速報 そば屋逮捕連行

MiniMax H3 チャプター別プロンプト（ch1〜ch6 逐語）

## ch1

Required attached input files: <Picture 1> = clip1_start.png — start keyframe, the video's FIRST frame; <Picture 2> = clip1_end.png — end keyframe, the video's LAST frame; <Audio 1> = clip1_line1_narrator_h3pad.wav — the on-screen anchor's spoken line, use AS-IS as the dialogue audio. These attachments are REQUIRED inputs and must remain attached for this generation. The video starts EXACTLY on <Picture 1> and ends EXACTLY on <Picture 2>. This chapter is 5.2 seconds long.

Cinematic photo-based realistic Japanese live-news studio shot. Exactly ONE generic male Japanese news anchor is on screen, in his forties, with short dark hair, seated behind a polished glass-topped news desk with both hands folded on it, wearing a dark navy suit, a white shirt and a navy striped tie with a small clip-on lapel microphone. Behind him are blank bright blue broadcast panels on the left and a blank deep red panel on the right, with brushed metal trim, no station logo and no lettering anywhere. Bright midday daylight-balanced studio lighting, realistic skin, cloth, brushed metal and glass, authentic Japanese live-news broadcast cinematography.

Camera: a MEDIUM news-anchor composition at eye level; the camera pushes in with small amplitude at slow speed across the whole chapter and never pans, tilts, shakes or cuts, ending on the slightly closer framing of <Picture 2>.

ONLY the anchor (S1) speaks — he says <d>[Japanese] 速報です。</d>, lip-syncing to <Audio 1>; he begins the line almost immediately and his mouth moves ONLY while <Audio 1> is playing; once the line ends his mouth stays naturally CLOSED for the rest of the chapter. Use <Audio 1> AS-IS as the dialogue audio and do NOT generate any voice — no synthesized speech, no off-screen narrator, no second voice. He stays seated and upright with only small natural head movement and one calm blink; his hands stay folded on the desk.

Exactly one person is on screen, appearing EXACTLY ONCE; no duplicate anchor, no extra people, no police officers, no masked man, no beer mug. The desk, the blank studio panels and the lighting stay unchanged from the first frame to the last. Do NOT render any on-screen text — no subtitles, no captions, no lower thirds, no station logo, no lettering, no Japanese characters; the video must contain no text at all.

Soundscape: quiet live-news studio room tone, subtle desk and studio equipment ambience, no applause, no audience. Music: no background music.

Cinematic photo-based realism with physically grounded textures, realistic skin, cloth, brushed metal, glass and studio surfaces, authentic Japanese live-news broadcast cinematography, bright midday daylight-balanced lighting; no anime, cartoon, or 3D toy look.


## ch2

The video starts EXACTLY on the attached first frame and ends EXACTLY on the attached last frame. This chapter is 5.9 seconds long.

Editorial CUT from a news studio to a bright midday police-station front walkway, shot as live-news B-roll. Exactly THREE people are on screen: Sobaya in the center — a hulking muscular gray-skinned man, approximately 180cm and 100kg, with short spiky black hair, wearing a plain fitted white short-sleeve T-shirt and dark jeans — flanked by TWO generic uniformed Japanese police officers in dark navy uniforms with peaked caps, white shirts and dark ties. Each person appears EXACTLY ONCE; there is only ONE Sobaya and exactly TWO officers, never duplicates, no extra officers, no reporters, no bystanders.

PRESERVE Sobaya's mask exactly as it is in the first frame — a smooth glossy white full-face mask with two large perfectly round black eye openings with no pupils, one horizontal black rectangular mouth slot, one small black dot high on the forehead, and exactly four red tapered markings arranged symmetrically as an upper pair and a lower pair on the two sides of the face. The count, shape, color and placement of those four red markings never change; no extra red streaks, spatter, drips or additional marks ever appear, no vents or holes are ever added, the mask never comes off and no face is revealed. His skin stays uniformly gray and his hair stays black and spiky.

Sobaya's hands are held together in front of his body by ONE pair of closed steel handcuffs; the cuffs stay CLOSED and in front for the entire chapter and never open.

Camera: a WIDE establishing composition at eye level, locked-off static camera — no pan, no tilt, no push-in, no shake.

Motion: all three begin standing calmly; the two officers then start a calm, non-violent escort, each holding Sobaya lightly by one upper arm, and by the last frame Sobaya has taken one small step forward with the officers moving with him. Nobody hurries, nobody pulls and nobody struggles.

