seedance-prompt-en
由 dexhunter 提供
Coding Agents & IDEs简介
Write effective prompts for Jimeng Seedance 2.0 multimodal AI video generation.
描述
该技能的功能
A prompt engineering guide for Jimeng Seedance 2.0, ByteDance's multimodal video generation model. Covers the @ reference syntax for assigning roles to uploaded images, videos, and audio, plus camera language, time-segmented scripting, and output constraints. Handles use cases from product ads to short dramas to educational clips.
Seedance 2.0's @ reference system requires explicitly assigning a role to every uploaded asset, and ambiguous or missing role assignments are the most common cause of poor output, so having the exact syntax patterns and structural formulas reduces trial-and-error significantly.
何时使用
- Writing a 15-second product showcase ad with beat-synced music
- Extending an existing video clip forward or backward in time
- Replicating specific camera movements from a reference video
- Scripting a dialogue-driven short drama with timed scene breakdowns
- Creating beat-matched montages from a set of landscape photos
示例工作流
User asks: I have a product image and a reference ad video. Write a 15-second product showcase prompt.
- 1
Ask the user to confirm their assets: one product image (@Image1) and one reference video (@Video1)
- 2
Map each asset to its role: @Image1 as hero product, @Video1 for editing style and camera transitions
- 3
Structure the prompt with four timed segments: 0-3s intro, 4-8s multi-angle transitions, 9-12s lifestyle context, 13-15s hero shot with branding
- 4
Add audio direction referencing @Video1's background music and product interaction sound effects
- 5
Check that total file count is within the 12-file limit and no real human faces are in the uploads
A complete time-segmented prompt using the @ reference syntax, ready to paste into Seedance 2.0
前置要求
- Jimeng Seedance 2.0 platform account (ByteDance)
安装
clawhub install dexhunter/seedance-prompt-en运行 CLI 命令,或把这段提示词粘贴到 Claude Code、Cursor、Codex 或你在用的任意编程智能体里。