Skip to content

MiniMax H3 prompts

Clips people made with MiniMax H3 and the prompts behind them. Press «Copy», paste into your ComfyUI and adapt. 46 examples so far.

Grim reaper asking for candy

by Moonbear_AIArt · 934 likes · 2026-10-27

Needs a start image: the prompt uses <Picture N>

subject_definitions:
<Subject 1> is the Grim Reaper from <Picture 1>. Preserve his skull, glowing orange eyes, black hooded robe, ornate scythe and illuminated pumpkin candy bucket. Speaker (S1): VERY DEEP, gravelly, hollow, bone-dry adult male voice, with a subtle ancient rasp and faint creaking resonance. He sounds unmistakably like Death, yet speaks with perfectly polite, casual understatement. Clearly intelligible, never shouting or growling. Identical voice in every line.

<Subject 2> is the elderly homeowner from <Picture 2>. Preserve his face, white hair, beard, beige vest and shirt. Speaker (S2): natural slightly raspy elderly male voice, restrained and believable.

<Subject 3> is the same Halloween house in <Picture 3> (wide exterior) and <Picture 4> (close porch). Preserve its single front door, lanterns, wreath, pumpkins and warm indoor lighting.

summary:
[reference generation] 15-second 16:9 cinematic Halloween dark comedy. The Reaper rings a doorbell, asks for candy and casually reveals the homeowner has eighteen minutes left to live. Original English dialogue, distinct voices, natural reactions, essential comic silences. Three shots only.

retention_analysis:
<Subject 1>: fully_preserved - exact Reaper, scythe and bucket.
<Subject 2>: fully_preserved - exact elderly man and clothing.
<Subject 3>: partially_preserved - same house and consistent doorway, using <Picture 3> for the wide shot and <Picture 4> for the main shot.

detailed_description:
Warm porch light, cool Halloween night; cinematic stylized 3D. Underplay reactions. No theatrical gestures, broad screaming faces, fast dialogue or overlapping lines.

CRITICAL BLOCKING:
The Reaper stays OUTSIDE on the porch, frame right. The homeowner is NOT initially visible and appears ONLY when the door opens INWARD; he stands INSIDE his hallway, frame left, behind the threshold, and never steps outside. The doorframe stays BETWEEN them. Camera stays OUTSIDE at normal eye level. Never put both characters on the porch or open the door behind them.

[Shot 1] 00:00-00:00.800
Very short wide shot based on <Picture 3>. Reaper is already near the porch and takes his last small steps toward the front door. No dialogue or doorbell.

[Shot 2] 00:00.800-00:10.000
HARD CUT to <Picture 4> near-door composition. The door is CLOSED. Reaper stands still, then raises one hand.

00:01.900-00:02.100:
He presses the doorbell ONCE, visibly touching the button. One audible chime, only now.

00:02.100-00:02.700:
Brief wait. Door stays shut.

00:02.700-00:03.100:
Door opens INWARD. The elderly man appears for the FIRST time, standing INSIDE.

00:03.100-00:03.500:
Brief visual recognition.

00:03.500-00:04.500:
Reaper gently raises his bucket. <Subject 1> (S1), deep and gravelly but pleasantly casual: <d>[English] Trick or treat!</d>

00:04.500-00:05.000:
SILENCE. The man stiffens, eyes widening slightly.

00:05.000-00:06.300:
<Subject 2> (S2), startled but controlled: <d>[English] Oh God! Is it my time?!</d>

00:06.300-00:06.700:
Pause. The Reaper tilts his head slightly.

00:06.700-00:08.200:
<Subject 1> (S1), same deep rasp, utterly matter-of-fact: <d>[English] No, no. I'm just collecting candy.</d>

00:08.200-00:08.700:
Silence; man exhales and relaxes.

00:08.700-00:10.000:
<Subject 2> (S2), with a small relieved chuckle: <d>[English] Puh! You scared me to death!</d>

[Shot 3] At 00:10.000, CUT to a slightly tighter LOCKED two-shot from the SAME outside camera side. Reaper remains outside, man inside.

00:10.000-00:10.700:
Silent beat. Reaper checks a small wristwatch under his sleeve.

00:10.700-00:11.900:
<Subject 1> (S1), in his deep, dry, creaking Death voice, calmly: <d>[English] Don't worry. Eighteen minutes left.</d>

00:11.900-00:13.250:
SILENCE. The man's smile slowly vanishes. His gaze fixes on Death, his lips part slightly; he freezes. No exaggerated expression.

00:13.250-00:13.550:
Still SILENT. The Reaper raises his candy bucket and holds it expectantly. Tiny intentional beat BEFORE the punchline.

00:13.550-00:14.650:
<Subject 1> (S1), same gravelly voice, sincerely polite and deadpan: <d>[English] Snickers would be nice.</d>

00:14.650-00:15.000:
SILENCE. Reaper keeps his bucket extended. Man is speechless. End on the same static two-shot.

overall_soundscape:
Quiet Halloween night, ONE doorbell exactly when pressed, a door creak, a relieved exhale. Exactly S1 and S2, consistent pitch and timbre. S1's deep raspy voice remains intelligible. Never compress or overlap lines. Silent beats as marked; no laughter track.

non_diegetic_music:
Very soft whimsical spooky ambience behind early dialogue. Single restrained ominous low note after "Eighteen minutes left." Then silence, including the tiny beat before "Snickers would be nice." No music competing with dialogue.
Original on Civitai

