Seedance reference to video
Seedance Reference to Video
Upload what the video should follow: a face, a place, a camera move, a voice. Say in the prompt what each file is for, and Seedance builds a new shot around them.
- Up to 9 images
- 3 clips · 3 audio files
- Type @ to name files
- Images at no extra cost
- Up to 30 s on 2.5
Reference-to-video examples
Each clip lists the model and the exact prompt it was made with. Start from one and change what you like.
“Shot 1: close-up of an old lighthouse keeper in a wet oilskin coat climbing a spiral iron staircase, a swinging lantern lighting his face, rain hammering the windows. Shot 2: wide shot from the sea as the lighthouse beam sweeps across huge storm waves crashing on the rocks, lightning splits the sky. Rolling thunder and howling wind”
Text · 2.0 · 720P · 5s
Use this prompt“Close-up of a young barista with a dark apron behind a café counter, she looks up from the espresso machine, smiles warmly and says "Your usual?". Soft morning light, shallow depth of field, café chatter and an espresso machine hissing in the background”
Text · 2.0 · 720P · 5s
Use this prompt“A street dancer in a yellow hoodie spins and drops into a freeze on a city rooftop at dusk, skyline lights coming on behind her, the camera circles her in one smooth orbit, a punchy hip-hop beat”
Text · 2.0 · 720P · 5s
Use this prompt“A red paper boat drifts down a rain-soaked city gutter at night, neon reflections, the camera follows low and close, soft rain sounds”
Text · 2.0 Fast · 480P · 5s
Use this prompt
How references guide a Seedance video
A reference is a file the model follows without putting it on screen as a frame: a person's face and clothes, the look of a room, the path of a camera, a voice or a piece of music. Mixed image, video and audio references arrived with Seedance 2.0, and they are what sets the family apart from frame-only image to video.
Files are numbered in upload order: Image 1, Image 2, Video 1, Audio 1. Type @ in the prompt to insert one, then give it a job: "The chef from Image 1 cooks in the kitchen from Image 2; copy the camera move in Video 1; Audio 1 is the music."
Limits here: up to 9 images, 3 clips and 3 audio files. Each clip is 2 to 15 seconds, with 15 seconds per kind at most, and audio needs at least one image or clip with it. Reference images cost nothing extra; a reference video adds its seconds to the output, billed at a lower rate.
For a character that holds, use one or two sharp, well-lit pictures of the face rather than many mixed ones. For a camera move, upload a short clip that shows only that move. Give every file a single job and name it in the prompt, so the model doesn't have to guess.
Seedance 3 AI is independent and not affiliated with or endorsed by ByteDance. Seedance, Dreamina, Jimeng and Doubao are trademarks of their owners.
How to make a video from references
Nothing to install and no invite to wait for. Most clips are ready in a few minutes.
- 01
Write the shot
Who is in it, what they do, how the camera moves, the light and what we hear. Or open an example and edit it.
- 02
Add an image or references
Optional: a first or last frame, or reference images, clips and audio for the model to follow.
- 03
Pick a model and render
Choose 2.0 Fast, 2.0 or 2.5, the resolution and the length. The button shows the credits; download the MP4 when it's done.
More ways to create
Seedance 3 AI home
Every Seedance model in one generator, plus Seedance 3 status.
Seedance 3.0
Release status, facts vs rumors, and what to use now.
Seedance 2.5
The newest Seedance: up to 30 s in one clip.
Text to video
Write the shot, get a video with sound.
Image to video
Animate a picture, with an optional end frame.
Reference to video questions
Pick Reference at the top, upload images, clips or audio, then describe the new video and what each file is for. Seedance keeps what you point to and creates the rest.
By its number: Image 1, Image 2, Video 1, Audio 1, in upload order. Type @ in the prompt box or tap a chip under it.
Up to 9 images, 3 clips and 3 audio files. Clips run 2 to 15 seconds each and 15 seconds per kind in total; audio needs an image or clip with it.
Images don't. With a reference video, its seconds are added to the output seconds and the total is billed at the lower reference rate. The button shows the final number.
A second clear picture of the same person from another angle, named together in the prompt ("the woman in Image 1 and Image 2"). Fewer, sharper references beat many mixed ones.
Yes. Give each one a clear picture, name both, and say who does what: "the boy from Image 1 hands a map to the woman from Image 2". That keeps the model from swapping them.
Image to video puts your picture on screen as the first (and last) frame. Reference mode treats pictures and clips as guides, so you get the same character or place in a new shot.
They're used only to make your video and are deleted after 30 days. Upload only files you own or have permission to use.
Turn your references into a video
Write one sentence, choose a length and get a video with sound. Your first one is free.
Start creating