MINIMAX H3 · AI VIDEO GENERATOR
Describe a shot or add frames, images, video and audio references to create short videos with native sound. Develop product showcases and visual stories with direction for both picture and audio.
Review results, manage assets and access generation history in AI Studio.
Watch videos from existing generation tasks and inspect their prompts and settings. Choose Use this example to load the same setup and reference materials in the studio.
Selected examples demonstrate workflows; repeated inputs may yield different details. WebM previews are capped at 1080p, while MP4 retains the original resolution.
Spaces & camera work
A route reference guides a flight through Istanbul, with timed directions for a low pass over water, a climb through the city and an orbit around a landmark.
Generation settings: 768P · 9:16 · 10s
Use this exampleProducts & brands
A coffee-making sequence uses close shots and separate directions for steam, liquid motion, latte art and ambient sound.
Generation settings: 2K · 9:16 · 10s
Use this exampleSpaces & camera work
An architectural showcase uses the building and landscape as spatial anchors, with directions for viewing order, lighting and camera movement.
Generation settings: 2K · 9:16 · 10s
Use this exampleVisual storytelling
A miniature world provides the setting for a visual story, with scale, environmental detail and shot pacing shaping its atmosphere.
Generation settings: 768P · 9:16 · 10s
Use this exampleProducts & brands
A locked camera follows an empty room becoming a bedroom, with furniture reveal, assembly and the final detail assigned to distinct time windows.
Generation settings: 768P · 9:16 · 10s
Use this exampleProducts & brands
The camera enters a steampunk miniature city through a laptop screen and returns to the desk, with the prompt directing the entry, journey and ending.
Generation settings: 2K · 9:16 · 10s
Use this exampleStart with one clear shot objective, then specify action, timing and sound. These original templates apply principles from the official guides; they have not been individually generation-tested.
Show a product’s shape, materials and details with one subject and one camera path.
Prompt template
10-second landscape shot. A matte white wireless-earbud case sits on a light-gray pedestal. The lid slowly opens as the camera makes a smooth quarter-orbit around the right side. Keep proportions consistent. Soft side lighting reveals the material. Only a subtle lid-opening sound; no music or captions.
Use macro framing, droplets and side light to explore materials; review brand labels separately.
Prompt template
5-second portrait shot. A frosted-glass serum bottle stands on dark stone. One clear droplet slowly slides down the bottle. Locked macro camera, narrow side light and a dark-gray background. A faint droplet sound only. No text or people.
Build the clip around one continuous action and distinguish preparation sounds from ambience.
Prompt template
10-second portrait shot. A close camera watches a barista pour warm milk into a latte, forming a simple heart on the surface. Keep the camera fixed and focus on the liquid under warm morning light. Milk pouring, a soft cup clink and distant café ambience; no dialogue or music.
Separate the introduction, opening and detail hold for a compact product reveal.
Prompt template
10-second square shot. An unbranded navy gift box rests on a pale table. 0–3s: slowly move closer to the closed box. 3–7s: two hands lift the lid to reveal a watch in velvet lining. 7–10s: hold on the dial. Natural side light, cardboard and fabric sounds, no text.
Use doors, windows and furniture as spatial anchors; avoid camera paths through solid objects.
Prompt template
10-second landscape shot. Slowly move forward from an open living-room entrance. A linen sofa is on the left, a wooden bookcase on the right and a full-height window faces a garden ahead. Follow the clear walkway and gently stop near the window. Afternoon light and faint birdsong; no narration.
Anchor the camera to a clear landmark and distance for an architectural concept shot.
Prompt template
10-second 16:9 shot. A white modern villa faces a bay. The camera moves slowly to the right above the front lawn, keeping the entire building and horizon in view. Warm evening light and softly lit windows. Sea breeze and distant waves; no people or text.
Choose a readable camera path and use environmental sound to establish place.
Prompt template
10-second landscape shot over a mountain lake at dawn. Move slowly just above the still water, pass a moored wooden boat, then gently rise to reveal distant snowy peaks. Thin mist, cool sky and golden highlights on the mountains. Water and birdsong; no narration or captions.
Check appearance continuity with a simple action before expanding the story.
Prompt template
5-second landscape shot. A small robot in an orange raincoat looks down at a puddle, hops over it and turns toward the camera. Fixed full-body framing. Keep the raincoat and body proportions consistent. Light rain and a soft landing sound; no dialogue.
Explore environmental layers and motion pacing as a concept clip, not a playable scene.
