First-party output evidence
MiniMax H3 Video Examples & Evidence Library
This library contains real MiniMax H3 video generations produced through MiniMax3.org workflows. Each example pairs the generated output with the original prompt and available metadata so you can compare what was requested with what H3 actually produced.
The goal is not to show only attractive outputs. Where data is available, records document settings, observed strengths, visible failures, continuity issues, and practical lessons.
Use this library to study MiniMax H3 behavior, not as a claim that every prompt will reproduce the same result. AI video generation is probabilistic.
Evidence scope: MiniMax3.org first-party generations · Model family: MiniMax H3 · Last library review: September 15, 2026
How These MiniMax H3 Examples Are Documented
Each record includes as much available information as possible: original prompt, workflow, input type, resolution, duration, aspect ratio, generation date, route, attempts, output behavior, strengths, artifacts, and reproduction lessons. Historical fields that were not recorded are marked Not recorded; missing seeds, settings, and attempt counts are never invented.
Level A — Reproducible Case Study
Output, prompt, important settings, run context, analysis, and reproduction guidance.
Level B — Documented Generation
Output, original prompt, and core settings; deeper test metadata is incomplete.
Level C — Historical Example
Output exists, but important generation metadata is incomplete.
MiniMax H3 Example Library Snapshot
| Real generated clips | 15 |
|---|---|
| Level A case studies | 0 |
| Level B documented generations | 15 |
| Level C historical examples | 0 |
| Text-to-video examples | 15 |
| Reference workflows | 0 |
| Examples with documented failure notes | 0 |
What We Test With MiniMax H3 Examples
Prompt adherence
Did the output follow the requested subject, action, and sequence?
Camera control
Did H3 follow the requested tracking, orbit, push-in, bullet-time, or other camera behavior?
Motion continuity
Did subject, camera, and environment motion remain spatially coherent?
Character and object consistency
Did important visual details survive motion and shot changes?
Visual physics
Did smoke, rain, debris, cloth, water, vehicles, or transformations behave plausibly?
Audio behavior
When audio is present, did environmental sound, dialogue, and audiovisual timing match the scene?
Failure behavior
Did the result show drift, deformation, discontinuity, unexpected cuts, or geometry changes?
Browse documented MiniMax H3 outputs
Showing 15 of 15 documented outputs. Filters do not create crawlable URLs.
Action & VFX · Text to Video · 720p · 15s
Giant Cat Destroys the Suspension Bridge
Evidence level: Level B — Documented Generation
What this example tests: Tests scale, structural-collapse causality, creature proportions, and spatial continuity.
View Prompt & Result Record
- Model
- MiniMax H3
- Workflow
- Text to Video
- Resolution
- 720p
- Duration
- 15 seconds
- Aspect ratio
- 16:9
- Generation date
- Not recorded
- Attempts
- Not recorded
- Post-editing
- Not recorded
- Platform
- MiniMax3.org
Original generation prompt
Create a daylight large-scale effects sequence on a suspension bridge. Introduce an enormous creature rising behind the bridge, establish its scale, then show one controlled strike causing progressive structural failure, debris, smoke, and vehicle movement. Keep the bridge layout and creature proportions stable. Make every collapse reaction follow the initial impact, with believable timing, weight, and spatial continuity.
Output observations and failure notes require human video review and are not inferred from the prompt or thumbnail.
Action & VFX · Text to Video · 720p · 15s
Burning Galleons in the Storm
Evidence level: Level B — Documented Generation
What this example tests: Tests ship orientation, weather, fire placement, and cause-and-effect across aerial and deck-level views.
View Prompt & Result Record
- Model
- MiniMax H3
- Workflow
- Text to Video
- Resolution
- 720p
- Duration
- 15 seconds
- Aspect ratio
- 16:9
- Generation date
- Not recorded
- Attempts
- Not recorded
- Post-editing
- Not recorded
- Platform
- MiniMax3.org
Original generation prompt
Create a stormy naval action sequence involving two large wooden ships. Begin with a wide aerial view, move to deck-level combat, then track a flaming projectile into a mast impact before returning to a wider sinking view. Preserve ship orientation, weather direction, fire placement, and character geography between shots. Make the escalation readable while maintaining continuous cause and effect.
Output observations and failure notes require human video review and are not inferred from the prompt or thumbnail.
Anime · Text to Video · 720p · 15s
Silver-Haired Sisters Anime Trailer
Evidence level: Level B — Documented Generation
What this example tests: Tests two-character design retention across several locations and editorial transitions.