Male developer and woman in office

by Moonbear_AIArt · 931 likes · 2026-10-13

Needs a start image: the prompt uses <Picture N>

subject_definitions:
<Subject 1> is the male developer from <Picture 1>: messy dark-brown hair, brown eyes, light stubble, charcoal hoodie with blue Civitai badge, navy shirt, jeans; same face and outfit.
<Subject 2> is the woman from <Picture 2>: auburn ponytail, green eyes, blue overshirt, white Civitai T-shirt, dark pants, white shoes; same identity and outfit.
<Subject 3> is their office from <Picture 3>-<Picture 9>: glowing blue Civitai wall logo, wooden desk, gray laptop, branded mug, plants, books, warm lamps, glass door on frame right.
<Picture 3> = panicking developer, frontal desk view.
<Picture 4> = Claude UI, over-the-shoulder.
<Picture 5> = woman entering through door.
<Picture 6> = woman at desk, developer answering.
<Picture 7> = questioning woman, palm up.
<Picture 8> = baffled developer close-up.
<Picture 9> = colorful laptop achievements, stunned woman.
summary:
[reference generation + keyframe completion] 15-second, 16:9 stylized 3D office comedy. Spoken English. Use six shots total. Preserve the exact characters, office, props, and the two important laptop screens.
retention_analysis:
<Subject 1>: fully_preserved - face, hairstyle, hoodie, shirt.
<Subject 2>: fully_preserved - face, ponytail, overshirt, shirt.
<Subject 3>: fully_preserved - office layout and desk props.
<Picture 3>, <Picture 4>, <Picture 7>, <Picture 8>, <Picture 9>: fully_preserved.
<Picture 5> to <Picture 6>: fully_preserved as one continuous action arc in a single shot, starting at the doorway composition and ending at the desk composition.
detailed_description:
High-quality expressive 3D animation, warm amber-and-blue office lighting, cinematic depth of field, same location and same outfits throughout. Use instant HARD CUTS only where explicitly stated. (S1) = young American-English man, mid-pitched voice, stressed at first, later proud and sincere. (S2) = American-English woman, clear mid-high voice, urgent, then skeptical. Accurate lip-sync. Natural dialogue timing.
[Shot 1] 00:00-00:02. HARD START on <Picture 3>. Frontal medium shot with a small slow push-in. <Subject 1> (S1) sits behind the laptop, sweating, gripping his hair, eyes wide, panicking. He says: <d>[English] Claude, hurry up! We need to deploy!</d>
[Shot 2] At 00:02.000, HARD CUT to <Picture 4>. Locked over-the-shoulder shot. Keep the Claude interface readable with completed green checks and the green banner text: "All tests passed. Ready to deploy." Nobody speaks. Screen text only. One soft UI confirmation chime.
[Shot 3] At 00:03.000, HARD CUT to a continuous shot that begins matching <Picture 5> and naturally evolves toward <Picture 6> with NO CUT inside the shot. Start as a fixed wide front-left office view: <Subject 1> turns toward the glass door in surprise while <Subject 2> (S2) rushes in through the doorway with raised hands and urgent breath, saying: <d>[English] Is it ready?! Everything's on fire!</d> As she continues moving from the doorway toward the desk and leans in at frame right, the framing subtly reframes to end near the <Picture 6> composition. Without a cut, <Subject 1> immediately answers in a natural, calm, matter-of-fact way, not overexcited, gesturing toward the laptop: <d>[English] Just finished. Deploying now!</d>
[Shot 4] At 00:07.000, HARD CUT to <Picture 7>. Static medium close-up of <Subject 2> (S2), one hand at her hip, the other palm up. Skeptical but hopeful expression. She asks: <d>[English] Great! So the bugs are fixed? Users will be happy again?</d>
[Shot 5] At 00:10.000, HARD CUT to <Picture 8>. Locked close-up. <Subject 1> (S1) freezes, brows furrow, confused eyes shifting, mouth parting in genuine bafflement. He says: <d>[English] Bu... bugs? What bugs?</d> Hold a brief awkward silence after the line.
[Shot 6] At 00:12.000, HARD CUT to <Picture 9>. Frontal medium-wide two-shot with a tiny slow push-in toward the laptop. <Subject 1> (S1) points proudly at the laptop screen and beams with innocent delight, stressing ACHIEVEMENTS: <d>[English] We have ACHIEVEMENTS now!</d> Preserve the referenced achievement UI, especially "Your Creator Journey", "Your Creator Score", "6,056,402", and "SUPERNOVA". <Subject 2> bends closer with one hand near her mouth, eyes wide in disbelief. Hold the reaction and HARD CUT TO BLACK at the end.
overall_soundscape:
Quiet office ventilation and laptop fan throughout. Light keyboard taps in Shot 1. One electronic UI ding in Shot 2. Door swish and quick footsteps at the start of Shot 3, then soft clothing movement as she crosses to the desk. Dialogue stays clean and foregrounded. Shot 2 remains speechless. After Shot 5, let the awkward silence breathe briefly.
non_diegetic_music:
Very quiet playful comedy score: light pizzicato strings and soft woodblock around 110 BPM. Slight tension in Shots 1-3, lighter playful tone in Shot 4, music pulls back in Shot 5, then a short bright comedic sting right after the word "ACHIEVEMENTS" in Shot 6. Keep music under dialogue at all times.
Original on Civitai

Man with pumpkin head

by The_Wall_of_Goat · 924 likes · 2026-10-30

This video is shot in RSNature style animation with defined yet rough sketch lines, saturated painterly colors, and concept art draft quality. This style remains consistent across movement and scene changes.

A cute Halloween style town at night with a crescent moon in the sky. The screen moves over to a Halloween style barber shop where a man with a plain pumpkin for a head sits down in the chair. A barber steps in front of him, blocking him from the camera for a moment while he works. 

Then the barber steps back, revealing the man with the pumpkin head now has a smiling jack-o'-lantern face
  • seed 298170517
  • steps 60
  • duration 6
Original on Civitai

Couple in car at night

by The_Wall_of_Goat · 903 likes · 2026-11-09

A comic book animation style, bold ink outlines and flat halftone shading.