There is no dialogue and no narration in this chapter — nobody speaks, and Sobaya and both officers keep their mouths CLOSED throughout. BRIGHT MIDDAY DAYLIGHT, sunlit exterior, DAYTIME NOT night; no darkness, no night sky. No struggle, no injury, no weapon, no abuse, no beer mug. Do NOT render any on-screen text — no subtitles, no captions, no lettering, no Japanese characters, no signage; the video must contain no text at all.

Soundscape: bright outdoor police-station ambience, soft footsteps beginning on paving, light uniform fabric movement, a faint handcuff chain tick, distant city traffic, no shouting and no speech. Music: no background music.

Cinematic photo-based realism with physically grounded textures, realistic skin, cloth, metal, glass and studio surfaces, authentic Japanese live-news broadcast cinematography, bright midday daylight-balanced lighting, and the official Sobaya character design integrated into a believable Japanese police escort scene; no anime, cartoon, or 3D toy look.


## ch3

The video starts EXACTLY on the attached first frame and ends EXACTLY on the attached last frame. This chapter is 5.2 seconds long.

Editorial CUT to a WIDE three-quarter side angle of the same arrest escort on the same bright midday police-station walkway, shot as live-news B-roll. Exactly THREE people are on screen: Sobaya — a hulking muscular gray-skinned man, approximately 180cm and 100kg, with short spiky black hair, wearing a plain fitted white short-sleeve T-shirt and dark jeans — and the SAME TWO generic uniformed Japanese police officers in dark navy uniforms with peaked caps. Each person appears EXACTLY ONCE; no duplicated Sobaya, no duplicated officer, no extra people, no reporters, no bystanders.

PRESERVE Sobaya's mask exactly as it is in the first frame — a smooth glossy white full-face mask with two large perfectly round black eye openings with no pupils, one horizontal black rectangular mouth slot, one small black dot high on the forehead, and exactly four red tapered markings arranged symmetrically as an upper pair and a lower pair on the two sides of the face. The count, shape, color and placement of those four red markings never change; no extra red streaks, spatter or additional marks ever appear, no vents or holes are ever added, the mask never comes off and no face is revealed. His skin stays uniformly gray and his hair stays black and spiky.

Sobaya's hands stay held together in front of his body by the SAME ONE pair of closed steel handcuffs; the cuffs stay CLOSED and in front for the entire chapter and never open.

Camera: a slight low angle; the camera tracks laterally with small amplitude at slow speed for the whole chapter, staying alongside the three walkers, and never cuts, whips or shakes.

Motion: all three continue walking calmly and steadily in one direction at the same unhurried pace. The officers guide Sobaya by the upper arms without force; Sobaya stays upright and calm and matches their pace, his cuffed hands swinging only slightly in front of him. By the last frame they have advanced a short distance along the walkway and the lateral tracking has moved with them.

There is no dialogue and no narration in this chapter — nobody speaks, and Sobaya and both officers keep their mouths CLOSED throughout. BRIGHT MIDDAY DAYLIGHT, sunlit police-station walkway, DAYTIME NOT night; no darkness, no night sky. No struggle, no injury, no weapon, no abuse, no beer mug. Do NOT render any on-screen text — no subtitles, no captions, no lettering, no Japanese characters, no signage; the video must contain no text at all.

Soundscape: synchronized outdoor footsteps on paving, light uniform fabric movement, a faint handcuff chain tick, distant city traffic, no shouting and no speech. Music: no background music.

Cinematic photo-based realism with physically grounded textures, realistic skin, cloth, metal, glass and studio surfaces, authentic Japanese live-news broadcast cinematography, bright midday daylight-balanced lighting, and the official Sobaya character design integrated into a believable Japanese police escort scene; no anime, cartoon, or 3D toy look.


## ch4

The video starts EXACTLY on the attached first frame and ends EXACTLY on the attached last frame. This chapter is 5.9 seconds long.

Editorial CUT to a MEDIUM waist-to-head detail angle of the same bright midday arrest escort, shot as live-news B-roll. Exactly THREE people are on screen: ONE Sobaya — a hulking muscular gray-skinned man, approximately 180cm and 100kg, with short spiky black hair, wearing a plain fitted white short-sleeve T-shirt — centered, with the SAME TWO generic uniformed Japanese police officers in dark navy uniforms and peaked caps framing him at the left and right edges without obscuring his cuffed hands. Each person appears EXACTLY ONCE; no duplicate Sobaya, no duplicate officers, no extra people, no reporters, no bystanders.

