Seedance

How to Prompt Seedance

Separate your primary and secondary motion, the key to cinematic quality.

Updated March 2026 · Official documentation + community tutorial sites, limited independent Reddit/Discord data

Generate from text. Use the 6-part framework and separate primary subject motion from secondary environmental motion.

🟡 Community pattern

The 6-Part + Layered Motion Framework

Seedance responds to a clear 6-part structure. The key technique: describe primary subject motion and secondary environmental motion as separate layers. This prevents the model from confusing what the subject does with what the environment does.

Subject + Action

A chef in a white apron skillfully flips a steak in a cast-iron skillet, flames rising briefly

💡 Lead with the primary action, what the subject is doing.

Visual Style

Photorealistic

💡 Specify style early: photorealism, cyberpunk, illustration. Seedance responds strongly to style anchors.

Camera Movement

Camera slowly dollies back to reveal the full kitchen behind him

💡 Use explicit language: "dolly back", "tracking shot", "crane up".

Secondary Motion (separate)

Secondary motion: other chefs moving in practiced rhythm, overhead lights casting warm moving shadows

💡 Describe secondary motion explicitly as a separate layer, prevents subject/background confusion.

Setting + Atmosphere

Steam curling upward, busy open kitchen visible behind him

💡 Describe environment as it interacts with the main action.

Lighting

Warm tungsten lighting from overhead, single directional source

💡 Seedance handles wide dynamic range well, lighting spec anchors the grade.

Example prompt

Photorealistic style. A chef in a white apron skillfully flips a steak in a cast-iron skillet, flames rising briefly, steam curling upward. Camera slowly dollies back to reveal a busy open kitchen. Secondary motion: other chefs moving in practiced rhythm, ambient steam and flickering flames. Warm tungsten lighting from overhead.

What works

Separate primary and secondary motion explicitly, prevents subject/environment confusion
Visual style anchor early (photorealism, cyberpunk, illustration)
Precise camera language: "dolly back", "tracking shot", "crane up"
Motion intensity adverbs: "quickly", "powerfully", "slowly" modulate speed
Multi-character interactions and complex action sequences are a strength

What fails

Vague or unstructured prompts produce generic results
Mixing subject and environment motion in the same sentence
Missing style anchor defaults to generic output

Example output

Generated on ApeFX

"Photorealistic style. A chef in a white apron skillfully flips a steak in a cast-iron skillet, flames rising briefly, steam curling upward. Camera slowly dollies back to reveal a busy open kitchen. Secondary motion: other chefs moving in practiced rhythm, ambient steam and flickering flames. Warm tungsten lighting from overhead."

Ready to generate?

Put these formulas to work on ApeFX, every model, in one place.

Generate with Seedance on ApeFX

Related guides

Kling

Kling AI Prompting Guide