A man in a black tuxedo and white gloves sits behind the wheel of a silver vintage luxury car; a woman in a red cocktail dress sits in the passenger seat. He grips the wheel with white knuckles, jaw set, eyes fixed on the road ahead. She presses her hand flat against the door, breath shallow, eyes flicking to the rearview mirror. The camera begins a slow dolly-out starting tight on his face, pulling back to reveal the coastal highway, the Monaco skyline, and a black car closing in behind them.

Overcast night, wet asphalt reflecting halftone highlights, no hard shadows. Audio: a low engine growl, a steady wind bed, one distant siren, only ambient room tone, no music.
  • seed 563056418
  • duration 6
  • resolution 768P
Original on Civitai

Alligator bursts from water

by The_Wall_of_Goat · 879 likes · 2026-10-16

This video is shot in RSNature style animation with defined yet rough sketch lines, saturated painterly colors, and concept art draft quality. This style remains consistent across movement and scene changes.

A dense wetland stretches across the frame, patches of murky water interspersed with overgrown marsh plants. A large alligator lies submerged, barely visible beneath the surface. Suddenly, it bursts upward, water streaming off its textured back, and snaps its powerful jaws directly at the camera lens. The camera remains locked in a tight medium shot, focusing on the alligator's emergence and the force of the snap. Overcast, diffused light illuminates the scene, highlighting the saturated greens and browns. Audio: a low, continuous wetland hum, the sharp crack of the alligator's jaws, and the splash of water, no other sounds.
  • seed 2092338673
  • steps 30
  • duration 6
Original on Civitai

Three cats in cowboy hats

by GuyFreely · 877 likes · 2026-10-14

Needs a start image: the prompt uses <Picture N>

subject_definitions:
<Subject 1> is the cat treat from <Picture 1>, a small orange square-shaped biscuit with a cat face embossed on it.
<Subject 2> is the tabby cat from <Picture 2>, wearing a dark cowboy hat and a red bandana around its neck.
<Subject 3> is the black cat from <Picture 3>, wearing a grey cowboy hat and small vest featuring intense eyes.
<Subject 4> is the white cat from <Picture 4>, wearing a tan cowboy hat and a collared shirt, with bright blue eyes.

summary:
[reference generation] The target video is a cinematic Western-style scene where three cats in cowboy hats, <Subject 2>, <Subject 3>, and <Subject 4>, compete for <Subject 1> in an old saloon. The sequence uses quick cuts between extreme close-ups of the cats' eyes and a final top-down shot showing their paws fighting over the treat, accompanied by tense Western music.

retention_analysis:
<Subject 1>: fully_preserved - the orange square biscuit with cat face embossing is retained.
<Subject 2>: fully_preserved - the tabby cat's identity, dark cowboy hat, and red bandana are retained.
<Subject 3>: fully_preserved - the black cat's appearance and eye intensity are retained.
<Subject 4>: fully_preserved - the white cat's identity, tan cowboy hat, and collared shirt are retained.

detailed_description:
50s_western. The target video is a cinematic, dramatic scene styled after classic old West films, characterized by high-contrast lighting, heavy shadows, and a gritty atmosphere.
[Shot 1] The scene opens with an extreme close-up of <Subject 1>, the orange cat treat, lying on a dark, textured wooden floor. A slow, dramatic zoom-out begins, revealing the environment of an old saloon with dim, warm light filtering through dust. As the camera continues to pull back, it reveals that <Subject 2> (the tabby), <Subject 3> (the black cat), and <Subject 4> (the white cat) are positioned in a tense triangle around the treat.
[Shot 2] At 00:01.500, there is a quick cut to an extreme close-up of the eyes of <Subject 2>. The tabby's green eyes dart intensely from side to side, reflecting the dim saloon light.
[Shot 3] At 00:02.500, a rapid cut shows an extreme close-up of the eyes of <Subject 3>. The black cat's dark, piercing eyes shift rapidly, capturing the high tension.
[Shot 4] At 00:03.500, another quick cut moves to an extreme close-up of the eyes of <Subject 4>. The white cat's blue eyes move with focused, predatory intent.
[Shot 5] At 00:04.500, the shot cuts to a top-down bird's-eye view of <Subject 1> centered on the wooden floor. Three different paws—one tabby, one black, and one white—suddenly enter the frame from the edges, frantically swiping and attempting to snag <Subject 1>. The cats produce competitive meows and sharp hisses as their paws scramble for dominance over the small treat.

overall_soundscape:
The heavy thud of paws on wooden floorboards, rhythmic cat hissing, and frantic meowing are heard during the final struggle.
Original on Civitai

