Reference to Video
Upload at least one image or video (audio-only is not allowed). In the prompt, tag assets with @image1, @video1, @audio1.
Ready for Generating
What is Reference to Video?
Reference to video lets you direct a clip with several inputs at once. Use images for characters, outfits and scenes, video clips for motion and camera work, and audio for rhythm or voice, then tell the model what each one is for. It is the most precise way to keep characters and style consistent across shots.
Images, video and audio together
Combine reference images, short video clips and audio files in one generation, depending on the model.
Tag references in the prompt
Refer to each input as @image1, @video1 or @audio1 so the model knows exactly what to use and how.
Consistent characters and style
Lock a character’s face, clothing or a visual style with reference images and reuse them across clips.
How to Create a Video from References
Four steps to direct a multimodal clip:
Upload references
Add images, video clips and audio. Seedance models accept all three; some models accept images only.
Tag them in your prompt
Write the scene and point to each file, for example "the woman from @image1 dances like @video1, with @audio1 as music".
Choose model and length
Pick a model, aspect ratio and duration.
Generate and review
Check how each reference was used, then adjust the tags or prompt and generate again.
Reference to Video FAQ
Direct Your Video with References
Combine images, motion and sound, and let the model bring them together.