Live workflow

MiniMax H3 - Motion Context

Guide one continuous shot with three reference images pinned to exact moments, and get video with matching audio.

Video Audio Image to video 3 references

1 Reference images 3/3 in use

Three fixed moments in the clip — start, middle and end.

0:00.04

0:05.08

0:10.13

Click a slot to upload a replacement, drag a file from your computer onto it, or drag one reference onto another to swap them. All three slots are required to run.

2 Shot description

Describe what happens, how the camera moves, what carries over between scenes, and what it sounds like — all in one block of prose. The references anchor the visuals; this sets everything between them.

Paste it into an assistant, describe your shot, upload your images, and it writes the prompt below.

Your prompt4,100

integrated_multimodal_description:
A single continuous cinematic shot with no cuts, rendered as a 2.5D painterly illustration: flat painted layers sliding in parallax at different rates, visible brushstrokes and ink grain that subtly boil as if each frame were repainted, dry-brush edges that smear in the direction of motion, and a palette locked to black ink, crimson red, and pale parchment. Depth reads as stacked painted planes, not photographic space. The painting never freezes: every held composition keeps living intrinsic motion — wind, flame, water, breath, falling debris — even while the camera lingers.

At the beginning, a warrior in a torn-crested helmet stands in the right foreground before a burning gate in a great city wall, red light pooling on his cheek guards. He plants his spear into the churned earth and turns his head to survey the fire, his crest and cloak whipping in the hot wind while the wreck behind him spits slow arcs of sparks and the smoke column leans and rolls up the left third of frame. Ash drifts between the painted planes at different speeds.

As the shot progresses toward the midpoint, the camera begins a long unhurried crane upward along the smoke column; the warrior lowers out of the bottom edge to join a line of black helmeted silhouettes, and the smoke thins and parts to reveal what it was hiding: the colossal dark wooden horse towering between two ruined towers.

The horse holds the frame and works: it rocks forward in slow heavy heaves as unseen ropes haul it toward the gate, timbers groaning and flexing, wheels grinding over rubble, black banners snapping below, rows of spear tips swaying in loose rhythm, ember flecks orbiting it on separate parallax planes. Give the labor one full heavy heave of the great body before the shot moves on.

In the final third, the camera pushes into the pitch darkness between the horse's plank-built legs; the black plane floods the frame like spilled ink, thin red brushstrokes bloom downward through it, and firelight flickers on to reveal that the vertical timbers have become vertical folds of marble: a cracked goddess statue seen from below against the darkness, red stains running and lengthening down her robes, small flames climbing her raised hand and staff.

In the closing seconds the statue's destruction plays out: a jagged fracture line spreads across her neck with trickling dust, chips of marble drop away and tumble toward the camera, the crack widens with each pulse of firelight — and her carved head breaks free, tips forward, and falls, tumbling slowly end over end past the camera and out of the bottom of frame, trailing dust, while the headless figure stands burning against the black. Hold on the decapitated silhouette, flames climbing higher, as the shot ends.

Camera:
One continuous evolving move with long dwells: a slow near-static drift on the warrior, a patient crane up the smoke column to the horse, a hold on the horse's labor, a push into the darkness beneath it, and a slow upward drift on the statue through the fall of the head. Every move is smooth and painterly, with foreground, midground, and background planes sliding at visibly different parallax rates.

Continuity:
No cuts, no teleportation, no sudden subject replacement, and no extra characters unless explicitly requested. Each transition is carried by an anchor: the smoke column concealing then revealing the horse, the horse's darkness revealing the statue. The progression should read clearly as:
[WARRIOR BEFORE THE BURNING GATE] -> [THE GREAT WOODEN HORSE HAULED FORWARD] -> [THE GODDESS STATUE, HEAD BREAKING FREE AND FALLING].

overall_soundscape:
Crackling fire, hot wind, and the crunch of the spear planting; swelling war drums and groaning timber as the crane rises; the horse's slow rhythmic heaves — rope strain, wheel grind, banner snap; the drums cut to near-silence in the darkness beneath it; then licking flame, the sharp spreading crack of stone, dust hissing down the marble, and the heavy tumbling rush of the falling head ending in silence and fire.

non_diegetic_music:
N/A

3 Output format

How long the finished video runs and what shape it comes out. Longer clips give the references more room to breathe between beats.

Clip length10s

Aspect ratio

Resolution

0.4 MP · 840×472 at this aspect ratio. Higher costs more time.

Fitting references

How images that do not match the aspect ratio are handled.

Seed

Same seed and settings give the same video back.