(Dasiwa_v2_singularity)MiniMax_H3双采参考生视频

Basic Model : MiniMax H3|Published : Sep 21, 2026 09:56

Workflow

ssy

image

prompt

integrated_multimodal_description:

A 15.00-second cinematic painterly psychological drama with six clearly separated shots and deliberate editorial cuts. The sequence must NOT play as one continuous camera take. Use explicit hard cuts between Shots 1–5, followed by one slow pullback only in the final shot.

VISUAL STYLE:

The visual language is inspired by Edward Hopper's American realist oil paintings, especially his treatment of isolated figures inside restaurants, cafés, hotel interiors and nocturnal urban architecture.

Do NOT interpret this style as a dark yellow-green cinematic filter.

Use clearly separated, moderately desaturated oil-paint color masses with restrained visible brush texture and opaque matte surfaces. Colors should feel mixed from physical pigments rather than digitally saturated.

The restaurant interior is clearly visible and adequately exposed.

Interior walls and furniture use muted warm cream, pale beige, faded ochre, subdued terracotta, dusty brown and restrained warm ivory. Human skin retains natural warm peach and muted rose notes under the restaurant lamps.

Exterior night colors are clean slate blue, muted ultramarine, cool gray-blue and soft blue-black rather than teal or green.

Avoid green color casts entirely.

Artificial restaurant light is warm ivory to soft amber, NOT saturated yellow, orange or chartreuse. Moonless city night outside is cool blue-gray, NOT cyan or neon teal.

Maintain medium overall brightness and readable shadow detail. The restaurant should feel quiet and nocturnal, not underexposed. Faces, hands, the photograph, plates, empty chair and architectural geometry must remain clearly readable.

Use Hopper-like compositional principles: strong horizontal and vertical architectural geometry, simplified perspective, large calm planes of wall and glass, carefully controlled negative space, isolated human figures, psychological distance between characters, clean silhouettes and hard-edged areas of artificial light.

Lighting forms broad geometric shapes rather than dramatic modern volumetric beams.

The image should resemble a restrained American realist oil painting brought subtly into motion: simplified but believable anatomy, slightly flattened painterly surfaces, confident shape design, visible but not exaggerated brushwork, minimal glossy reflection and no photographic microtexture.

No noir-black crushed shadows. No horror lighting. No sickly yellow-green palette. No cyan-orange blockbuster grading. No neon colors. No high saturation. No heavy green tint. No sepia filter. No excessive vignette.

SETTING AND CONTINUITY:

Deep night inside an almost empty old-fashioned American corner restaurant.

A large plate-glass window occupies one side of the restaurant. Beyond it lies a silent blue-gray city street with simple dark building façades.

A small dining table stands beside the window.

An adult man and adult woman sit opposite each other.

A third wooden chair stands beside their table and remains empty for the entire sequence.

The man has {{MAN_APPEARANCE}} and wears {{MAN_CLOTHING}}.

The woman has {{WOMAN_APPEARANCE}} and wears {{WOMAN_CLOTHING}} with an accessible coat pocket.

The man has a white ceramic plate with steak, a fork and a dinner knife.

The woman's food remains untouched.

The woman carries one small black-and-white photograph inside her coat pocket. The photograph contains a human figure, but the person's face must remain unreadable throughout.

Preserve identical character identity, clothing, seating positions, table geometry, window position, photograph, tableware and empty-chair position across every cut.

The woman remains on the same screen side and the man remains on the opposite screen side. Respect the 180-degree rule and preserve correct eyelines throughout all dialogue coverage.

[Shot 1 — 00:00.000–00:03.200]
ESTABLISHING TWO-SHOT — QUIET DISTANCE

A static 35mm medium-wide two-shot establishes the entire dramatic relationship.

The man sits on one side of the small window table. The woman sits opposite him. The third empty chair is clearly visible beside the table from the beginning.

The window and a substantial portion of the nearly empty restaurant remain visible around them. They occupy only part of the frame, surrounded by quiet architectural negative space.

The warm ivory restaurant light creates one clean illuminated zone around the table. Outside the window, the street remains cool slate blue and gray-blue.

The man looks down and slowly cuts his steak.

