MiniMax H3 终极提示词指南

integrated_multimodal_description:
[Shot 1] STYLE + INITIAL FRAMING + SUBJECT + ENVIRONMENT.
REFERENCE OR OPENING-STATE CONTINUITY.
SMALL NATURAL MOVEMENT.
ACTION ONSET.
SPEAKER DESCRIPTION + SPEAKER ID + DELIVERY:
<d>[Language] Exact dialogue.</d>
VISIBLE PHYSICAL RESPONSE.
CAMERA MOTION TYPE + AMPLITUDE + SPEED.
INTERMEDIATE COMPOSITIONAL CHANGES.
SYNCHRONIZED DIEGETIC SOUND.
FINAL CHARACTER STATE + FINAL COMPOSITION.

overall_soundscape:
The full environmental and physical sound environment.

non_diegetic_music:
The audience-only background score, or N/A.

subject_definitions:
Define what each reference file controls. Assign a narrow, explicit task to every Picture, Video, and Audio.

summary:
One-sentence overview of the creative direction.

retention_analysis:
For each reference, state what is preserved and what is changed.

detailed_description:
[Shot 1] STYLE + FRAMING + SUBJECT + ENVIRONMENT...

overall_soundscape:
Environmental and physical sound.

non_diegetic_music:
Audience-only background score, or N/A.

subject_definitions:
<Picture 1> defines the adult woman's exact facial identity, skin texture, and brown hair only.
<Picture 2> defines the red silk evening dress and its drape behaviour only.
<Video 1> defines the slow walking gait and hip sway timing only.
<Audio 1> defines the authorised voice timbre only.

summary:
A woman in a red silk dress walks through a rainy Shanghai alley at night, speaking one line in Chinese.

retention_analysis:
- <Picture 1>: fully_preserved (facial identity, skin, hair)
- <Picture 2>: fully_preserved (dress design, colour, material)
- <Video 1>: attribute_transfer (gait timing applied to new environment)
- <Audio 1>: fully_copy (voice timbre)
- Background from <Picture 1>: weak_reference

detailed_description:
[Shot 1] Live-action, cinematic. The woman defined in <Picture 1> wears the red silk dress from <Picture 2>. She walks through a narrow Shanghai alley at night, neon signs reflecting on wet pavement. The camera tracks behind her with medium amplitude at slow speed. She stops, turns to face the camera.

The woman with a calm, clear voice (S1) says softly:
<d>[Chinese] 雨停了，该走了。</d>
Her lips, jaw, and throat move naturally with each word.

overall_soundscape:
Light rain on wet stone, distant car horns, one passing bicycle, subtle silk fabric movement, footsteps on wet ground. No other speech.

non_diegetic_music:
N/A

integrated_multimodal_description:
[Shot 1] Live-action, cinematic. A young woman stands in a Shanghai alley at dusk, neon signs reflecting on wet pavement. She adjusts her coat and looks toward the camera. The camera holds a static medium shot. No dialogue, no voiceover. Her lips remain completely closed. Synchronized diegetic sound: distant traffic, light rain on fabric.

overall_soundscape:
Steady rain on wet stone, distant car horns, one passing bicycle with a bell, subtle fabric movement as she adjusts her coat. No speech, no vocalization.

non_diegetic_music:
Sparse solo piano, slow tempo, minor key, fading out at the end.

integrated_multimodal_description:
[Shot 1] A brushed-steel automatic watch rests on black volcanic stone. Fine rain lands on the glass while the second hand moves naturally. The camera begins with an extreme macro, then makes a slow 30-degree orbit, ending on the watch face. High-contrast luxury commercial, deep black background. Keep the logo and dial geometry unchanged. Subtle rain and room-tone audio.

overall_soundscape:
Soft mechanical ticking, light rain on glass, one distant low thunder rumble.

non_diegetic_music:
One restrained low cinematic pulse at 0s, fading by 3s. No melody.

Constraints:
No dialogue. No on-screen text. No soft dissolves.

integrated_multimodal_description:
[Shot 1] Live-action, cinematic. The young woman shown in <Picture 1> sits in a hospital corridor at night, preserving her exact facial identity and dark hair. 0 to 5s she returns with two paper cups and sits. 5 to 11s she speaks. 11 to 15s neither answers and one takes the other's hand. Medium-close coverage, restrained shot-reverse-shot.

The young woman with a calm, low, natural voice (S1) says with controlled sadness:
<d>[Chinese] 有些话，我们都不想说出口。</d>
Her lips, jaw, and throat move naturally with each word. Her eyes remain fixed forward.

overall_soundscape:
A distant trolley, the vending machine's hum, fabric movement as she sits.

non_diegetic_music:
N/A

Constraints:
No music. No subtitles. No on-screen text. No soft dissolves.

integrated_multimodal_description:
[Shot 1] Image 1 is the character and art-style reference: keep this exact hand-painted 2D webcomic look, the same watercolour texture, the same ink outline weight, the same dog, the same little hat, the same mug. Do not redraw him in 3D, do not change his proportions, do not clean up the linework.

Put him in a different room and keep his composure. A cramped open-plan office at night, fluorescent tubes flickering, a wall of monitors all showing a red error state, printer paper drifting down through the frame, a small electrical fire in the corner. He sits in an office chair at the centre of the frame, mug in paw, looking straight at the camera, completely relaxed.

Camera: locked off, static wide shot. No push in, no cuts.

overall_soundscape:
Fluorescent hum, printer grinding, soft alarm chirp every 2s, distant electrical crackle, one calm sip at 4s.

non_diegetic_music:
N/A

Constraints:
No on-screen text. No subtitles. No other characters.

How the reference pictures align with the generated video:
- Picture 1 aligns with 0.00s (the first frame).
- Picture 2 aligns with 15.00s (the last frame).

integrated_multimodal_description:
Hold the opening composition briefly. Daylight gradually fades as shop windows illuminate, traffic begins moving, and reflections appear across the wet street. Keep the buildings and camera position aligned. Resolve smoothly into the supplied nighttime frame without a sudden cut. No camera movement. No dialogue.

overall_soundscape:
Day ambience fading into evening: distant traffic increasing, one car horn, wet tire sounds, shop door chimes.

non_diegetic_music:
N/A

integrated_multimodal_description:
[Shot 1] A 21:9 ultra-wide of a single-track causeway running dead straight across a tidal flat, the incoming sea already crossing it in thin sheets. A small pale car slows, hesitates, and begins a three-point turn as water climbs the tyres. Flat overcast light, silver water against dark wet sand, the horizon a low hard line across the middle of the frame. Slow lateral drift of the camera, the emptiness on either side doing the work.

overall_soundscape:
Shallow water moving over tarmac, wind, one distant gull.

non_diegetic_music:
N/A

Constraints:
No music, no dialogue, no camera cuts, no push-in.