Seedance 2.0
适合高控制深度与重参考电影化片段。
Up to 9 images for style, subject, and composition.
Generate the final video with synced audio when the selected model supports it.
你基于参考素材生成的视频会显示在这里。
Preview the kinds of prompts and outputs this tool flow is designed for.
Reference Images
Use character and style references to keep the subject consistent across a new cinematic scene
Reference Video
Guide the motion, pacing, and camera rhythm with a reference clip while generating a fresh result
Multimodal Reference
Combine images, video, and audio references to steer look, motion, and atmosphere in one workflow
Reference to Video 面向一致性优先的生成任务,可通过图像、视频和多模态参考约束风格、身份与动作,而不只依赖提示词。
Text to Video 更快做探索,Reference to Video 更适合严格连续性要求。
Image to Video 偏单锚点动化,Reference to Video 可处理更复杂多源约束。
当角色身份、场景风格或动作节奏必须跨输出稳定跟随参考素材时优先选择。
先给最小必要集合(定义身份与动作),仅在出现特定不一致时再增量补充。
在对齐测试阶段先用短时长,确认一致性后再提升时长和质量。