Elf and bear in magical forest

by Moonbear_AIArt · 857 likes · 2026-10-21

Needs a start image: the prompt uses <Picture N>

subject_definitions:
<Subject 1>: Moonbear from <Picture 1>. Preserve exact appearance, explorer outfit and proportions. Speaking voice (S2): warm, youthful, gentle mid-pitched male voice, natural conversational English. SAME voice throughout the entire video.
<Subject 2>: Magical elf from <Picture 2>. Preserve exact appearance, outfit and wings. Speaking voice (S1): bright, light, expressive female voice, natural conversational English. SAME voice throughout.
<Subject 3>: Enchanted forest from <Picture 3> South, <Picture 4> East, <Picture 5> North and <Picture 6> West. The stone doorway and landing in <Picture 6> are the MAIN setting. Other views provide environmental continuity.
<Subject 4>: Green book "How to: World Peace" and ornate glasses from <Picture 7>.
summary:
reference generation. Create a 15-second 16:9 cinematic 3D animated fantasy comedy in English. Natural dialogue pacing, clear pauses, expressive acting and one controlled camera orbit. Moonbear already stands almost at the stone doorway. Two consistent speaking voices ONLY: elf (S1), Moonbear (S2). No overlapping dialogue. Each line has one speaker from beginning to end. The final joke is completely visual.
retention_analysis:
<Subject 1>: fully_preserved.
<Subject 2>: fully_preserved.
<Subject 3>: partially_preserved; prioritize the WEST doorway and landing, retain forest visual consistency during the camera arc.
<Subject 4>: fully_preserved; recognizable book, readable title and matching glasses.
detailed_description:
Warm afternoon sunlight, lush enchanted forest, high-quality expressive 3D character animation. Natural physical acting. Deliberate comedic pauses. Dialogue is spoken at a normal conversational pace, never rushed.
[Shot 1] Moonbear is ALREADY on the small WEST stone landing, only two short steps away from the round wooden door shown in <Picture 6>. Medium shot. He takes TWO tired little steps, stops directly at the door, sighs from exhaustion and knocks twice. No long walking sequence. No traveling through the forest. Camera remains near the doorway.
[Shot 2] At 00:01.500, the door opens and the magical elf floats out. She happily recognizes Moonbear. She gently flies in ONE smooth arc around him while the camera follows her in a controlled 90-degree tracking orbit, revealing some surrounding forest. No repeated circles. She slows down and hovers facing Moonbear. <Subject 2> (S1), with a bright, playful female voice, says naturally: <d>[English] You found me! One wish. No miracles.</d> Allow a tiny pause between the wish offer and the warning.
[Shot 3] At 00:05.200, the camera cuts to a medium close-up of Moonbear. His tired expression instantly turns delighted. <Subject 1> (S2), with his established warm youthful male voice, says enthusiastically: <d>[English] World peace, please!</d> He smiles expectantly and waits for her answer.
[Shot 4] At 00:06.800, cut to the elf. She looks stern, raises one finger and firmly corrects him. <Subject 2> (S1), using exactly her established female voice, says: <d>[English] No miracles!</d> She holds her annoyed expression briefly. A clear small pause follows.
[Shot 5] At 00:08.500, cut back to a stable medium close-up of Moonbear. This is ONE CONTINUOUS UNCUT SHOT throughout his entire sentence. He thinks for a brief moment, then brightens with a new idea. His eyes grow progressively brighter with excitement. <Subject 1> (S2), using EXACTLY the SAME warm youthful male voice as Shot 3, says at a natural conversational pace: <d>[English] Okay, then a high-end AI PC. 128 gigs of RAM. RTX 5090.</d> Small natural pauses separate the specifications. ONLY Moonbear speaks. Keep the camera on Moonbear until his entire line has completely finished. No reaction cuts, no voice changes, no interruption.
[Shot 6] At 00:12.400, cut to the elf. She stares at Moonbear, completely speechless. Her confident expression collapses into nervous disbelief. Her eyes dart slightly, and she swallows anxiously. She remains SILENT, with her lips closed. Hold this awkward reaction briefly.
[Shot 7] At 00:13.100, HARD CUT to the elf wearing the ornate glasses from <Subject 4>, desperately studying the open green book "How to: World Peace". The book title is clearly visible on the cover facing camera. She silently scans the pages with worried, intense concentration, turns one page and continues reading. Hold the visual joke until the end. No dialogue, no narration.
overall_soundscape:
Gentle forest ambience, two distinct knocks, soft magical wing flutter. Natural conversational voices with breathing space between lines. Small comedic accents accompany the elf's reactions. Completely consistent voices for S1 and S2. After the PC request, an awkward silence, followed by quiet page rustling in the final shot.
non_diegetic_music:
Soft whimsical fantasy music, gentle and unobtrusive under dialogue. Light magical flourish during the elf's entrance. Subtle playful tension during Moonbear's second wish. Music drops into awkward silence at the final book reveal. No music overpowering speech.
Original on Civitai

Girl on a skateboard in mountains at sunset

by Skynet_erobert_die_Welt · 854 likes · 2026-11-11

(masterpiece, best quality),
(Makoto Shinkai inspired),
(Weathering With You atmosphere),
beautiful detailed anime animation,
cinematic anime movie scene,
stunning hand-painted anime background,
(extremely vibrant saturated anime colors:1.5),
(rich luminous color palette:1.5),
dreamy magical atmosphere,
15-second continuous animated tracking shot,
A young blonde woman is riding her skateboard
down an empty mountain road,
seen completely from behind.
She has long golden blonde hair flowing
naturally in the wind.
She wears an oversized dark forest green
puffer jacket, loose light gray pants,
and dark sneakers with white soles.
Her skateboard has bright orange-red wheels.
The woman continuously skateboards forward
at a steady speed.
(CAMERA PERFECTLY TRACKING THE SKATEBOARDER:1.5),
(camera moving forward at exactly the same speed:1.5),
(constant distance between camera and woman:1.5),
(subject remains perfectly fixed in the frame:1.5),
(subject maintains identical size throughout:1.5),
The woman remains in the exact same position
in the lower center of the screen
throughout the entire 15-second sequence.
The camera smoothly follows directly behind her.
The road surface moves continuously beneath her skateboard.
The yellow center lines flow toward the bottom of the screen.
Nearby trees, grass, rocks and roadside objects
move backward past the camera.
Strong environmental parallax,
foreground scenery moving rapidly backward,
middle-distance scenery moving more slowly,
distant mountains moving very slowly.
IMMENSE SNOW-COVERED ALPINE MOUNTAINS,
gigantic jagged mountain peaks,
monumental rocky cliffs,
brilliant white snow with glowing blue highlights,
deep cobalt blue and vivid violet mountain shadows,
lush intensely saturated emerald green alpine meadows,
brilliant vivid green mountain slopes,
rich deep green evergreen forests,
vibrant lime green sunlit grass,
(MAGNIFICENT SUNSET TRANSFORMATION:1.5),
(a breathtaking sunset gradually unfolds over 15 seconds:1.5),
At the beginning of the clip,
the sky is brilliant deep azure blue,
with enormous luminous white cumulus clouds,
bright golden afternoon sunlight
illuminating the green alpine landscape.
As the skateboarder continues moving,
the sun slowly descends toward
the distant mountain horizon.
The sunlight progressively becomes warmer and softer.
Over the entire 15-second sequence,
the sky gradually transforms
from deep sapphire blue into
brilliant golden yellow,
vibrant orange,
intense coral pink,
luminous magenta,
and rich violet.
(beautiful slow sunset color transition:1.5),
(spectacular vibrant anime sunset sky:1.6),
(intensely saturated orange and pink clouds:1.5),
(radiant golden sunlight illuminating the mountains:1.5),
The enormous clouds slowly change
from brilliant white into glowing peach,
warm orange, vivid pink and lavender.
Golden sunlight spreads across the mountain peaks,
creating breathtaking glowing highlights.
The snow-covered mountains gradually reflect
soft pink, orange and violet sunset colors.
Long warm shadows slowly stretch
across the emerald green alpine meadows.
The woman's blonde hair becomes
beautifully illuminated by warm golden backlight.
A delicate glowing rim light gradually appears
around her hair and green jacket.
During the final seconds,
the sky becomes an extraordinary
mixture of fiery orange,
brilliant magenta,
deep purple and glowing pink.
The sun approaches the mountain horizon,
creating a magnificent golden sunset glow.
The landscape remains clearly visible,
with vibrant green meadows,
deep blue mountain shadows
and luminous colorful clouds.
(magical luminous anime lighting:1.5),
(extremely rich saturated colors:1.6),
(high color contrast and brilliant luminosity:1.5),
(vivid jewel-toned color palette:1.5),
(extreme sense of scale:1.4),
(monumental mountain landscape:1.5),
dreamlike scenery,
breathtaking fantasy atmosphere,
smooth continuous forward motion,
perfectly stabilized tracking camera,
constant framing,
consistent perspective,
fluid anime animation,
(single uninterrupted 15-second shot:1.5),
(smooth continuous sunset progression over 15 seconds:1.5),
(no sudden lighting changes:1.5),
no cuts, no transitions,
no camera rotation,
no camera zoom.
  • seed 1392890661
  • steps 30
  • duration 15
Original on Civitai