Fork holds the meat.
Knife moves.
Knife lightly touches ceramic.
He continues cutting.

The woman does not eat.

She looks through the window toward the empty street.

Neither character looks at the other.

Hold the composition long enough to establish the silence, distance and repetitive knife movement.

At 00:03.200:

HARD CUT TO SHOT 2.

Do not pan, zoom or morph from Shot 1 into Shot 2.

[Shot 2 — 00:03.200–00:05.700]
INSERT / HAND AND PHOTOGRAPH — THE INTERRUPTION ENTERS THE SCENE

Cut to a tighter 55mm table-level medium insert from the woman's side.

The composition shows the woman's lower torso and hand, the tabletop, the man's hands and plate in the opposite portion of frame. Faces may be partially outside the composition.

The man's knife continues its same slow cutting rhythm established in Shot 1.

The woman first withdraws her gaze from the window.

Her hand enters her coat pocket.

She removes one small black-and-white photograph.

She places it flat against the wooden table.

Her index and middle fingers remain on its edge.

She slowly pushes the photograph across the table.

The photograph moves through the foreground toward the man's plate.

The photograph enters the man's lower field of view.

Only then—

his knife abruptly stops halfway through a cut.

The still knife becomes the visual reaction.

Hold approximately 0.6–0.8 seconds.

The photograph remains between his hands and the woman's fingers.

The image on the photograph is visibly photographic but the depicted person's identity remains unclear.

At 00:05.700:

HARD CUT TO SHOT 3.

[Shot 3 — 00:05.700–00:08.000]
MAN'S REACTION — CONCEALMENT

Cut to a restrained 65mm medium close shot of the man from slightly over the woman's shoulder.

The blurred edge of the woman's shoulder may occupy a small portion of foreground, establishing their eyeline.

Do not use an extreme facial close-up.

The man's face, hands, plate and edge of the photograph remain readable.

His knife is still frozen.

His eyes are directed downward.

He does not immediately look at the photograph.
He does not look at the woman.

After the brief pause, he deliberately resumes the cutting motion.

The knife presses through the steak.

One controlled cut.

His face remains almost neutral, but the jaw becomes slightly firmer and the movement is more deliberate than before.

The important emotional information is:

he reacted,
he knows that he reacted,
and he is now deliberately pretending that nothing happened.

The woman remains silent outside or at the edge of frame.

At approximately 00:07.300, the man's knife completes the cut.

His gaze remains lowered.

At 00:08.000:

HARD CUT TO SHOT 4.

[Shot 4 — 00:08.000–00:11.700]
REVERSE / DIALOGUE AND EYE CONTACT — THE TENSION PEAKS

Cut to a 50mm medium two-shot from a slightly different angle across the table.

Both faces are now clearly readable.

The photograph occupies the lower central region of the composition between them.

The empty chair remains visible deeper or beside the table whenever composition allows, but do not distort the geography merely to force it into frame.

The woman is already watching the man.

Her expression is restrained.

No anger.
No accusation gesture.
No leaning aggressively forward.

She asks quietly:

"You've seen her before?"

CUT WITHIN THE SAME SHOT GROUP to a restrained 65mm medium shot of the man immediately after the question.

Do not change screen direction.

The man remains looking downward.

Hold approximately 1.5 seconds of silence.

No knife movement.

No unnecessary facial movement.

Then, still looking at his plate, he answers:

"No."

A short pause follows the word.

Only now do his eyes rise.

First from the plate—

then briefly across the photograph—

then toward the woman.

His head follows the eyes only slightly.

CUT BACK to the 50mm two-shot as their gazes finally meet.

The woman does not react dramatically to the lie.

She simply continues looking at him.

The man holds her gaze.

The repetitive dinner movement has stopped completely.

The tension comes from the difference between what he says and what his earlier physical reaction already revealed.

At 00:11.700:

HARD CUT TO SHOT 5.

[Shot 5 — 00:11.700–00:13.000]
PHOTOGRAPH AND ABSENT PERSON — VISUAL REFRAMING

Cut to a low, restrained 70mm tabletop insert.

The black-and-white photograph lies sharply readable as an object near the center foreground, but the face of the person inside it remains indistinct.

