Multimodal Masterclass: Universal Reference Guide

The core breakthrough of Seedance 2.0 lies in its Universal Reference system. No longer limited to simple text-to-video, it supports a "full-dimensional" input of text, images, videos, and audio.

1. Understanding the "12-File Limit"

In a single generation flow, you can combine:

• Up to 9 Images (defining visual style, characters, or first/last frames)
• Up to 3 Videos (specifying motion patterns and camera movement)
• Up to 3 Audio files (driving rhythm, ambient sound, or lip-sync)

The total number of files cannot exceed 12.

2. Core Syntax: Dynamic @Mention

In the prompt box, use the @ symbol to "tag" your assets just like tagging a friend on social media:

Example Prompt: "Use @Image1 as character reference, replicate the side-tracking camera movement from @Video1, with @Audio1 driving the bass rhythm."

3. Advanced Tactic: Cross-Modal Synergy

When you want to replicate complex motions, video references are far more effective than text.

• Motion Replication: Upload a video of yourself dancing and tag @Video1, then tell the AI: "A medieval woman in a dark green dress, mirroring @Video1's movements."
• Start/End Frame Control: Upload Image1 (start) and Image2 (end) to let the AI interpolate the seamless transition.

4. Best Practices

1. Define Character Weight: When using multiple image references, ensure the primary character image is clear.
2. Avoid Instruction Conflicts: If the text says "run fast" while the video shows "strolling," the output may suffer from distortions.
3. Audio-Driven Dynamics: Seedance 2.0 can automatically adjust camera push/pull intensity based on audio amplitude.

Start combining your assets today and unlock true "director-level" creativity!