Fairy and gnome by mushroom

by phatestronaught · 852 likes · 2026-10-17

Beautiful pixie fluttering resting on a giant mushroom with her gossamer wings sparking. There’s a door in the mushroom and a small window next to it.
The door opens and a tiny and very ugly gnome with a red hat shambles out of the mushroom house. He looks up at the pixie and shakes his fist at her, shouting, "Oy, git off me house. Sod off!"
  • seed 1840569253
  • duration 8
  • resolution 768p

Ferris wheel in snow

by Petishamiaou · 788 likes · 2026-09-03

Needs a start image: the prompt uses <Picture N>

For the target video, at 0.00 seconds into the target video, <Picture 1> (from [Shot 1]) is fully referenced.

integrated_multimodal_description: [Shot 1] Ferris wheel is going down a ski jump ramp sliding and spinning. [Shot 2] At 00:05.000, The Ferris wheel reach the end of the ramp and take off. The wheel is flying. [Shot 3] At 00:10.042, The wheel is landing in the snow.

overall_soundscape: Ambient noises, sport event

non_diegetic_music: N/A
Original on Civitai

Two characters duel with lightsabers

by ravemry9 · 772 likes · 2026-10-30

Needs a start image: the prompt uses <Picture N>

subject_definitions
<Subject 1>: Revan. Use <Picture 4> for face, and <Picture 5> and <Picture 6> for outfit. Preserve appearance throughout.
<Subject 2>: Bastila Shan. Use <Picture 1> for face, and <Picture 2> and <Picture 3> for outfit. Preserve appearance throughout.
<Subject 3>: Sith warrior dueling Revan.
<Subject 4>: Two Sith guards fighting Bastila.
<Subject 5>: Metal footlocker containing exactly 10 credits and one basic med-pack.
<Subject 6>: Star Forge interior, metallic floors, industrial structures and dramatic ambient lighting.
summary
15-second cinematic, realistic KOTOR scene. Revan effortlessly wins a dual-lightsaber duel, then deliberately abandons Bastila while she fights two guards to loot a footlocker. Bastila reacts with an exasperated glare. Deadpan RPG-player humor; no dialogue or subtitles.
retention_analysis
<Subject 1>: Fully preserved; consistent face, outfit, dual lightsabers and movement.
<Subject 2>: Fully preserved; consistent face, outfit and combat behavior.
<Subject 3>: Present during the opening duel; defeated by 0:06.
<Subject 4>: Both remain fighting Bastila during Revan’s detour.
<Subject 5>: Same container and contents throughout.
<Subject 6>: Consistent location, lighting and spatial layout.
detailed_description
[Shot 1 | 0:00–0:04]
Wide, low-angle tracking shot. Revan fights <Subject 3> with two lightsabers. Fast, fluid, dance-like pivots, precise parries and sweeping attacks. Revan clearly outclasses the opponent. Keep the choreography and both blades visible.
[Shot 2 | 0:04–0:06]
Cut to a closer side angle. Revan redirects an attack and disarms the Sith with a decisive movement. The red saber falls away. Revan remains composed; the opponent is defeated.
[Shot 3 | 0:06–0:08]
Cut to a medium-wide over-the-shoulder shot. Reveal Bastila fighting both members of <Subject 4>. Revan notices <Subject 5>, turns away from the defeated Sith and deliberately walks toward it. He pumps his hand high in the air and hisses "yes!".
[Shot 4 | 0:08–0:11]
Lateral tracking shot following Revan to the footlocker. Bastila continues fighting both guards. She notices Revan leaving and glares at him in exasperation while defending herself. Bastila says, "uh, some help?". Revan calmly opens the container.
[Shot 5 | 0:11–0:15]
Brief close-up of exactly 10 credits and one basic med-pack inside the footlocker. Revan collects both. Cut to Bastila’s incredulous glare as she continues fighting, with Revan looting in the background if framing permits. Close up of Bastila's face as she facepalms in exasperation.
Maintain screen direction, character positions, weapon colors and visual continuity between shots. Natural movement, realistic anatomy and consistent reference appearance. Revan must not appear confused, robotic or glitched. No teleporting, slapstick or exaggerated expressions.
overall_soundscape
Star Forge mechanical ambience throughout. Lightsaber hums, clashes, impacts and footsteps during the duel; continued saber clashes from Bastila’s fight during the looting. Include the footlocker latch and a subtle credit clink. No dialogue, voice-over or comedic sound effects.
non_diegetic_music
Dramatic orchestral Star Wars-style score. Build intensity during the duel and maintain a serious cinematic tone through the looting. No comedic musical sting or abrupt genre change.
  • seed 55548193
  • steps 20
  • duration 15
Original on Civitai