View Prompt & Result Record
- Model
- MiniMax H3
- Workflow
- Text to Video
- Resolution
- 720p
- Duration
- 15 seconds
- Aspect ratio
- 16:9
- Generation date
- Not recorded
- Attempts
- Not recorded
- Post-editing
- Not recorded
- Platform
- MiniMax3.org
Original generation prompt
Create an original cold-toned anime trailer featuring two related protagonists moving through classrooms, rooftops, subway platforms, rainy streets, and high-rise corridors. Maintain identical character designs, hair, clothing, and color treatment across every location. Use restrained expressions, purposeful movement, sharp editorial transitions, and a clear suspense arc. Avoid logos, celebrity likenesses, or recognizable copyrighted characters.
Output observations and failure notes require human video review and are not inferred from the prompt or thumbnail.
Action & VFX · Text to Video · 720p · 15s
Mecha Ninja Final Showdown
Evidence level: Level B — Documented Generation
What this example tests: Tests two-subject choreography, silhouette retention, impact readability, and landing continuity in rain.
View Prompt & Result Record
- Model
- MiniMax H3
- Workflow
- Text to Video
- Resolution
- 720p
- Duration
- 15 seconds
- Aspect ratio
- 16:9
- Generation date
- Not recorded
- Attempts
- Not recorded
- Post-editing
- Not recorded
- Platform
- MiniMax3.org
Original generation prompt
Create an original futuristic rooftop duel during heavy rain. Choreograph a fast exchange between a light stealth fighter and a heavier armored opponent using low-angle tracking, lateral camera movement, a controlled leap, and a precise landing. Keep both silhouettes, weapons, and armor designs consistent. Use neon reflections and brief slow motion to clarify impacts without breaking the continuity of the fight.
Output observations and failure notes require human video review and are not inferred from the prompt or thumbnail.
Cinematic · Text to Video · 720p · 15s
Silver-Haired Warrior Transformation
Evidence level: Level B — Documented Generation
What this example tests: Tests sequential armor assembly while requesting stable face, pose, proportions, and environment.
View Prompt & Result Record
- Model
- MiniMax H3
- Workflow
- Text to Video
- Resolution
- 720p
- Duration
- 15 seconds
- Aspect ratio
- 16:9
- Generation date
- Not recorded
- Attempts
- Not recorded
- Post-editing
- Not recorded
- Platform
- MiniMax3.org
Original generation prompt
Create a science-fiction armor transformation on a damaged rooftop. Begin with a determined silver-haired warrior activating a waist-mounted device as violet energy spreads through the suit. Assemble white and purple armor in a logical sequence while preserving the face, body proportions, pose, and rooftop environment. Finish with the completed fighter in a stable combat stance as particles settle and the camera holds.
Output observations and failure notes require human video review and are not inferred from the prompt or thumbnail.
Cinematic · Text to Video · 720p · 15s
Werewolf Transformation in a Foggy Forest
Evidence level: Level B — Documented Generation
What this example tests: Tests progressive anatomical transformation during a continuous orbit without an abrupt identity jump.
View Prompt & Result Record
- Model
- MiniMax H3
- Workflow
- Text to Video
- Resolution
- 720p
- Duration
- 15 seconds
- Aspect ratio
- 16:9
- Generation date
- Not recorded
- Attempts
- Not recorded
- Post-editing
- Not recorded
- Platform
- MiniMax3.org
Original generation prompt
Create a dark fantasy transformation in a foggy moonlit forest. Use a continuous orbital camera as a distressed character gradually changes into a powerful creature. Show progressive changes to facial structure, eyes, hands, posture, and surface texture without identity jumps. Keep the environment, lighting direction, wardrobe remnants, and camera distance consistent. Complete the transformation only at the end of the orbit.
Output observations and failure notes require human video review and are not inferred from the prompt or thumbnail.
Cinematic · Text to Video · 720p · 15s
Bullet-Time Race Car in the Rain
Evidence level: Level B — Documented Generation
What this example tests: Tests whether a high-speed racing shot can transition into an orbiting bullet-time beat while requesting stable vehicle direction and design.
View Prompt & Result Record
- Model
- MiniMax H3
- Workflow
- Text to Video
- Resolution
- 720p
- Duration
- 15 seconds
- Aspect ratio
- 16:9
- Generation date
- Not recorded
- Attempts
- Not recorded
- Post-editing
- Not recorded
- Platform
- MiniMax3.org
Original generation prompt
Create a cinematic racing sequence on a rain-soaked circuit. Follow a low-profile race car at extreme speed, then transition into bullet time as suspended rain droplets surround the vehicle. Orbit the car smoothly, reveal tire spray and reflections in slow motion, then snap back to full speed as the car disappears into mist. Keep the vehicle design consistent, preserve its direction of travel, and make every camera transition continuous.
Output observations and failure notes require human video review and are not inferred from the prompt or thumbnail.
