Figure 1. Weight and momentum become readable through preparation, travel, contact, and response.
Motion feels real when a visible force creates a matching response. Describe preparation, acceleration, contact, recoil, secondary motion, and rest in order. Effective AI physics prompts show where weight shifts, what absorbs the impact, and how momentum ends. That detail prevents floating objects, weak collisions, instant stops, and weightless body movement.
Why Does AI Motion Often Look Weightless?
Many prompts name only the action: “a person lifts a box” or “a ball hits a wall.” The model must invent the speed, mass, effort, path, contact, and result. It may create an attractive action that lacks physical evidence.
Weight is visible through resistance and delay. A heavy box changes posture before it moves. Momentum is visible through continued travel. A runner needs several steps to stop. Impact is visible through compression, recoil, vibration, sound, dust, or movement in the receiving object.
Research confirms that this remains difficult. The VideoPhy-2 benchmark found that video models especially struggle with conservation laws such as mass and momentum. Your prompt should therefore state the event chain rather than assume the model will infer it.
What Is the Simplest Force-First Prompt Structure?
Use six ordered parts:
- 1
- Subject and physical traits: size, rigidity, balance, or apparent mass. 2
- Applied force: push, pull, gravity, kick, throw, motor, or wind. 3
- Acceleration and path: how speed changes and where the subject travels. 4
- Contact: the exact point and moment of collision. 5
- Response: recoil, compression, rotation, vibration, dust, or surface movement. 6
- Final state: where everything stops and whether it settles or keeps moving.
A heavy wooden crate rests flat on concrete. A worker braces both feet and pushes its left side with both hands. The crate resists at first, then slides slowly 40 centimeters to the right. Friction keeps it level and produces a short scrape. The worker leans forward, takes one recovery step, and stops as the crate settles. Locked side camera, no floating, no sudden speed change, no floor penetration.
This is ready to test in an AI video generator because every stage has visible evidence.
How Do You Show That an Object Is Heavy?
Figure 2. A heavy lift shows a braced stance, slow rise, close load position, and controlled set-down.
Show preparation, slow acceleration, limited travel, and recovery. The word “heavy” helps, but body and surface reactions make it readable.
For a person lifting something, include:
- Feet spread for balance.
- Knees and hips bend before the lift.
- Hands establish a firm grip.
- The load rises slowly near the body.
- Shoulders and torso respond to resistance.
- The person takes a small balance step if needed.
- The load lands and the body relaxes after release.
Copy-Ready Heavy Box Prompt
Full-body side view of a delivery worker lifting one large sealed cardboard box from the floor onto a waist-high table. The worker plants both feet shoulder-width apart, squats with a straight back, grips the lower corners, and lifts slowly with visible effort. The box stays rigid and close to the torso. The worker takes one short step, lowers the box onto the table, releases both hands, and straightens. The table vibrates slightly at contact. Locked camera, natural speed, no floating box, no bent edges, no sliding hands, no sudden teleport.
Keep the apparent mass consistent. A large box should not leap upward from a fingertip unless the scene is intentionally superhuman.
How Do You Show That an Object Is Light?
Use quick acceleration, small effort, and a stronger response to air or contact. A tissue, paper cup, or balloon can change direction easily.
One empty paper cup rests on a smooth desk. A brief fan gust pushes it from left to right. The cup accelerates quickly, tips onto its side, rolls in a shallow curve, taps a closed notebook, rebounds a few centimeters, and stops. The notebook does not move. Locked overhead camera, soft office light, no heavy impact, no cup deformation beyond a slight rim flex.
The stationary notebook provides a mass comparison. The cup moves while the heavier object barely responds.
How Do You Prompt Acceleration and Braking?
Describe speed as a change over time. “Fast car” defines a state, while “accelerates from rest, reaches steady speed, then brakes over two car lengths” defines motion.
Copy-Ready Car Braking Prompt
Low side tracking shot of a dark blue sedan moving from left to right on a dry test road. The car begins at moderate speed, brakes firmly at a marked cone, and slows over two car lengths. The nose dips slightly, the front suspension compresses, the wheels keep aligned, and the body settles once the car stops. The camera tracks at matching speed, then eases to a stop. No instant freeze, no sideways drift, no tire deformation, no changing car shape.
Use Seedance after you have fixed the vehicle, road direction, braking point, and final position. A reference frame with clear wheel and body geometry gives the motion a stronger anchor.
How Do You Prompt a Throw With Momentum?
Figure 3. Braking and throwing both need visible force, continued travel, and a final response.
Include the wind-up, release, flight arc, follow-through, and landing. The hand should release before the object travels freely.