A small robot explores space museum

by The_Wall_of_Goat · 762 likes · 2026-11-08

This video is shot in RSNature style animation with defined yet rough sketch lines, saturated painterly colors, and concept art draft quality. This style remains consistent across movement and scene changes.

Subject: a tiny brass-and-iron robot with a single round glass eye and two scuff-marked treads wanders a cosmic museum of lost Earths. Its movement is small and deliberate, head tilting as it studies each case. The exhibits sit in tall glass tanks, each holding a different Earth: one wrapped in dense green forest, one grown into a lattice of crystal, one broken into slowly drifting fragments. The final display holds a pale, unfamiliar world.

Setting: a vast dim hall of glass cases and floating dust, the floor a dark mirror, the far wall opening onto deep space so the exhibits seem to float in the void. As the robot reaches the last case, the whole hall begins to dissolve, the glass cases and dust lifting into the air like ash, the floor fading, the structure of the museum coming apart into drifting paint.

The take is one single unbroken move, with no cuts and no separate shots. The robot first stops before the final display, then reaches out and presses a small hand against the glass, and only then does the museum begin to fall away around it. As the cases dissolve, the camera slowly dollies forward through the parting glass, the hall peeling open into a sunrise over an unfamiliar planet, the robot now tiny and centered in the new sky. The robot watches the sunrise settle, its single eye reflecting the light.

Camera: a slow continuous dolly forward through the dissolving cases, the robot tracked so it stays small in the lower frame, no cuts, no reframing into a new shot, the move landing on the sunrise over the foreign planet. A shallow lens keeps the robot sharp while the far cases and the emerging sky sit at the edge of focus.

Lighting and style: the hall is lit by soft cool case-lights, the robot catching a warm rim as it nears the end. The dissolve carries the palette from the cool museum blues into the warm gold of the sunrise. Overcast, painterly, concept-draft quality, the lines rough and defined throughout, the colors saturated and the surface kept in draft.

Audio: a low ambient hum of the museum as a continuous bed, the soft scrape of the treads on the floor, a faint glassy shimmer as the cases dissolve, then the bed lifting into open wind and the low drone of a new world as the sunrise fills the frame, only ambient sound throughout, no music.

Constraints: the robot's design and the three named Earth exhibits must not change. The museum must dissolve into the sunrise in one continuous move, never cut. No on-screen text.
  • seed 1254847127
  • duration 6
  • resolution 768P
Original on Civitai

Bear looks at talking food on plate

by Moonbear_AIArt · 761 likes · 2026-10-21

Needs a start image: the prompt uses <Picture N>

subject_definitions:
<Subject 1> is Moonbear from <Picture 1>. Preserve his exact white fur, black beret, colorful paint-splatter shirt, face and full-sized bear proportions. He does not speak.
<Subject 2> is the sausage from <Picture 2>. Preserve its exact shape, face, arms and legs. A small, adult anthropomorphic food mascot. Speaker (S2): natural, upbeat male conversational voice.
<Subject 3> is the hotdog bun from <Picture 3>. Preserve its exact split-bun shape, golden crust, face, eyelashes, arms and legs. A small, adult anthropomorphic food mascot. Speaker (S1): warm, confident female conversational voice with playful comic timing.
<Subject 4> is the kitchen from <Picture 4>. Preserve the cabinets, window, countertop, white plate and warm daylight.
summary:
[reference generation] A 13-second 16:9 cinematic animated food-comedy sketch. Two adult food mascots exchange cheeky double-entendre jokes on a plate. Moonbear overhears, recoils in horrified disgust and chooses salad instead. Clear spatial scale, one gentle push-in, one upward camera tilt and one final hard cut. Natural dialogue, short pauses, strong facial acting.
retention_analysis:
<Subject 1>: fully_preserved - exact character design, clothing, face and full-sized proportions.
<Subject 2>: fully_preserved - exact sausage design and voice.
<Subject 3>: fully_preserved - exact bun design and voice.
<Subject 4>: partially_preserved - kitchen architecture, counter, plate and daylight.
detailed_description:
Warm cinematic 3D animation. Use the reference sheets only for character identities, not as grids or studio backgrounds. The bun and sausage remain separate, standing on the SAME white plate.
CRITICAL STAGING:
The plate is ON TOP of the kitchen counter, close to camera. Moonbear stands ON THE KITCHEN FLOOR BEHIND the counter, several feet behind the plate. The counter physically separates him from the tiny food mascots. Moonbear is normal bear-sized, dramatically taller than both; the food mascots are only a few inches tall. Moonbear never stands on the counter or plate. Preserve this scale and blocking throughout.
[Shot 1] Start on a medium shot of the plate in the foreground. Moonbear stands behind the counter, on the floor; only his colorful shirt and upper torso are visible in soft focus. His head is cropped ABOVE the frame. The camera makes ONE short, smooth push-in toward the plate.
The bun gives the sausage a sideways glance, one raised eyebrow and a tiny playful sway, then (S1) says with dry, mischievous confidence: <d>[English] Oh, handsome... come on, be my wiener.</d>
A short natural pause. The sausage smiles, straightens proudly, leans forward just a little and (S2) answers with cheerful confidence: <d>[English] Oh yeah... let's get dirty.</d>
They exchange a knowing grin. Hold a brief silent beat. No extra movement or dialogue.
[Shot 2] At 00:06.600, the camera tilts UPWARD from the plate, past the countertop, to reveal Moonbear's face. He remains standing on the FLOOR BEHIND the counter, towering over the tiny mascots. Stop in a steady medium close-up.
Moonbear freezes, wide-eyed and speechless. His eyebrows rise unevenly, his jaw drops, then his nose wrinkles and his mouth twists with disgust. He recoils, blinks once and raises a paw to his mouth. ONE sudden dry gagging sound, as if struggling to keep his breakfast down. No vomiting and no words. Hold his appalled stare for a comic beat.
[Shot 3] At 00:09.700, HARD CUT to the same kitchen dining table. Moonbear sits on a normal chair with a bowl of green salad; the hotdog characters are absent. Static medium shot.
He takes ONE deliberate bite of lettuce. An absurdly loud CRUNCH.
He stops chewing and stares deadpan at camera, clearly done with hotdogs. Hold to 00:13.000. No dialogue.
overall_soundscape:
Soft kitchen room tone. Exactly two distinct voices: bun (S1), natural and playful; sausage (S2), upbeat and conversational. No overlap or voice swapping. Small pause between lines, then a brief silence before the upward tilt. One audible gag from Moonbear. One exaggerated salad crunch in the final shot. No laughter track.
non_diegetic_music:
Very light, deliberately cheesy lounge jazz under the two food mascots. Abruptly stop on Moonbear's horrified reveal. Silence under the gag and final salad crunch.
Original on Civitai

