Prompt guide
Tie one short spoken line, room tone, and each important sound cue to the visible action that should produce it.
Prompt structure
Quiet café, one continuous conversation beat. The woman says, 'I kept the last ticket.' The man pauses before answering. Warm room tone, distant cups, and one clear ceramic spoon tap exactly when her hand touches the cup. Natural speech, restrained performance, no music, no subtitles.
Replace the variables and reference labels with your own licensed material. Treat the output as probabilistic and review the full clip.
Exact route workspace
The generator is pinned to the recommended model ID. Confirm current controls and credits before submitting.
Seedance 2.0 Pro: Complex multimodal references and selected final takes. The generator on this page is pinned to that exact connected model ID.
Use it as a shot structure, then replace the subject, references, actions, dialogue, and constraints with your own licensed material. Remove instructions that do not serve this specific result.
Keep the subject or product anchors, core action, reference roles, aspect ratio, duration, and ending composition fixed. Change only the variable the current test is meant to answer.
Change one production variable at a time: one camera move, action beat, reference assignment, sound cue, or composition choice. Multiple simultaneous changes make the result difficult to diagnose.
Upload only assets needed to control identity, shape, setting, motion, or sound for this task. Give every asset one explicit role and remove duplicates or references that compete.
Identify whether the failure comes from reference conflict, excessive action, camera competition, duration, or an incompatible ending. Simplify the brief and rerun the same baseline before adding detail.
Review the full clip frame by frame for identity, faces, hands, geometry, text, logos, continuity, dialogue, sound timing, crop, and compliance with the brief.
Commercial use depends on your rights and applicable terms. Confirm permission for every uploaded person, brand, voice, image, video, and audio asset, and manually verify the generated result before publishing.