Copy-Ready Ball Throw Prompt
Medium-wide side view of an athlete throwing one red medicine ball toward a padded wall. The athlete steps forward with the left foot, rotates the torso, extends both arms, and releases the ball at chest height. The ball travels in a shallow forward arc, strikes the center pad, compresses slightly, rebounds toward the floor, and rolls to a stop. The athlete's arms continue forward after release, then lower. Locked camera, natural motion blur on the ball only, no ball duplication, no release after impact, no floating rebound.
The follow-through proves that momentum existed before the collision. The rebound and roll show where it went afterward.
How Do You Prompt a Fall Under Gravity?
State the starting support, loss of support, downward acceleration, impact, and settling. Avoid an unexplained midair start if the cause matters.
Copy-Ready Product Drop Prompt
Macro side view of one closed matte-white earbud case resting at the edge of a 30-centimeter platform. A small mechanical arm nudges the case outward. The case tips past the edge, falls under gravity without changing shape, rotates one-quarter turn, and lands on a gray rubber mat. The mat compresses slightly, the case bounces once at low height, slides five centimeters, and stops. Static camera, crisp case silhouette, no levitation, no extra bounce, no parts opening.
If product damage is unacceptable, use a soft surface and a low drop. For a polished AI ad video, the impact should support the message without making the product look unsafe.
How Do You Prompt a Collision Between Two Objects?
Give both objects an initial state and describe momentum transfer after contact. One should not react before the other arrives.
Copy-Ready Billiard Ball Prompt
Locked overhead view of a white cue ball rolling in a straight line toward a stationary blue ball on green felt. The cue ball contacts the blue ball at its center. At contact, the cue ball slows sharply while the blue ball begins moving forward along the same line. Both balls keep their size, color, and round shape. Felt friction gradually slows the blue ball until it stops. No movement before contact, no overlap, no duplicate balls, no change in direction without collision.
Avoid describing a perfect scientific demonstration unless you need one. The goal is visible causality: approach, contact, transfer, and stop.
How Do You Prompt a Person Landing From a Jump?
Figure 4. Different masses and materials produce different contact, rebound, and absorption patterns.
Show that the ground absorbs downward momentum through ankles, knees, hips, and a short recovery. A rigid upright landing looks weightless or unsafe.
Copy-Ready Jump Landing Prompt
Full-body front three-quarter view of a basketball player dropping from a low box onto a gym floor. Both feet contact the floor together. The ankles flex, knees bend to about 60 degrees, hips move back, and the torso leans slightly forward to absorb the landing. The arms swing forward for balance. The shoes compress against the floor, then the player rises to a stable stance. Locked camera, natural speed, no floor penetration, no stiff knees, no extra jump.
Use a wide enough frame to keep the feet visible. A cropped landing hides the strongest evidence of weight.
How Do You Prompt Recoil?
Recoil should follow a brief force and move in the opposite direction. Its size depends on the subject, support, and event.
For safe everyday examples, use a spring, sports equipment, or a tool on a test rig:
Close side view of a tennis racket striking one yellow ball. The ball compresses against the strings for a brief moment, then leaves forward at high speed. The racket head moves back a few centimeters after contact, the strings vibrate, and the player's wrist stabilizes the handle. The ball follows a clear rising arc. Static camera, high-speed visual treatment, no contact before the swing, no ball passing through the strings.
Do not exaggerate recoil until the object loses its expected mass. A heavy supported machine should respond less than a handheld racket.
How Do You Prompt Soft and Hard Impacts Differently?
A hard surface produces short contact, little deformation, sharper sound, and more rebound. A soft surface produces longer contact, visible compression, quieter impact, and less rebound.
Hard Impact Prompt
One steel ball falls onto a thick steel plate. The ball makes brief centered contact, rebounds to one-third of its drop height, and falls again with a smaller bounce. The plate remains fixed but vibrates subtly. Crisp metallic click at each contact, locked macro camera, no dent, no sparks, no sideways jump.
Soft Impact Prompt
One foam practice ball falls onto a folded cotton towel. The towel compresses and wrinkles under the ball. The ball deforms slightly, makes almost no rebound, rolls a few centimeters into a fold, and stops. Quiet fabric thump, locked close camera, soft daylight, no hard metallic response.
Material and reaction must agree. A towel should not launch a heavy object high into the air without an added spring.
How Do You Show Momentum in Running?
Use forward lean during acceleration, repeated ground contact, arm opposition, and several deceleration steps. A person cannot move at full speed and stop on one planted foot without a visible force.
Side tracking shot of a runner accelerating along a marked track. The first steps are short and forceful with a forward body lean. Strides lengthen as speed increases. At the final marker, the runner gradually straightens, shortens each stride, takes four deceleration steps, and comes to a balanced stop. Clothing continues moving briefly after the body stops. Smooth tracking camera, feet contact the track, no sliding, no instant stop.