Cat in crown rides a robot vacuum

by moshie · 654 likes · 2026-11-02

Needs a start image: the prompt uses <Picture N>

For the target video, at 0.00 seconds into the target video, <Picture 1> (from [Shot 1]) is fully referenced.

integrated_multimodal_description: [Shot 1] She is a fully grown adult woman with a mature, adult body and adult proportions. A digital speed painting with loose bold brushstrokes and warm window sunlight; there are no people in the scene, and the only characters are the white cat wearing a gold crown and the robot vacuum with glowing blue eyes, which keep their appearance and positions from the image, the cat seated upright in the gold chariot on top of the vacuum holding up a silver spoon. The cat taps the vacuum's lid twice with the silver spoon, and the vacuum's blue eyes blink as it glides forward across the carpet with a steady hum. The cat waves the spoon slowly side to side like a queen greeting her subjects, and the red flag flutters behind her. The vacuum swings in a wide turn past the sofa, leaving a bright clean stripe in the carpet, and the cat lifts her chin even higher. The camera pans to follow. No dialogue. No voiceover. No singing. No lyrics.

overall_soundscape: A sunny living room, the low steady hum of the vacuum, two light clinks of a spoon on plastic, and soft carpet brushing under the wheels.

non_diegetic_music: N/A
  • steps 20
Original on Civitai

Female android crumbles in blazing flames

by Takah0x · 586 likes · 2026-10-23

A female android is engulfed in blazing flames and begins to crumble in parts. Parts fall off, and her arm detaches. Finally, her entire body collapses.
The camera remains stationary.
with no music.
  • seed 428332256
  • steps 40
  • duration 6

Woman walks in eerie abstract space

by Takah0x · 581 likes · 2026-09-23

A woman stands alone in purgatory. An eerie black hole stirs.  Energy Like a black mist swirling. When the girl begins to walk quietly, It ends with a terrifying, demon-like face appearing from the hole. terrifying, terrifying.
japanese sumi artwork motion .

Man pushes a giant cat uphill

by 1aDD · 579 likes · 2026-10-28

A surreal, darkly humorous scene inspired by the myth of Sisyphus, rendered in the style of a classic vintage philosophical illustration.

COMPOSITION AND ENVIRONMENT:

A completely static, wide side view. Pale, warm sepia sky occupies approximately the upper half and left side of the frame. A massive, rugged rock formation dominates the right half and lower portion. A steep, uneven rocky slope rises diagonally from the lower left toward the upper right. Open sky contrasts with ancient, weathered stone. Muted sandy beige tones, realistic brown rock textures, soft shadows, subtle film grain, timeless vintage illustration aesthetic. The landscape is an exposed rocky hillside beneath an open sky.

SISYPHUS:

A dark-skinned, powerfully muscular adult man resembling the classical depiction of Sisyphus. Bare chest, broad shoulders, muscular arms, defined back and powerful legs. He wears only a simple, rough, beige ancient Greek loincloth tied around his waist. Barefoot, without armor, shirt or accessories. His appearance is primitive and mythological. His movements are physically realistic, showing tremendous effort as he pushes a heavy object uphill.

THE GIANT CAT:

One absurdly enormous, fluffy domestic cat, several times larger than the man. The cat is curled tightly into a compact, nearly spherical ball, with its paws tucked against its body, tail wrapped around itself, and head nestled close to its chest. Its fluffy fur creates a rounded silhouette, but its feline face and small ears remain recognizable. The cat is peaceful and sleepy, resembling a giant furry ball.

ACTION:

OPENING: The rocky landscape and pale sepia sky are completely empty and motionless.

THE UPHILL PUSH: Sisyphus enters from the lower left, pushing the enormous curled-up cat uphill. He places both hands firmly against the back of the cat's spherical body and leans forward, using his powerful legs to drive it up the steep slope.

The man must physically ROLL the curled-up cat uphill, like a giant spherical object. The cat's entire curled body rotates continuously around its center as it rolls over the rocky ground. Its body visibly turns with every rotation, and its fur makes natural contact with the uneven stone surface. The cat rolls forward through successive rotations, moving diagonally uphill from the lower left toward the upper right. It must not slide, glide, float, or simply move uphill without rotating.

