanishing, teleporting, or being knocked flying. The CRT monitor remains firmly in its original spot on the desk throughout, still showing the AI footage of the woman running in front of the manor. 12:85-14:20: The male lead has already fallen out of frame. The female lead stands by the desk, still holding the same "AI PROMPTS" booklet in her right hand, her body leaning slightly forward as she looks down at him lying on the floor. She first holds the booklet up for him to see, cover facing the camera so "AI PROMPTS" is clearly legible, then says in a calm, sarcastic tone still tinged with lingering anger: Female lead: "Use this blueprint, you stubborn fool!" After saying this, she releases the booklet from her hand toward the floor where he lies. Do not fully show it landing within this medium shot — use a hard cut to deliver the closing punchline instead. 14:20-14:35: Hard cut suddenly to a slightly high-angle close-up of the floor. The male lead lies flat on his back on an intricately patterned rug, his face fully exposed at first. His eyes are closed, mouth slightly open, both arms limp at his sides, palms up, legs naturally straightened, presenting an exaggerated, "utterly worn out" comedic state. At this instant, nothing — no monitor, no book, no other object — may be on the male lead's face. 14:35-14:55: After a pause of about 0.1 seconds, the same "AI PROMPTS" hardcover booklet drops in fast from the upper left of the frame. The booklet's cover faces the camera throughout, falling in a short diagonal path from upper-left toward the center of the frame. It lands gently on the male lead's face, fully covering it, producing a second, softer "smack." The complete three stages must be shown: the male lead's bare face lying on the floor → the booklet entering the frame from the upper left → the booklet landing to cover his face. This action must not be skipped via instant morphing or appearing out of nowhere. 14:55-15:00: The booklet bounces once slightly on the male lead's face, then settles still. The final frame holds on the booklet's bold "AI PROMPTS" text, the three film concept images, and the male lead's limp, splayed-out arms. The female lead does not appear in this final floor close-up. The CRT monitor also does not appear in the floor shot, since it remains on the computer desk throughout. Performance: Before the first booklet strike, the female lead should not shout or yell — instead she keeps a calm, almost unsettlingly composed expression; the male lead's fall must show three clear, continuous stages: head jerking forward, body lunging onto the desk, and sliding off the chair. The final booklet-on-face beat is a lightweight-prop comedic effect and must not depict any real injury. Camera: The first action uses a fixed medium shot that does not follow the male lead's fall; the ending uses a slightly high-angle, roughly 50mm floor close-up. At the 14.20-second mark, a clean, crisp hard cut must be used — no dissolve, morph transition, or camera move-through. Audio: The first crisp booklet-strike sound; a faint sound of the chair shifting; a dull thud as the male lead lands on the rug; the female lead's English dialogue; the second, softer sound of the booklet landing on his face at the end; finally, about 0.3 seconds of quiet room ambience. No background music, no subtitles. --- NEGATIVE / Negative Constraints: Do not let the CRT monitor fall, move, fly up, tip over, or cover the male lead; Do not let the computer smash into the male lead; Do not turn the AI booklet into a CRT monitor, tablet, photo frame, blueprint paper, ordinary folder, or any other object; Do not render the first action as hitting the desk; Do not let the booklet strike the keyboard or the desktop; Do not omit the male lead's continuous "lunging onto the keyboard → sliding off the chair → falling out of frame" motion; Do not let the male lead teleport to the floor; Do not cover the male lead with the booklet at the very start of the final shot — his bare face must be shown first, then the booklet falls in from the upper left to cover it; Do not mistakenly render the booklet cover's three images onto the CRT screen; The CRT screen must only show the AI-generated video of the long-haired woman running straight at the camera in front of the manor; Apart from the computer interface and the booklet cover, no text of any kind may appear; No dialogue subtitles, lower-third captions, titles, stickers, watermarks, or platform logos; No blood, wounds, swelling, fractures, or realistic injury effects; No horror-movie atmosphere, no brutal violence; No cartoon rendering, anime style, plastic-looking CG skin, or excessive skin smoothing; No face-swapping, identity-swapping, age changes, costume drift, or hairstyle changes; No extra characters suddenly appearing; No facial melting, misaligned mouths, double mouths, or out-of-sync lip shapes; No extra arms, extra fingers, fused palms, or the booklet clipping through hands; No props floating, clipping through geometry, teleporting, or randomly changing size; Background characters must not be completely static, nor should they steal focus; No camera-position drift, changes to the room's layout, or sudden jumps in the position of the desk, chair, or CRT monitor.
✍️ Editor’s Notes

这条的核心难点不是动作而是道具不能跑偏,所以专门列了一节必须锁定关键道具,逐条写死CRT显示器不能掉、资料册封面必须清晰印着「AI PROMPTS」——情景喜剧的笑点全靠这本书最后精准砸脸,模型一旦把它换成别的东西梗就废了。三个SHOT按秒拆到00:40这种精度,台词也一句句写死,这种极端精细化在国产短剧类prompt里很常见,本质是不信任模型的即兴发挥,宁可自己把每一拍都嚼碎了喂进去。

Create with AI
Related Works
×