Action & VFX · Text to Video · 720p · 15s
Colossal Serpent Attacks a Skyscraper
Evidence level: Level B — Documented Generation
What this example tests: Tests large-scale creature, building, aircraft, explosion, and camera geography in one disaster sequence.
View Prompt & Result Record
- Model
- MiniMax H3
- Workflow
- Text to Video
- Resolution
- 720p
- Duration
- 15 seconds
- Aspect ratio
- 16:9
- Generation date
- Not recorded
- Attempts
- Not recorded
- Post-editing
- Not recorded
- Platform
- MiniMax3.org
Original generation prompt
Create a large-scale city disaster shot in which a colossal creature coils around a glass skyscraper. Establish the full scale from street level, move closer to show surface detail and structural stress, then rise toward the building crown as aircraft approach. Finish with a controlled explosion and layered smoke. Keep the creature, building geometry, city perspective, and direction of motion spatially consistent.
Output observations and failure notes require human video review and are not inferred from the prompt or thumbnail.
Cinematic · Text to Video · 720p · 15s
Mage Contract in the Alley
Evidence level: Level B — Documented Generation
What this example tests: Tests prop placement, mixed-color lighting, subtle secondary motion, and a controlled reveal.
View Prompt & Result Record
- Model
- MiniMax H3
- Workflow
- Text to Video
- Resolution
- 720p
- Duration
- 15 seconds
- Aspect ratio
- 16:9
- Generation date
- Not recorded
- Attempts
- Not recorded
- Post-editing
- Not recorded
- Platform
- MiniMax3.org
Original generation prompt
Create an original fantasy scene in a narrow, damp alley lit by warm practical lamps and a contrasting blue magical flame. Follow an eccentric contract maker through a slow handheld push-in as paper charms, mechanical birds, and potion bottles move subtly around the set. Keep the lighting direction and prop positions coherent. End on a controlled expression change and a sudden blue-flame reveal.
Output observations and failure notes require human video review and are not inferred from the prompt or thumbnail.
Action & VFX · Text to Video · 720p · 15s
Bullet-Time Tank Strike in the Desert
Evidence level: Level B — Documented Generation
What this example tests: Tests cause-and-effect continuity from recoil and muzzle blast through projectile travel to a distant impact.
View Prompt & Result Record
- Model
- MiniMax H3
- Workflow
- Text to Video
- Resolution
- 720p
- Duration
- 15 seconds
- Aspect ratio
- 16:9
- Generation date
- Not recorded
- Attempts
- Not recorded
- Post-editing
- Not recorded
- Platform
- MiniMax3.org
Original generation prompt
Create a cinematic desert action sequence focused on the timing of a tank firing. Begin close to the barrel, show recoil and a controlled muzzle blast, then follow the projectile through heat-distorted air in slow motion. End with a distant armored impact, sparks, sand, and smoke. Maintain realistic scale, stable vehicle geometry, clear cause and effect, and coherent motion through every stage of the shot.
Output observations and failure notes require human video review and are not inferred from the prompt or thumbnail.
Cinematic · Text to Video · 720p · 15s
Private Jet Emergency Thriller
Evidence level: Level B — Documented Generation
What this example tests: Tests character and cabin continuity when moving from a calm interior to turbulence and an exterior aircraft view.
View Prompt & Result Record
- Model
- MiniMax H3
- Workflow
- Text to Video
- Resolution
- 720p
- Duration
- 15 seconds
- Aspect ratio
- 16:9
- Generation date
- Not recorded
- Attempts
- Not recorded
- Post-editing
- Not recorded
- Platform
- MiniMax3.org
Original generation prompt
Create a compact airborne emergency sequence inside a private aircraft. Begin with a calm cabin moment, introduce sudden turbulence, track the main character moving toward the cockpit, then cut to an exterior view of the aircraft banking through storm clouds. Preserve the character's appearance, cabin layout, lighting progression, and direction of movement. Build tension clearly without skipping the cause-and-effect chain.
Output observations and failure notes require human video review and are not inferred from the prompt or thumbnail.
Action & VFX · Text to Video · 720p · 15s
Squad Advance Through a Ruined City
Evidence level: Level B — Documented Generation
What this example tests: Tests multi-character identity, equipment, and relative-position continuity across movement and cover changes.
View Prompt & Result Record
- Model
- MiniMax H3
- Workflow
- Text to Video
- Resolution
- 720p
- Duration
- 15 seconds
- Aspect ratio
- 16:9
- Generation date
- Not recorded
- Attempts
- Not recorded
- Post-editing
- Not recorded
- Platform
- MiniMax3.org
Original generation prompt
Create a tense urban action sequence following a small squad through a ruined city. Track the group down a narrow alley, react to a nearby blast, and follow them into a damaged building with volumetric light. Keep every person recognizable, preserve their equipment and relative positions, and make the transitions between cover, movement, and the final rooftop silhouette feel spatially continuous.