Secondary motion in fabric and hair should lag behind the main body, then settle later. This delay helps sell momentum without adding chaos.
How Do You Prompt a Push or Pull?
Show the force connection. Hands, rope, handle, feet, and the moved object must stay linked during the effort.
Copy-Ready Furniture Push Prompt
Wide side view of one person pushing a heavy wooden table across a smooth studio floor. Both palms stay flat against the near edge. The person leans forward, braces one foot behind, and increases force. The table resists, then slides slowly in a straight line. Its four legs remain on the floor and the tabletop stays level. The person takes two controlled steps and stops pushing. Friction brings the table to rest. Locked camera, no floating legs, no hand separation, no sudden acceleration.
For a pull, define rope tension or handle position. A rope should become taut before the load moves.
How Should the Environment React to Impact?
Use one or two secondary reactions that match the scale of the event. A footstep can compress grass and lift a little dust. A heavy crate landing can vibrate a table. A fast ball can flex a net.
Useful reactions include:
- Dust rising from a dry surface after contact.
- Water or mud moving away from a footfall.
- Fabric, hair, or straps continuing after the body stops.
- A net stretching in the direction of the ball.
- A cushion compressing under a load.
- Loose objects rattling on a struck table.
- A shadow staying attached to the moving subject.
Keep the response proportional. A small dropped key should not shake the whole room.
How Do You Keep the Camera From Hiding Physics?
Choose an angle that shows the force path and contact point. A locked side view is best for many tests because it reveals distance, acceleration, and landing. An overhead view works for planar collisions. A front view can show compression but may hide travel depth.
If the camera tracks, match one moving subject and keep the horizon stable. Avoid a fast orbit during a collision. The changing viewpoint can hide whether objects touch or pass through each other.
You can prepare a simple, high-contrast setting with an AI background generator. Clear floor lines, wall edges, mats, and markers make distance and scale easier to judge.
What Should You Fix When Physics Looks Wrong?
Name the missing cause or response and change one instruction.
Do not solve a weak collision by adding more cinematic adjectives. First make the physical chain readable.
Which Frames Prove That the Motion Works?
Figure 5. Reviewing approach, contact, recoil, and rest exposes broken event order quickly.
Review five moments: preparation, maximum speed, first contact, largest response, and final rest. A plausible clip should connect them without unexplained jumps.
Ask:
- Is the force visible before acceleration?
- Does the path remain continuous?
- Does contact occur before recoil or deformation?
- Does the larger mass respond less than the smaller mass?
- Does secondary motion continue briefly after the main motion?
- Does friction, balance, or another force explain the stop?
- Do people, products, and surfaces keep their identity and shape?
Approved physical shots can then be arranged in an AI film where cuts preserve the event rather than hide errors.
Make Every Motion Earn Its Result
Believable action connects force to consequence. Define the subject's apparent mass, applied force, acceleration, contact, response, and final state. Keep the camera on the physical event and add only proportional secondary motion. Use these AI physics prompts in Pippit to replace weightless movement with action the viewer can understand.
FAQs
Q1. How Do I Make an AI-Generated Object Look Heavy?
Show resistance before movement, slow acceleration, limited travel, and a visible surface response. If a person moves it, include a firm grip, bent knees, braced feet, and a recovery motion. The word “heavy” helps, but physical evidence makes the weight believable.
Q2. What Is Momentum in an AI Video Prompt?
Momentum is continued motion that must be redirected or reduced by another force. In a prompt, show the subject accelerating, traveling along a clear path, and slowing through friction, braking, collision, or recovery steps. Avoid instant stops unless a visible barrier causes them.
Q3. How Can I Prompt a Realistic Collision?
Give both objects a clear starting state, show approach, name the exact contact point, and describe what each object does after contact. Add one proportional response such as rebound, compression, rotation, vibration, or transferred movement. Keep the camera fixed so contact remains visible.
Q4. Why Does an AI Character Float During a Jump?
The prompt may omit takeoff force, gravity, or landing response. State that the feet push against the ground, the body rises and slows, then accelerates downward. On landing, show ankle, knee, and hip flexion. Keep the full body and ground inside the frame.
Q5. Should I Put Physics Equations in a Video Prompt?
Usually no. Observable motion is easier to direct than equations. Describe how speed changes, where contact occurs, how surfaces deform, and how the subject stops. Simple measurements such as distance, direction, bounce height, or step count can help when they are visible in the shot.
Q6. How Many Physical Events Should One Clip Contain?
One main event is safest. A lift, throw, collision, or landing already contains several connected stages. When a prompt also adds a second collision, camera orbit, transformation, and scene change, causality becomes harder to preserve. Generate separate clips and combine approved results during editing.