Subject + action
Name the focal person, product, object, or place and one visible event.
Seedance 2.0 prompt guide
Build one clear scene from text, or assign separate jobs to image, video, and audio references. Keep appearance, movement, camera, and sound under distinct control.
Text-to-video structure
Start with what happens on screen. Then add the scene, camera, timing, and sound details that change the result.
Name the focal person, product, object, or place and one visible event.
Set the location, light, color, texture, weather, and visual medium.
Choose the frame, viewpoint, movement, direction, speed, and rhythm.
Identify the speaker, short line, key effect, ambience, music, or silence.
Multimodal direction
State what transfers from each file and what does not. One source of truth per detail reduces conflicts.
Face, wardrobe, product, location, composition, palette, or art direction.
Action, body mechanics, camera path, transition, effect, or performance timing.
Voice timbre, dialogue content, melody, ambience, emotional build, or edit rhythm.
9 images · 3 videos · 3 audio files · audio needs an image or video
Copy, then rebuild
Text only · native audio · 8 seconds
A ceramic artist in a charcoal apron lifts a finished blue bowl from a spinning wheel and holds it toward the window. Medium close-up, slow dolly in, warm late-afternoon side light, visible clay texture, quiet documentary realism. Audio: the wheel slows, one soft ceramic tap, and light workshop ambience. No dialogue.
First frame · controlled motion · 6 seconds
Use [Image 1] as the exact appearance and composition anchor. Preserve the model's face, silver jacket, wet hair, and neon street. She takes one calm breath, turns her eyes toward a passing light, and tightens her grip on the umbrella. Hair and jacket edges move gently in the wind. Very slow handheld push in. End on her steady gaze. Audio: light rain, one car passing through water, and distant traffic.
Image + video + audio · 10 seconds
[Image 1] controls the runner's face, red windbreaker, black shorts, and silver shoes. Preserve those details throughout. [Video 1] controls only the sprint rhythm, hurdle movement, and low side-tracking camera. Do not copy its performer, warehouse, or gray clothes. [Audio 1] controls the rising drum rhythm and edit intensity. Begin restrained, accelerate through the hurdle, and land on the final beat. Dawn city street after rain, realistic sports commercial, no dialogue.
Image to video
Repeat only the face, product, wardrobe, logo, color, or composition details that must not drift.
Separate subject motion, environmental motion, and camera motion into clear sentences.
Use first-and-last-frame mode when the final pose, product angle, or composition must match exactly.
Enable generated audio, then write the short dialogue, key effect, and ambience that fit the scene.
Seedance prompt questions
Use subject and action, scene and style, camera and timing, then audio and dialogue. For reference-to-video, also state exactly what each image, video, or audio file controls.
Read the full answer →Assign each file a job in the prompt. Use an image for appearance or style, a video for action or camera motion, and audio for voice, music, sound, or editing rhythm. State what must carry over and what must not.
Read the full answer →Let the input frame define the appearance and composition. Use the prompt for subject motion, environmental motion, camera behavior, sound, and the final state. Add a last frame when the ending must match exactly.
Read the full answer →Enable generated audio, identify the speaker, keep the exact line short enough for the clip, and add only the important effects and ambience. Reference audio can control voice timbre, music, dialogue content, or visual rhythm.
Read the full answer →The current Seedance 2.0 open platform supports up to nine image references, three video references, and three audio references. Text can be combined with those assets, but audio-only input is not supported for 2.0.
Read the full answer →Build your Seedance prompt
Official sources