The man's plate and motionless knife occupy one side of the background.

The woman's untouched plate occupies the other.

Beyond the photograph, slightly out of focus but unmistakable, stands the third empty chair.

This composition visually connects:

the photograph
→ the empty chair
→ the absent third person.

Do not explain this connection literally.

No hands touch the photograph.

Neither character moves significantly.

At 00:13.000:

HARD CUT TO SHOT 6.

[Shot 6 — 00:13.000–00:15.000]
WIDE PULLBACK — ISOLATION AND UNRESOLVED ENDING

Cut to a new 32mm symmetrical-to-slightly-asymmetrical wide composition from farther inside the restaurant.

This is clearly a new camera position, not a continuation of Shot 5.

The man and woman sit opposite each other beside the large window.

The third chair is clearly empty.

The photograph remains on the table between them.

Both characters continue holding eye contact.

Now begin one slow, smooth backward dolly.

This is the ONLY substantial continuous camera movement in the sequence.

As the camera retreats, the table and characters become progressively smaller.

More empty tables, unused chairs, pale wall planes, dark wooden architectural lines and the large blue-gray window enter the frame.

Negative space grows around the characters.

The restaurant architecture gradually becomes more visually dominant than the people.

Outside, one car passes along the street.

Cold neutral-white headlights move across the exterior.

The moving light crosses the window glass first.

Then it sweeps briefly across the restaurant interior.

For a fraction of a second, the passing headlight catches the photograph on the table and the edge of the empty third chair.

It also passes softly across the two characters.

Neither person looks away.

The car leaves.

The moving light disappears.

The restaurant returns to its stable warm-ivory interior light against the cool slate-blue night outside.

During the final 0.50 seconds, the camera stops.

Hold the final composition.

The man and woman remain small within the large restaurant.

They continue looking at one another.

The photograph remains between them.

The third chair remains empty.

No one moves.

End without revealing who the person in the photograph is and without resolving whether the man's answer is truthful.

EDITING AND CAMERA RHYTHM:

The sequence must contain clearly perceptible editorial cuts.

Required visual rhythm:

00:00.000 — medium-wide establishing two-shot
→
00:03.200 — HARD CUT to table/photograph insert
→
00:05.700 — HARD CUT to man's reaction shot
→
00:08.000 — HARD CUT to dialogue coverage and reverse reaction
→
00:11.700 — HARD CUT to photograph + empty-chair narrative insert
→
00:13.000 — HARD CUT to distant wide shot, then slow pullback.

Do NOT merge these shots into one continuous take.

Do NOT interpolate camera positions as one smooth orbit.

Each hard cut establishes a distinctly different framing, focal length and narrative purpose.

The editing rhythm should evolve from observational to psychologically focused and then back to spatial isolation:

space
→ object
→ reaction
→ confrontation
→ clue
→ space.

PERFORMANCE CAUSALITY:

The dramatic progression must remain strictly ordered:

ordinary repetitive dinner behavior
→ woman decides to introduce photograph
→ photograph enters man's field of view
→ knife stops involuntarily
→ pause
→ man consciously resumes cutting
→ woman observes the concealment
→ woman asks the question
→ silence
→ man verbally denies recognition while avoiding eye contact
→ only after answering does he look up
→ mutual gaze
→ photograph and empty chair become visually associated
→ camera withdraws
→ unresolved silence.

Do not allow emotional reactions before their triggering information.

The man's reaction to the photograph must be visible BEFORE his verbal denial, because that contradiction carries the dramatic information.

The woman's suspicion is expressed through sustained observation rather than overt accusation.

VISUAL CONTINUITY:

Preserve identical man and woman throughout all cuts.

Preserve identical clothing, hair, face, age and body proportions.

Respect the 180-degree rule.

Maintain correct left-right orientation and matching eyelines between coverage.

The photograph stays in the same position after the woman pushes it toward the man.

The steak does not regenerate between cuts.

The knife remains beside or above the man's plate after he stops cutting.

The woman's meal remains untouched.

The third chair remains empty and stationary throughout all six shots.