PRESERVE Sobaya's mask exactly as it is in the first frame — a smooth glossy white full-face mask with two large perfectly round black eye openings with no pupils, one horizontal black rectangular mouth slot, one small black dot high on the forehead, and exactly four red tapered markings arranged symmetrically as an upper pair and a lower pair on the two sides of the face. The count, shape, color and placement of those four red markings never change; no extra red streaks, spatter or additional marks ever appear, no vents or holes are ever added, the mask never comes off and no face is revealed. This is the closest view of the mask in the whole video, so its geometry must stay exactly as the first frame shows it. His skin stays uniformly gray and his hair stays black and spiky.

Sobaya's hands stay held together low in the frame in front of his body by the SAME ONE pair of closed steel handcuffs, which stay clearly visible, CLOSED and in front for the entire chapter and never open.

Camera: a slight low angle; the camera pushes in with small amplitude at slow speed across the whole chapter and never pans, tilts, shakes or cuts, ending on the slightly closer framing of the last frame.

Motion: the three keep walking slowly and calmly together. Sobaya turns his masked face only a few degrees toward camera with comic resignation and then holds it there; the officers keep their calm grip on his upper arms and keep facing forward.

There is no dialogue and no narration in this chapter — nobody speaks, and Sobaya and both officers keep their mouths CLOSED throughout. BRIGHT MIDDAY DAYLIGHT, sunlit exterior, DAYTIME NOT night; no darkness, no night sky. No struggle, no injury, no weapon, no abuse, no beer mug. Do NOT render any on-screen text — no subtitles, no captions, no lettering, no Japanese characters, no signage; the video must contain no text at all.

Soundscape: close outdoor footsteps on paving, quiet handcuff metal movement without violent rattling, uniform fabric movement, distant city traffic, no shouting and no speech. Music: no background music.

Cinematic photo-based realism with physically grounded textures, realistic skin, cloth, metal, glass and studio surfaces, authentic Japanese live-news broadcast cinematography, bright midday daylight-balanced lighting, and the official Sobaya character design integrated into a believable Japanese police escort scene; no anime, cartoon, or 3D toy look.


## ch5

The video starts EXACTLY on the attached first frame and ends EXACTLY on the attached last frame. This chapter is 5.2 seconds long.

Editorial CUT to a MEDIUM frontal news B-roll angle on the same bright midday police-station walkway. Exactly THREE people are on screen, walking toward camera: ONE Sobaya — a hulking muscular gray-skinned man, approximately 180cm and 100kg, with short spiky black hair, wearing a plain fitted white short-sleeve T-shirt and dark jeans — centered between the SAME TWO generic uniformed Japanese police officers in dark navy uniforms with peaked caps, one on either side. Each person appears EXACTLY ONCE; no duplicates, no extra officers, no reporters, no bystanders.

PRESERVE Sobaya's mask exactly as it is in the first frame — a smooth glossy white full-face mask with two large perfectly round black eye openings with no pupils, one horizontal black rectangular mouth slot, one small black dot high on the forehead, and exactly four red tapered markings arranged symmetrically as an upper pair and a lower pair on the two sides of the face. The count, shape, color and placement of those four red markings never change; no extra red streaks, spatter or additional marks ever appear, no vents or holes are ever added, the mask never comes off and no face is revealed. His gray skin, spiky black hair and white T-shirt stay consistent for the whole chapter.

Sobaya's hands stay held together in front of his body by the SAME ONE pair of closed steel handcuffs; the cuffs stay clearly visible, CLOSED and in front for the entire chapter and never open.

Camera: eye level; the camera pulls back with small amplitude at slow speed across the whole chapter and never pans, tilts, shakes or cuts, revealing slightly more of the sunlit walkway by the last frame.

Motion: the three continue their calm, even walk toward camera at an unhurried pace, the officers keeping a light guiding hold on Sobaya's upper arms. Sobaya stays upright and matches their pace; his cuffed hands swing only slightly in front of him.

