Write effective prompts for ByteDance's Jimeng Seedance 2.0 multimodal video model - using the @ reference system for images, video, and audio, plus camera moves, effects, and beat-matching.
---
name: seedance-prompt-en
description: Write prompts for Jimeng Seedance 2.0 multimodal video generation. Use when crafting text/image/video/audio prompts with the @ reference system for camera moves, effects, video extension, and music beat-matching.
---
Prompt-engineering guide for ByteDance's Jimeng Seedance 2.0 AI video model.
## How to use
1. Respect input limits: up to 9 images, 3 videos (2-15s total), 3 audio clips (<=15s), 12 files combined; output 4-15s at 480p-720p with auto sound. No realistic human faces allowed.
2. Use the `@` reference system to assign a role to each asset, e.g. `@Image1 as the first frame`, `@Image2 as the last frame`, `@Video1` for motion reference, `@Audio1` for beat.
3. State the role of every reference explicitly, then describe camera movement, subject action, style, and timing in natural language.
4. Apply the recipes for common jobs - effect replication, video extension/editing, music beat-matching, e-commerce ads, short dramas, and educational content.
Full skill & source: https://github.com/dexhunter/seedance2-skill/tree/e06c7c63a766d623004a2807881c30685ce517af