Output observations and failure notes require human video review and are not inferred from the prompt or thumbnail.
Action & VFX · Text to Video · 720p · 15s
Hypersonic Mecha Launch Sequence
Evidence level: Level B — Documented Generation
What this example tests: Tests armor and body-proportion continuity during rapid vertical acceleration and a continuous upward camera path.
View Prompt & Result Record
- Model
- MiniMax H3
- Workflow
- Text to Video
- Resolution
- 720p
- Duration
- 15 seconds
- Aspect ratio
- 16:9
- Generation date
- Not recorded
- Attempts
- Not recorded
- Post-editing
- Not recorded
- Platform
- MiniMax3.org
Original generation prompt
Create a high-speed science-fiction launch sequence featuring a titanium mecha with cyan thrusters. Start from a low angle as it crouches and builds thrust, track the vertical takeoff through dust and debris, then follow it into the clouds. Preserve the armor design and body proportions during acceleration. Finish with a controlled sonic boom, a bright vapor trail, and a continuous camera path toward the horizon.
Output observations and failure notes require human video review and are not inferred from the prompt or thumbnail.
Action & VFX · Text to Video · 720p · 15s
Desert Soldier Mecha Armor Transformation
Evidence level: Level B — Documented Generation
What this example tests: Tests logical mechanical assembly around a human body while requesting stable face and proportions.
View Prompt & Result Record
- Model
- MiniMax H3
- Workflow
- Text to Video
- Resolution
- 720p
- Duration
- 15 seconds
- Aspect ratio
- 16:9
- Generation date
- Not recorded
- Attempts
- Not recorded
- Post-editing
- Not recorded
- Platform
- MiniMax3.org
Original generation prompt
Create a detailed mechanical armor transformation inside a dusty industrial hangar. Begin with a close orbit around a standing subject as plates, pistons, cables, and joints assemble from the feet upward. Each component must connect logically without intersecting the body. Preserve the person's face and proportions until the helmet closes. End with steam venting from the completed armor and a stable grounded pose.
Output observations and failure notes require human video review and are not inferred from the prompt or thumbnail.
Nature · Text to Video · 720p · 15s
Amazon Python Jungle Ambush
Evidence level: Level B — Documented Generation
What this example tests: Tests tracking of multiple people and a large animal through foliage, water, mud, and fast evasive motion.
View Prompt & Result Record
- Model
- MiniMax H3
- Workflow
- Text to Video
- Resolution
- 720p
- Duration
- 15 seconds
- Aspect ratio
- 16:9
- Generation date
- Not recorded
- Attempts
- Not recorded
- Post-editing
- Not recorded
- Platform
- MiniMax3.org
Original generation prompt
Create a humid jungle survival sequence centered on reliable subject tracking. Follow an expedition team through dense foliage, reveal movement in nearby water, then introduce a large jungle predator from the shadows. Track the action through a fast strike, evasive movement, mud, spray, and displaced equipment. Preserve the identity and position of every character while keeping the animal's scale and movement physically coherent.
Output observations and failure notes require human video review and are not inferred from the prompt or thumbnail.
Are These Real MiniMax H3 Outputs?
Yes. These are generated MiniMax H3 outputs documented by MiniMax3.org. Each entry includes the original prompt and available generation metadata. Missing historical fields are labeled rather than reconstructed from memory.
A successful example demonstrates that H3 produced that result in at least one generation. It does not prove the same prompt will succeed every time.
Want prompts to copy?
The Examples Library records what actually happened. The Prompt Library owns reusable templates.
Browse reusable MiniMax H3 promptsWant prompting methodology?
The Prompt Guide explains how to structure dialogue, camera direction, references, timing, and audio.
Read the MiniMax H3 Prompt GuideMiniMax H3 Example FAQ
Are these real MiniMax H3 videos?
Yes. The library contains generated MiniMax H3 outputs documented by MiniMax3.org. Metadata completeness varies and is labeled transparently.
Can I copy the prompts?
Yes. You can adapt the original prompts, but AI video generation is probabilistic and the same prompt may produce a different result.
Do these examples prove MiniMax H3 always produces this quality?
No. Each example represents one generated result or documented run, not a universal quality or success-rate measurement.
Where can I find reusable H3 prompt templates?
Use the MiniMax H3 Prompt Library for reusable templates. This library focuses on generated outputs and evidence.
Where can I learn how the prompts work?
Use the MiniMax H3 Prompt Guide for prompt structure, reference roles, timing, camera direction, dialogue, and troubleshooting.