OpenZhenZhen AI Prompt Gallery · MiniMax H3 · video

H3 Progressive Sampling: SD2.5 Bionic Flow, True Cinematic Quality

H3 Progressive Sampling: SD2.5 Bionic Flow, True Cinematic Quality — explore this MiniMax H3 creation on OpenZhenZhen, with its original prompt and a reusable generation example.

H3 Progressive Sampling: SD2.5 Bionic Flow, True Cinematic Quality — AI video creation
H3 Progressive Sampling: SD2.5 Bionic Flow, True Cinematic Quality

Prompt

English prompt

subject_definitions: <Subject 1> is 羊二蛋, whose identity comes exclusively from <Picture 1>: original facial features, face shape, body proportions, curved brown horns, short black hair, pale cream wool, thick eyebrows, drooping sheep ears, brick-red ancient-style robe, cream crossed collar, teal-green belt, and dark-blue loose trousers. He is already unsealed and remains physically free throughout this sequence. <Subject 2> is the scripture-library environment from <Picture 2>: dark wooden bookshelves, wooden desk, computer, intact lattice windows, cool blue moonlight outside, warm lanterns, and the original room layout. <Subject 3> is 大肥鱼, whose identity comes exclusively from <Picture 3>: original adult facial features and slender proportions, long gradient-blue hair, blue eyes, whale ear fins, white maid headdress, dark-blue-and-white maid dress, white apron, and blue whale tail. <Subject 4> is 克劳德, whose identity comes exclusively from <Picture 5>: original facial features and proportions, long orange hair, golden-orange eyes, sunflower hair ornament, black bow, white-and-orange-gold dress with its original patterns, and original book. Preserve her white-gold halo with the visible text "Claude!" above it. She holds the book in her left hand throughout and casts with her right hand. Her demeanor is cold, haughty, and controlled. <Picture 4> is the four-panel storyboard composition reference, read top-left, top-right, bottom-left, bottom-right. Use it for shot order, framing, relative positions, and effect placement. Character identity references and the environment reference take priority over storyboard inconsistencies. Render four successive full-screen shots, never a collage. summary: [reference generation] Generate a 15-second, 21:9 cinematic 3D animated sequence with exactly four shots inside <Subject 2>. Continue after <Subject 1> has already been freed and a blue cultivation-sharing network has already formed. An anomaly attracts <Subject 4>, who descends continuously through a spatial rift and attempts to seal the entire network. Her golden seal suppresses most nodes, but surviving independent nodes reconnect around the sealed central source. End in a tense three-character confrontation, with the blue network operating and the golden seal still present. There is exactly one spoken event, delivered by <Subject 4> in Shot 2. retention_analysis: <Subject 1> (appears in [Shot 1], [Shot 3], [Shot 4]): fully_preserved - retain his complete identity, costume, proportions, and already-unsealed physical state from <Picture 1>. <Subject 2> (appears throughout): fully_preserved - preserve the room layout, bookshelves, desk, computer, intact windows, lanterns, and consistent light directions from <Picture 2>. <Subject 3> (appears in [Shot 1], [Shot 3], [Shot 4]): fully_preserved - retain her complete adult identity, maid costume, whale features, and slender proportions from <Picture 3>. <Subject 4> (appears in [Shot 2], [Shot 3], [Shot 4]): fully_preserved - retain her complete identity, costume, halo, halo lettering, and continuously held book from <Picture 5>. <Picture 4> (composition reference across all four shots): partially_preserved - preserve panel order and target compositions while making the speaker's mouth readable during the single dialogue event. detailed_description: Cinematic 3D animation, 21:9 widescreen, soft natural facial shading, restrained highlights, and detailed hair and fabric. Warm lanterns contrast with cool moonlight inside the same intact library. Preserve the confrontation axis and established physical positions across camera changes. The blue network consists of translucent rectangular nodes containing seated cultivator silhouettes, connected by blue data lines; these silhouettes never become additional physical characters. Do not replay the previous unsealing or create another "昇腾" formation. [Shot 1] A side-front medium