Sisyphus walks behind the cat, continuously pushing it with both hands. His feet grip the rocks, his muscles tense, and his torso leans heavily into the enormous furry sphere. He repeatedly adjusts his footing while maintaining contact with the cat. His movements are slow, exhausting, and determined. The cat remains tightly curled into a spherical shape throughout the entire ascent.

THE SUMMIT: After immense effort, Sisyphus rolls the giant cat over the crest of the rocky slope. The man and the still-curled cat disappear beyond the upper-right edge of the frame.

THE COMIC TWIST: After a brief pause, the same giant cat reappears at the top of the slope. It stops rolling, uncurls naturally, stretches its legs, and stands on all four paws. Without any concern for the man's enormous effort, it suddenly sprints downhill. The enormous fluffy cat runs diagonally toward the lower left, its paws moving rapidly, its round body bouncing comically with each stride, and its tail raised. It scrambles over the uneven rocks like an ordinary cat, but on an absurdly gigantic scale. The cat runs all the way down the slope and disappears beyond the lower-left edge. It does not roll downhill. It runs naturally on four legs. Sisyphus does not reappear.

ENDING: The frame returns to the completely empty rocky hillside beneath the pale sepia sky. The landscape remains perfectly motionless, matching the opening composition.

STRICT VISUAL CONSISTENCY:
The camera is completely locked off. No camera movement, panning, zooming, camera shake, cuts, or transitions. The sky, rock formations, slope, lighting, shadows, and composition never change. Only Sisyphus and the single giant cat move.

The man's appearance remains consistent: dark skin, muscular bare torso, ancient beige loincloth, barefoot. The cat retains the same fur color, size, and recognizable feline anatomy throughout the video.

During the ascent, the cat remains tightly curled and physically rolls uphill through visible body rotation as the man pushes it. At the summit, the cat uncurls and runs downhill on four legs. These are two distinct, clearly visible actions.

No sliding during the uphill push, no floating, no teleportation, no duplicate cats, no additional characters, no environmental deformation.

The humor comes from the contrast between Sisyphus's exhausting physical labor and the cat's casual decision to run back down immediately after reaching the top. Serious mythological atmosphere, realistic physical movement, absurd proportions, and timeless philosophical symbolism.
  • model video\DasiwaMinimaxH3_dasiwaHybridV4Turbo_int8
  • steps 8
  • denoise 1
  • sampler Euler
  • CFG 1
  • scheduler simple
Original on Civitai

Angel and demon arguing at monitor

by TheVideoGuy · 555 likes · 2026-10-20

Use the attached image as the exact first frame. Preserve the same monitor view, desk, room, devil on the left side of the screen, angel on the right side of the screen, red R button on the left, blue PG button on the right, and the mouse pointer already visible near the lower middle area of the screen.

6 seconds, single continuous shot.

At the start, the red devil mascot on the left points at the red R button and says in a cute mischievous cartoon voice, “Click here.” Immediately after that, the blue angel mascot on the right points at the blue PG button and says in a cute pleading cartoon voice, “No no, click here.”

After completion of blue angel mascot's dialogue The human user moves the mouse pointer only toward the blue PG button and clicks only the blue PG button. Do not hover over or click the red R button at any point.

After the click, a cute cat video is generated on the screen for a brief moment, clearly readable as a harmless cute cat clip. Then a large bold red warning panel pops up on screen with clean readable text: “Rated R”. No extra gibberish text, no random symbols, no fake UI junk, only the clear big warning text “Rated R”.

Only after the red “Rated R” warning has fully appeared, the devil starts laughing triumphantly with cute evil laughter. After the devil begins laughing, the angel reacts with disappointment and does a facepalm.

Important action order must be exact:
1) devil says “Click here”
2) angel says “No no, click here”
3) mouse moves only to blue PG
4) blue PG is clicked
5) cute cat video appears
6) big red “Rated R” warning pops up
7) devil laughs
8) angel facepalms

Keep both mascots on their own sides the whole time. Devil always stays on the left side, angel always stays on the right side. No switching sides, no crossing, no flying around randomly. Keep the shot visually clear, funny, and readable. Cute expressive mascot animation, clean UI animation, cinematic cozy room lighting, no gibberish speech, no subtitles, no extra text besides the big red “Rated R” warning.
  • steps 4
Original on Civitai

Two characters having a serious conversation

by MonkeyBrain_JS · 547 likes · 2026-10-25

<Subject 1> is having a heart-to-heart conversation with<Subject 2>. <Subject 1> said, "I have decided!, I don't need Potter!". <Subject 2> said, "Very wise of you, Draconia!, very wise indeed!".
  • steps 8
  • sampler Euler
  • CFG 1

Woman with folding fan performs dance

by Takah0x · 545 likes · 2026-11-07

Accompanied by the sound of the shamisen, she performs a traditional Japanese dance with graceful movements. She extends an arm—holding a folding fan—and slowly turns. The camera gradually zooms in
  • seed 1162705654
  • duration 6
  • resolution 768p

Girl in white suit rides motorcycle

by Takah0x · 539 likes · 2026-10-09

cg anime style ,cyberpunk . Detailed white leather suit.  A futuristic cyberpunk biker doing a perfect vertical wheelie on a neon-lit street. As the motorcycle speeds by, the neon sign flickers. 

And she raises the gun driving with one hand, aiming it forward Just as the camera zooms in on her face,

The camera follows.
  • seed 1281014775
  • steps 40
  • denoise 1
  • sampler res_multistep
  • scheduler simple

Giant fox jumps in front of camera

by Takah0x · 524 likes · 2026-11-05

A giant nine-tailed fox jumps, flies off-screen, and lands right in front of the camera. Dynamic. reaching for vierwer.
  • seed 63432883
  • steps 30
  • denoise 1
  • sampler res_multistep
  • scheduler simple