How to Generate “Weight” in AI Video
- Noemi Kaminski
- Apr 26
- 4 min read

Most AI video looks good at a glance.
Then something feels off.
Not the lighting. Not the detail. It’s the weight.
Movement doesn’t carry force.Impacts don’t transfer energy.Everything just kind of… glides.
So I started testing prompts specifically designed to break that illusion—heavy motion, particle interaction, slow motion, close-ups.
Here’s what actually worked.
Tip 1: Force the model to simulate impact, not just motion
If your prompt only says “running” or “galloping,” the model will fake it.
You need to describe what happens when mass hits something.
Prompt:
A hyperrealistic slow-motion close-up of a powerful warhorse’s chest and shoulders driving forward mid-stride. Massive muscles contract and roll beneath taut skin, veins subtly visible as the force of motion ripples across the body. Dust and debris burst upward from just out of frame, drifting past the camera in ultra-detailed particle simulation—fine grains catching warm sunlight as they spin and collide. The base of the neck flexes with each stride, showing layered muscle movement and natural skin deformation.
The rider’s armored thighs and worn leather saddle gear move with realistic inertia, straps tightening and loosening with each impact. The camera is positioned low but angled upward toward the torso, tracking alongside the horse with shallow depth of field—sharp focus on muscle detail, background cavalry blurred into motion.
Warm directional sunlight grazes across the horse’s body, highlighting sweat, dust, and texture while casting soft shadows into the contours of muscle. Motion blur remains subtle and physically accurate, emphasizing weight and forward momentum.
Breakdown:
“driving forward mid-stride” → implies force, not just movement
“muscles contract and roll” → forces anatomical realism
“dust and debris burst upward” → adds cause-and-effect
“particle simulation… spin and collide” → stops particles from feeling fake
“straps tightening and loosening” → introduces secondary motion (key for realism)
Most models skip all of this unless you explicitly demand it.
Tip 2: Separate heavy mass from light elements
Realism isn’t just weight—it’s contrast.
Heavy things should feel heavy.Light things should feel reactive.
Prompt:
A hyperrealistic slow-motion shot focused on the upper neck and flowing mane of a galloping warhorse. The powerful neck arches and flexes with each stride, muscles shifting naturally beneath dust-coated skin. Strands of the mane whip through the air in layered motion, catching light and casting fine shadows across the horse’s body.
Particles of dirt and dust drift upward from below, moving through the frame in detailed slow motion, interacting with airflow around the horse. The rider’s hands and reins enter frame partially, leather straps pulling taut and relaxing with the horse’s movement.
Camera tracks tightly alongside the neck at a low angle, maintaining shallow depth of field—mane strands and skin texture in sharp focus while the background dissolves into motion blur. Warm sunlight glances across the scene, creating rim lighting along the mane and subtle highlights across dust particles.
Breakdown:
Neck = mass → controlled, powerful movement
Mane = light → loose, reactive, layered motion
This contrast is what makes:
Hair feel bouncy and airy
Body feel grounded and heavy
If everything moves the same way, the illusion breaks instantly.
Tip 3: Control the camera to reinforce weight
Camera placement is doing more work than you think.
Low angles + tracking = perceived mass
Prompt:
A hyperrealistic ultra-low angle shot skimming just above the ground, angled upward toward the horse’s torso and rider as it charges past. The horse’s body dominates the frame, its mass passing overhead with immense presence. Dirt and debris fly across the foreground, occasionally crossing the lens.
The camera remains focused on the underside of the torso and chest, capturing shifting muscle tension and the interplay of light and shadow across the body. The rider’s silhouette looms above, partially framed against the sky.
Foreground particles are razor sharp while the horse moves through shallow depth of field, creating layered realism. Sunlight flares slightly through dust clouds.
Breakdown:
“skimming just above the ground” → puts the viewer under the mass
“body dominates the frame” → removes visual escape → feels heavier
foreground particles crossing lens → adds depth + realism
slight camera imperfection (micro shake / flare) → breaks “AI smoothness”
This is why cinematic shots feel heavier than flat ones.
Tip 4: Add inertia and delayed reactions everywhere
Nothing in real life moves instantly.
Everything lags, catches up, and settles.
Prompt:
A hyperrealistic three-quarter angle shot of a galloping warhorse, capturing both forward motion and body depth. The camera is positioned slightly ahead and to the side, tracking backward as the horse moves into frame. The chest, neck, and head are all visible, creating a powerful, balanced composition.
Muscles flex dynamically beneath the skin, the mane flowing backward in layered strands. Dust trails behind and wraps around the body, interacting naturally with movement. The rider is clearly visible, armor shifting subtly with inertia, leather straps lagging slightly behind each stride before snapping back into place.
Lighting is warm and directional, emphasizing contours and surface detail. Depth of field isolates the horse while the environment softens into cinematic blur.
Breakdown:
“lagging slightly… before snapping back” → this is inertia
dust trailing, not just appearing → reinforces motion direction
camera tracking backward → adds resistance (feels like force is chasing camera)
Without inertia, everything feels weightless—even if it looks detailed.
Final Thought
Most people are trying to make AI video look better.
But realism doesn’t come from prettier frames—it comes from movement that holds up under pressure.
If you focus on:
impact
inertia
particle interaction
and contrast between heavy vs light
…your results jump instantly.



Comments