Motion Generation in Seedance

A director doesn’t guesses movement— it observes, measures, sequences, and tests. That sentence sums up the entire lesson. Seedance, the video generator for this stage, was designed to respond to physics, not adjectives: it understands force, time, and weight much better than it understands words like “epic” or “dramatic.”1 Asking for “an incredible movement” produces something generic; describing what moves, how strongly, in what order, and over how much time brings back real movement.

Every movement rests on three elements. Force — what drives the action (impact, gravity, push, jump). Time — how long each action lasts. Weight — how objects behave physically. When you provide these three, the generator has what it needs to produce something the eye accepts as real. When you provide adjectives, it fills in the gaps on its own — and the result floats, slips, and feels weightless.

Before writing: five questions for a director

Before writing a scene, the source suggests defining five things: camera (where it is, how it moves), force (explosion, gravity, shove, jump), sequence (what happens first → what happens last), delays (time offsets between actions) and perception (what the viewer experiences). The synthesis is simple and powerful: movement is a controlled sequence of events. Not a single effect—a chain ordered over time.

Every strong prompt begins with force: what starts the movement? It could be a impact, a acceleration, a gravity or a interaction. Force is the trigger—without it, the scene has no cause, and movement without a cause looks like drifting. Defining the force first is what anchors everything that follows.

Next comes the sequence: the order of events over time. Main action first, secondary movement afterward, residual effects last. The source’s golden rule is brief and applies to almost anything that explodes, falls, or collides: large → small → particles (dust, smoke, debris). This order isn’t aesthetic—it’s physics. In the real world, the large mass moves first and pushes the smaller ones; the particles are the last link in the chain. Reversing this (dust before impact) is what makes a scene feel “wrong” without anyone knowing why.2

A movement prompt built around force and sequence, ready for Seedance—notice how each action has an order and timing:

A camera defines perception — it decides what the viewer experiences of the event. Three controls: position (low, wide, close-up), movement (pan, tilt, dolly, orbit, shake) and behavior (stable vs. handheld). In the lesson examples, the camera appears on a helmet (helmet cam), tracking shots (tracking), drone follow, continuous orbit, and constant pullback—each placing the viewer in a different position in relation to the action.3

A human reaction is the fourth pillar, and perhaps the most subtle. Emotion is shown physically: breathing, micro-movements, gaze focus, body tension. The rule is absolute: never state the emotion—show it through movement. Instead of asking “he is afraid,” describe the fingers slipping, the eyes widening, the breathing changing. The generator doesn’t know how to interpret “fear”; it knows how to animate slipping fingers. Translating emotion into the body is what makes the performance believable.

A prompt that combines camera behavior and human reaction— the emotion appears only through the body, never named:

O time is what creates realism. The source is specific: think of real-time action, in 30% slow motion when a moment needs weight, and in precise durations — 0.3s, 1.2s — instead of “fast” or “slow.” Numbers give the generator a target; timing adjectives leave everything adrift. Measuring time means treating movement like engineering, not wishful thinking.

O secondary movement is what separates the amateur from the professional. Everything remains after from the main action: dust lingers, smoke leaves a trail, water hangs in the air before falling. These residual layers create depth and realism — they’re proof that the world has inertia. An explosion that ends abruptly, with nothing continuing, feels like a cut; one that leaves dust settling for two seconds feels real. The secret, once again, is layers over time: primary, secondary, residual.4

A prompt focused on timing control and secondary movement, with measured durations—ready for Seedance:

The source's key rules bring the idea together: no vague descriptions; use measurable details (time, distance, angle); focus on observable actions; always think in sequences; always define camera behavior. In one sentence: believable, measured movement, linked together and shown—never described with adjectives.

The final lesson ties it all together: you don’t need to describe every detail of the scene. You only need four pieces — force, sequence, camera, and human reaction — and Seedance turns that into movement. The rest is noise. That's the practical relief of the method: instead of trying to predict every frame, you define the four levers and let the model's physics fill in the space between them.

In this frame, the emotion is never stated—it’s staged through movement and timing. A character isn’t “exhausted”: their shoulders droop, their steps drag, their breathing slows. The camera isn’t “tense”: it shakes slightly and moves closer slowly. Time isn’t “dramatic”: it slows down 30% at the moment that matters. When you direct these levers, emotion emerges from the physical details—and the viewer feels it without you having to name it.

The assignment—and the bridge to the next lesson

The source assignment is straightforward: build your own scene using the method, including force, sequence, camera, and human reaction—and then explain why that speed, why that camera, and what the viewer feels (without naming emotions). Show everything through movement and timing. Sum up the module in one sentence: Seedance reads physics, not adjectives— from force, sequence, camera, and body, with measured timing, comes movement that feels real. In the next lesson, we shift the focus from the physics of movement to the camera: fine control over how it moves, in Kling.

Notice how the method connects with the scene algorithm you saw at the beginning of the course: force triggers the rupture, the sequence controls the escalation, timing (with 30% slow motion) marks the peak, and human reaction is what makes the impact matter. The tools change; the grammar of a powerful moment stays the same—you’re just writing it now in a language the motion generator understands.