Cinelyo creation guide
Seedance 2.0 Prompt Guide
Build more controllable AI videos with clear subjects, deliberate motion, camera direction, references, and audio cues.
Prompt basics
Write in natural language, but give every important decision a clear place in the prompt. Start with the action, then add visual direction.
Subject
RequiredWho or what appears, including defining visual traits.
Motion
RequiredWhat the subject does and how the action develops.
Environment
Location, time, weather, lighting, and surrounding details.
Camera
Shot size, angle, lens behavior, and camera movement.
Aesthetics
Medium, genre, color treatment, texture, and pacing.
Audio
Dialogue, ambience, sound effects, music, and timing.
Prompt formula
[Subject] + [Motion] + [Environment] + [Camera] + [Aesthetics] + [Audio]
Example prompt
A ceramic astronaut walks through a rain-soaked night market, studying the glowing food stalls. The camera begins with a low wide shot, then slowly tracks beside the astronaut. Cinematic realism, wet reflections, soft neon light, restrained film grain. Distant chatter, rainfall, and a quiet electronic score.Put the most important action first. Keep later details consistent instead of stacking conflicting styles or camera moves.
Text, dialogue, and audio
State the exact words, speaker, timing, placement, and sound character. Quotation marks make dialogue and on-screen copy easier to distinguish.
On-screen text
Specify exact copy, position, style, and when it appears.
Dialogue
Identify each speaker and write their exact line in quotes.
Sound direction
Describe voice, ambience, effects, music, and synchronization.
Prompt formula
[Speaker/action] + says “…” + [voice] + [sound bed] + [subtitle or title timing]
Example prompt
A chef places the finished dessert on the counter and says in a calm, warm voice: “Small details make the moment.” Soft kitchen ambience continues underneath. As the camera settles into a close-up, the title “Made with patience” fades in at the lower center in clean white type, then disappears after two seconds.Use short, common words for generated text. Ask for subtitles explicitly when the spoken line must also appear on screen.
Image references
Name each uploaded image by order and state exactly what should be borrowed from it. Separate identity, styling, environment, and composition instructions.
Subject consistency
Reference face, clothing, product shape, or other identity traits.
Combine references
Assign a distinct job to each image instead of asking the model to blend everything.
Storyboard control
Describe the order of shots when an image contains multiple panels.
Prompt formula
Use [element] from Image N + [new scene/action] + preserve [specific features]
Example prompt
Use the character appearance and red jacket from Image 1. Use the glass observatory environment from Image 2. The character walks toward the telescope, pauses, and looks up as a meteor shower begins. Preserve the character’s facial features and jacket details. Start with a rear medium shot, then arc smoothly to a close-up.Avoid vague requests such as “reference all images.” Give every upload one clear responsibility.
Video and camera references
A reference video can guide action timing, camera choreography, transitions, or effects. Say which property to copy and which content to replace.
Action
Follow gesture, performance, rhythm, or physical motion.
Camera movement
Reuse a push, orbit, handheld path, tilt, or reveal.
Visual effect
Match the timing and behavior of particles, transformations, or transitions.
Prompt formula
Follow [property] from Video N + generate [new content] + preserve [timing/path/effect behavior]
Example prompt
Follow the camera path and pacing from Video 1, but replace the scene with the futuristic greenhouse from Image 1. Keep the same first-person forward movement and final upward reveal. Plants emit subtle blue bioluminescence, with natural footsteps and greenhouse ambience.Requesting only one or two properties from a reference usually gives more predictable results than copying its entire visual language.
Editing and extension
For edits, identify the source video, the exact change, and everything that must remain untouched. For extensions, describe the new action and continuity point.
Add, remove, replace
Name the element, position, time range, and unchanged details.
Extend a shot
Continue before or after the source while preserving motion and lighting.
Connect clips
Define the visual or action cue that bridges one clip into the next.
Prompt formula
[Edit operation] in Video N + [time/position] + preserve [unchanged content] + [continuation]
Example prompt
Extend Video 1 after its final frame. Preserve the same camera speed, overcast lighting, and the runner’s appearance. The runner turns the corner into a quiet plaza, slows to a stop beside a fountain, and looks back toward the camera. Continue the original footsteps naturally, then fade into soft city ambience.Continuity improves when you repeat the source clip’s lighting, direction of motion, camera speed, and audio bed.