Seedance 2.0 Reference To Video

Seedance 2.0 reference-to-video generation using text plus optional image, video, and audio references.

Model Overview

Seedance 2.0 Reference To Video generates short video clips from a prompt and optional reference media. It supports image, video, and audio references, 480p, 720p, and 1080p output, 4-15 second duration controls, flexible aspect ratios, and optional generated audio.

Best At

• Multimodal video generation with reference images, videos, or audio.
• Prompt-directed clips that borrow subject, style, or motion cues from supplied media.
• Reference-heavy creative work where output quality matters more than fast iteration.

Limitations / Not Good At

• Reference files must be aligned with prompt instructions to avoid conflicting guidance.
• Video-reference pricing includes input video duration as well as output duration.
• 1080p provides the highest fidelity, but costs more than 720p and 480p.

Ideal Use Cases

• Creating video variations from visual and audio reference sets.
• Producing short campaign, storyboard, and concept clips with multimodal guidance.
• Generating final candidates after testing reference direction in faster modes.

Input & Output Format

• Input: required prompt; optional image_urls, video_urls, audio_urls, aspect_ratio, duration, resolution, generate_audio, and seed.
• Output: generated video URI returned on response.

Performance Notes

• Pricing scales with generated video size and duration.
• Runs with video references also account for input video duration, with Fal applying its video-input multiplier.