Seedance Text-to-Video Prompt Examples for Complete Scenes

Text-to-video starts without a visual reference. That gives Seedance more creative freedom, but it also means the prompt must define the subject, setting, physical action, camera, light, and continuity that an image-to-video workflow inherits from its first frame. A strong prompt reads like a compact shooting brief.

ByteDance describes Seedance as supporting text- and image-driven generation, camera control, and cohesive multi-shot storytelling. The exact inputs, duration, audio options, and reference controls vary by Seedance version and interface, so use the structure below with the controls available in your workspace.

Use a seven-part text-to-video brief

Write in this order: format and duration, subject, setting, action sequence, camera plan, physical and visual continuity, and ending. A single-shot example is: “Six-second cinematic product shot, 16:9. A brushed-steel wristwatch rests on dark slate at blue hour. The second hand begins moving as a narrow cool highlight travels across the bezel. The camera makes a slow push-in from a low three-quarter angle. Preserve the watch geometry, readable dial, and grounded shadow. Finish on a stable close-up with the face centered.”

Each phrase answers a production question. Remove style filler before removing the action or camera direction.

Example 1: documentary street scene

“Eight-second documentary-style street scene at dawn. A bakery owner lifts the metal shutter and warm interior light spills onto the wet pavement. Two early commuters pass in the distant background. Medium-wide locked camera with subtle handheld micro-movement, natural pace, cool exterior light contrasting with warm interior light. Keep the owner’s clothing and storefront geometry consistent. End as the shutter reaches the top and the owner pauses in the doorway.”

Why it works: the light change is caused by a physical action, the background supports the place without taking over, and the ending completes the event.

Example 2: cinematic character entrance

“Seven-second cinematic scene inside a quiet train carriage at night. A woman in a rust-red coat enters from the rear door, takes three measured steps down the aisle, and looks toward an empty seat. The camera tracks backward at matching speed in a medium shot. Reflections move naturally across the windows; overhead lights remain stable. Preserve her face, coat, and body proportions. Finish as she stops beside the seat.”

If the model changes the face, reduce the walking distance and expression before changing the setting.

Example 3: food commercial

“Five-second overhead food commercial, vertical 4:5. A chef’s right hand places one bright green garnish at the center of a steaming noodle bowl, then exits the frame. Locked overhead camera, soft daylight from the left, realistic steam and contact shadows, dark oak table with restrained props. Preserve the bowl shape and ingredient arrangement. End with the complete dish still for one second.”

This prompt limits the hand to one clear action and gives the final food image time to read.

Example 4: landscape scale reveal

“Nine-second wide landscape shot at sunrise. Begin behind a lone hiker standing on a dark ridge. The camera rises slowly above shoulder height to reveal a vast fog-filled valley and distant mountains. The hiker remains anchored in the lower center; jacket fabric responds to a light wind. Keep the horizon level and the light direction consistent. End on the full valley as camera motion settles.”

The foreground figure gives the reveal scale. The camera moves, but the subject does not need a complex performance.

Example 5: stylized illustration

“Six-second hand-painted editorial animation in a restrained gouache style. A small red tram travels from left to right through a snowy town while window lights switch on one by one. The camera tracks alongside in a wide shot. Maintain the same brush texture, limited navy-and-red palette, building shapes, and tram proportions. Snow falls gently in one direction. Finish as the tram stops beneath the station sign.”

Style prompts still need physical direction. Name texture and palette, then describe the same action and camera logic used for realistic footage.

Plan a short multi-shot sequence

When the enabled Seedance workflow supports multi-shot prompting, divide the scene into numbered beats with a shared continuity block. For example: “Maintain the same woman, rust-red coat, blue-hour station, cool lighting, and realistic cinematic style across all shots. Shot 1: wide establishing view as the train arrives. Shot 2: medium side view as she turns toward the doors. Shot 3: close-up as warm carriage light reaches her face.”

Each shot should add new information. Do not repeat three versions of the same framing. Keep character description, wardrobe, time of day, and color treatment identical across the sequence.

Control timing with action size

Do not write a feature-film scene for a short generation. One door opening, three steps, a single product reveal, or one camera rise can complete within several seconds. If an action repeatedly stops halfway, reduce its distance or divide it into another shot.

Diagnose the output against the brief

Check subject identity, action order, camera direction, environment response, continuity, and ending separately. If the camera is wrong, remove secondary motion. If the subject changes, simplify the angle and action. If a multi-shot sequence loses continuity, shorten the shared identity block to the few visible details that matter most and repeat it exactly.

Text-to-video prompting is scene design under a small time budget. Define what the viewer sees first, what changes, how the camera observes it, what remains constant, and what final image proves the action is complete.