close two-shot follows the top-left panel of <Picture 4>. <Subject 1> occupies the left foreground in right-facing three-quarter profile; <Subject 3> stands center-right beside him. Among the existing blue nodes behind them, a central red anomaly node flashes with a red silhouette and triangular warning symbol. Both react with widened eyes and tense brows, keeping their lips closed without speech-like mouth movement. The camera pushes in slowly with small amplitude, retaining both faces and the warning node. A short electronic alert sounds. No speech, rift, attacker, upward tilt, or additional cut. [Shot 2] At 00:03.000, the camera cuts to an extreme low-angle view from near the library floor. A white-gold spatial rift opens within the upper room without damaging the ceiling. The halo emerges first; <Subject 4> then becomes continuously visible through the opening and descends, her hair and skirt responding to the airflow. Her left hand retains the book. The camera pushes in slowly with small amplitude, keeping her entire face inside the frame. By 00:05.000, she has settled into a stable hover matching the top-right storyboard panel. Stop the camera push before she speaks; do not continue revealing her body during the line. From 00:05.100 to 00:06.500, <Subject 4> (S1), the only speaker, looks down at the pair and says with a clear young adult female Mandarin voice, a flat synthetic timbre, and restrained contempt, <d>[Chinese]又是你们。</d> Her unobstructed mouth articulates the complete sentence once, then closes. Hold her cold expression through the cut. Reduce rift and halo effects beneath her voice. [Shot 3] At 00:07.000, the camera cuts to a static medium-wide rear view matching the bottom-left panel. <Subject 3> occupies the lower left, <Subject 1> the lower center, and <Subject 4> floats upper center. Without releasing her book, <Subject 4> raises her right hand to cast a network-wide seal. Golden chains extend from her casting energy, cross throughout the network, and carry suspended talismans marked "封". The chains target the network, never bind either companion. As they tighten, blue connections break and most nodes dim. A few independent nodes retaining local technique copies remain faintly illuminated. Near 00:10.500, one surviving node sends a blue pulse toward another, which begins flickering. No speech; everyone keeps their lips closed. [Shot 4] At 00:11.000, the camera cuts to the bottom-right storyboard composition: <Subject 3> on the left, <Subject 1> in the center, and <Subject 4> in the right foreground, facing left in profile. From 00:11.000 to 00:12.200, pulses spread from the surviving nodes; blue connections progressively route around the golden seal and restore the distributed network. The central source remains sealed. From 00:12.200 to 00:13.000, <Subject 4>'s still-held book emits red warning light and displays the visible text "全网封锁失败", angled legibly toward the camera without an insert shot. A short nonverbal error beep sounds. She glances down, her faint smirk fading into a cold stare. From 00:13.000 to the end, <Subject 1> raises his clenched fists near his chest with a restrained, defiant smile, without punching, jumping, or celebrating. <Subject 3> remains alert beside him. <Subject 4> returns his gaze, still floating. The camera pushes in very slowly with small amplitude while preserving all three positions. End with the blue network active, the central source sealed, and golden chains and talismans still suspended. No new attack, surrender, departure, or final victory. No speech. overall_soundscape: Quiet library ambience, subtle blue-network pulses, one initial electronic alert, restrained rift resonance, halo hum, chain extension and tightening, progressive reconnection sounds, and one short nonverbal book-error beep. The only human vocal event is S1 in Shot 2, confined to 00:05.100–00:06.500, with no repetition, overlap, or carryover across cuts. The other two characters remain silent with no speech-like mouth movement; the book warning is visible text accompanied by a beep, never a spoken system announcement. No narration, extra vocalizations, dialogue subtitles, speech bubbles, time labels, panel borders, watermarks, or background music. non_diegetic_music: N/A

View API model

Explore more AI image and video prompts · Try image generation · Try video generation · Create on the AI canvas