EDWARD HOPPER-INSPIRED VISUAL CONSTRAINTS:

Prioritize:
quiet architectural geometry,
clean rectangular windows,
strong horizontal table lines,
vertical window and wall divisions,
large areas of negative space,
isolated human figures,
restrained gestures,
clear silhouette separation,
matte oil-painted surfaces,
broad opaque color masses,
subtle visible brush texture,
controlled hard-edged illumination,
psychological distance,
and ordinary urban spaces made emotionally uncanny through composition.

Use a restrained palette of:

warm ivory,
muted cream,
pale ochre,
soft beige,
dusty terracotta,
dark natural brown,
muted brick red,
slate blue,
gray-blue,
desaturated ultramarine,
neutral charcoal.

Warm and cool areas should remain visually distinct rather than blending into muddy green.

Keep saturation low to moderate.

Keep faces and foreground objects clearly exposed.

Do not make the image predominantly yellow.

Do not make the image predominantly green.

Do not combine yellow light and cyan shadows into a green cast.

Do not crush the interior into darkness.

Do not imitate modern teal-and-orange cinema grading.

Do not use neon signage as a dominant light source.

Do not use high-saturation emerald, lime, chartreuse, cyan or fluorescent yellow.

Do not turn the restaurant into noir, horror, cyberpunk or a modern glossy diner.

Do not make every surface dark brown.

The final image should feel quiet, spacious, painterly and psychologically isolated rather than gloomy, dirty or sinister.

No exaggerated acting, no shouting, no crying, no aggressive gestures, no table slamming, no sudden standing, no random smiles, no excessive blinking, no theatrical eyebrow movement, no visible panic.

No extreme close-ups.

No handheld camera.

No Dutch angles.

No whip pans.

No crash zooms.

No action-style camera motion.

No continuous 15-second single take.

No duplicated photograph, no moving empty chair, no additional restaurant customers, no waiters, no extra human figures, no prop teleportation, no face drift, no costume changes, no extra limbs, no fused hands, no visible subtitles, no captions, no logo and no watermark.

overall_soundscape:

A nearly silent late-night restaurant.

The opening shot contains a faint refrigerator hum, subtle room tone and extremely distant city ambience through the glass.

The man's knife produces a slow recurring scrape against the ceramic plate, establishing the initial rhythm.

At 00:03.200, coat fabric rustles softly.

The photograph touches the wooden table with a dry paper sound and slides quietly across it.

When the knife stops, the recurring ceramic scrape stops with it.

Use the sudden absence of that sound as part of the dramatic reaction.

When the man deliberately resumes cutting, one controlled knife scrape breaks the silence.

At 00:08.000, the woman asks quietly:

"You've seen her before?"

Leave approximately 1.5 seconds of genuine silence.

The man answers:

"No."

No further dialogue.

During the final pullback, a distant car approaches outside.

Its tire noise moves spatially across the window and recedes as the headlights pass.

After it leaves, only the refrigerator hum and faint distant city ambience remain.

non_diegetic_music:

Use almost no conventional score.

A barely audible sustained low-string texture may enter only after the photograph reaches the man.

Do not use percussion.

Do not musically accent the knife stopping.

Do not add a suspense sting when the photograph appears.

During the silence before "No," allow the underscore to become even quieter.

When the man finally raises his eyes, introduce one restrained unresolved low-string interval.

During the final pullback, sustain this unresolved harmony at very low volume.

The passing car receives no musical accent.

End without cadence or emotional resolution.

aspect_ratio

duration

Instructions

Uses DasiwaMinimaxH3 Hybrid V2 as the primary model, enhanced by a high/low dual-sampling architecture and multiple specialized LoRAs, including Singularity, Combat, Motion Repair, Cinemagrade, and MiniMax LMS. The workflow balances visual detail, motion stability, and dynamic performance, making it suitable for both cinematic character-driven scenes and fast-paced action sequences. It also integrates high-quality upscaling for clearer and more refined final output. A 15-second high-quality upscaled video can be generated in approximately 8 minutes at a cost of around $0.455.

(Dasiwa_v2_singularity)MiniMax_H3双采参考生视频. Uses DasiwaMinimaxH3 Hybrid V2 as the primary model, enhanced by a high/low dual-sampling architecture and multiple specialized LoRAs, inc…

Gallery

No creation yet