There is no dialogue and no narration in this chapter — nobody speaks, and Sobaya and both officers keep their mouths CLOSED throughout. BRIGHT MIDDAY DAYLIGHT, sunlit exterior, DAYTIME NOT night; no darkness, no night sky. No struggle, no injury, no weapon, no abuse, no beer mug. Do NOT render any on-screen text — no subtitles, no captions, no lettering, no Japanese characters, no signage; the video must contain no text at all.

Soundscape: outdoor footsteps approaching on paving, distant city traffic, subtle uniform fabric movement, a faint handcuff chain tick, no shouting and no speech. Music: no background music.

Cinematic photo-based realism with physically grounded textures, realistic skin, cloth, metal, glass and studio surfaces, authentic Japanese live-news broadcast cinematography, bright midday daylight-balanced lighting, and the official Sobaya character design integrated into a believable Japanese police escort scene; no anime, cartoon, or 3D toy look.


## ch6

Required attached input files: <Picture 1> = clip6_start.png — start keyframe, the video's FIRST frame; <Picture 2> = clip6_end.png — end keyframe, the video's LAST frame; <Picture 3> = Sobaya_sheet.png — Sobaya's character model sheet, identity/design reference only, NOT a composition reference (PRESERVE: mask geometry, the four red markings, gray skin, spiky black hair, white T-shirt and body build; do NOT carry over: the sheet's poses, panel layout, text labels, plain background or the beer mug); <Audio 1> = clip6_line1_sobaya_h3pad.wav — Sobaya's spoken line, use AS-IS as the dialogue audio. These attachments are REQUIRED inputs and must remain attached for this generation. The video starts EXACTLY on <Picture 1> and ends EXACTLY on <Picture 2>. This chapter is 3.75 seconds long.

Editorial CUT to a MEDIUM shot of Sobaya centered between the SAME TWO generic uniformed Japanese police officers on the same bright midday police-station walkway. Exactly THREE people are on screen: ONE Sobaya (<Picture 3>, the hulking 180cm/100kg muscular gray-skinned man whose face is always covered by a white mask with red markings, wearing a white T-shirt) and TWO officers in dark navy uniforms with peaked caps, one on either side of him. Each person appears EXACTLY ONCE; no duplicates, no extra officers, no reporters, no bystanders.

PRESERVE Sobaya's mask exactly as it is in <Picture 1> and <Picture 3> — a smooth glossy white full-face mask with two large perfectly round black eye openings with no pupils, one horizontal black rectangular mouth slot, one small black dot high on the forehead, and exactly four red tapered markings arranged symmetrically as an upper pair and a lower pair on the two sides of the face. The count, shape, color and placement of those four red markings never change; no extra red streaks, spatter or additional marks ever appear, no vents or holes are ever added, the mask never comes off and no face is revealed. His hands stay held together in front of his body by the SAME ONE pair of closed steel handcuffs, which stay CLOSED and in front for the entire chapter.

Camera: eye level, locked-off static camera — no pan, no tilt, no push-in, no shake.

ONLY Sobaya (S1) speaks — he says <d>[Japanese] スーパードライが美味かった。</d>, lip-syncing to <Audio 1>; he begins the line almost immediately and the mouth slot of his mask moves ONLY while <Audio 1> is playing, then stays CLOSED for the rest of the chapter. Use <Audio 1> AS-IS as the dialogue audio and do NOT generate any voice — no synthesized speech, no narrator, no second voice. The two officers do NOT speak; their mouths stay CLOSED the whole time and they keep facing forward.

Motion: the three keep walking slowly and calmly toward camera while Sobaya delivers the line and gives one small satisfied nod; the officers keep their light guiding hold on his upper arms.

The reference sheet's text labels must NOT appear in the video. BRIGHT MIDDAY DAYLIGHT, sunlit exterior, DAYTIME NOT night; no darkness, no night sky. No struggle, no injury, no weapon, no abuse, no beer mug. Do NOT render any on-screen text — no subtitles, no captions, no lettering, no Japanese characters, no signage; the video must contain no text at all.

Soundscape: close outdoor footsteps on paving, restrained handcuff metal movement, quiet uniform fabric, distant city traffic, no shouting. Music: no background music.

Cinematic photo-based realism with physically grounded textures, realistic skin, cloth, metal, glass and studio surfaces, authentic Japanese live-news broadcast cinematography, bright midday daylight-balanced lighting, and the official Sobaya character design integrated into a believable Japanese police escort scene; no anime, cartoon, or 3D toy look.