Prompt template
10-second 21:9 shot. A forgotten stone gate stands in a forest, moss covering its steps and pale blue light escaping through a narrow opening. Slowly push the camera forward as leaves sway and low fog drifts. Deep wind and a distant bell; no people, captions or interface.
Explore an art direction with a simple character action and stable framing.
Prompt template
10-second landscape shot in a soft watercolor picture-book style. A small fox wearing a red scarf walks beneath an apple tree, watches a leaf settle on its nose, then blinks. Fixed medium shot in warm sunset light. A breeze and rustling leaves; no dialogue or text.
Name the speaker and keep the line short; review timing and pronunciation in the output.
Prompt template
10-second landscape shot. A baker behind a bright shop counter gently places a paper bag on the surface, looks up and smiles: “Fresh from the oven. Have a good day.” Stable medium-close framing. Quiet shop ambience; no narration, music or captions.
Make one small movement the visual focus for a calm motion study.
Prompt template
5-second portrait macro shot of a green leaf at dawn. A droplet on its tip grows and falls, and the leaf gently rebounds as the weight is released. Locked camera, soft background blur and clear morning light. A delicate water sound and distant birds; no music.
These are the options currently available on ImgAI. Duration, resolution and references determine the setup for each task.
Adaptive sizing is for frame or reference workflows. Choose a fixed aspect ratio for text-only generation.
| Resolution | 5s | 10s | 15s |
|---|---|---|---|
| 768P | 408 | 816 | 1,224 |
| 2K | 654 | 1,307 | 1,960 |
Calculated from the current catalog, with native audio included. Video references add their duration to the calculation; check the generate button for the final credit amount.
View credit plansAdapted from MiniMax’s official text, image and reference prompting guides. Give each instruction one purpose: establish the scene, then describe what changes.
Identify who or what is in the scene, where it is and what it does. Add light, materials and visual style after the action is clear.
Prompt template
A ceramic cup rests on a wooden table beside a window. Hot tea pours in and steam rises toward the light on the left. Hold a close shot.
Build the clip around one main action. Use time windows when the order matters, and avoid crowding a few seconds with many locations, characters and transitions.
Prompt template
0–3s: show the closed gift box. 3–7s: the lid slowly opens. 7–10s: hold on the product and lining details.
The image establishes the subject and composition. Focus on motion, camera work and details to preserve. Start and end frames should allow a coherent transition.
Prompt template
Preserve the cup’s shape, color and position from the first frame. Steam rises gently as the camera slowly moves closer. Keep the background arrangement unchanged.
Use the editor’s @ reference picker to identify each uploaded asset. Specify whether it supplies appearance, movement, camera direction or sound, and avoid conflicting instructions.
Prompt template
Use the image for product appearance, the video only for the camera orbit, and the audio for background rhythm. Preserve the product’s color and proportions.
Specify the speaker and exact line when dialogue is needed. Describe ambience and music separately, and state clearly when no music or narration is wanted.
Prompt template
The baker softly says, “Fresh from the oven.” Keep paper-bag rustling and quiet shop ambience. No narration or music.
Check labels, on-screen text, hands, contact between objects and dialogue timing. Add exact brand text or captions in post-production when reliable typography is essential.
Prompt template
Check the bottle outline and hand motion first, then review label spelling. Lay out the final promotional text in the editor.
MiniMax H3 is a multimodal generation model that accepts text, images, video and audio references and produces video with native stereo sound. ImgAI provides text, frame and reference workflows.
Choose 5, 10 or 15 seconds at 768P or 2K. Available aspect ratios include landscape, portrait, square and widescreen. The controls in the generator show the current options.
Model capabilities and platform settings can differ. ImgAI currently accepts up to 5 reference images, 3 reference videos and 3 reference audio files. Video and audio references each have a total duration limit of 15 seconds.
H3 outputs include native audio, and ImgAI does not provide a separate audio toggle for this model. Describe dialogue, music or ambience in the prompt. Remove the audio track in post-production if you need a fully silent file.
The videos come from existing MiniMax H3 tasks on ImgAI, with their prompts and reusable settings. The 12 scenario prompts are separately written creative templates, not records of tested outputs.
Credits depend on output duration and resolution. Reference-video duration also contributes to the cost. The table shows combinations without video references; the generate button displays the final amount.
They are different variants. This page describes and runs MiniMax H3. H3 Max speed or specification claims should not be applied to it; check the full model name in the generator.
No. Complex movement, precise text, subject continuity and audio timing still need review. Test the shot with a shorter clip, then refine the prompt and references to keep revisions manageable.
Choose an existing example or describe the subject, motion and sound, then set the duration and resolution.