Reference to Video AI
Create a new video from reference images, clips, or audio. Show the model what to keep and describe what happens in the new shot.
How to turn references into video
1.Choose your references
Add clear views of your character, product, or setting. With Seedance 2.0 or 2.5, you can also add a short clip as a motion reference or audio as a rhythm guide.
2.Direct the new shot
Describe the action, framing, and camera movement. Say which subject, outfit, or product details should carry over from each reference.
3.Generate and review
Choose the duration, resolution, and aspect ratio. Check the credit estimate, generate your video, and compare the result with your references.
Reference images for consistent characters
A clear portrait and a second angle give the model two useful views of the same character. Use the prompt to describe how that character moves in the new scene.
Reference to video FAQ
Which model should I choose?
Happy Horse 1.0 and 1.1 use one to nine reference images. Choose Seedance 2.0 or 2.5 if you also want to guide the new shot with clips or audio. All models require a prompt.
What is the difference between reference to video and image to video?
Reference to video uses your media to guide a new composition. Image to video animates a particular starting image. Video to video transforms an existing clip.
How many reference files can I use?
Seedance 2.0 supports up to 9 images, 3 video clips, and 3 audio files, with separate total duration limits of 15 seconds for video and audio. Seedance 2.5 supports up to 30 images, 10 video clips, and 10 audio files, with separate totals of 30 seconds for video and audio.
How is the credit cost calculated?
The estimate depends on the model, duration, and resolution. Seedance 2.0 costs 15 credits per generated second. Seedance 2.5 also factors in the duration of any reference video clips.