coolxxxx/video-prompt-reverse-engineer
Developer toolsThis page was translated automatically. Technical details may differ from the original. View the original
Reverse-engineer reference videos into executable AI video prompts (Sora, Kling, Runway, Vidu, Pika, Seedance 2.0/2.5 native format)
Overview
A skill for AI video generation workflows: Breaks down reference videos into concrete, directly usable AI video prompts, supporting Sora, Kling, Runway, Vidu, Pika, and . This skill only analyzes and prompt text and does not directly submit generation tasks. Only learning camera logic and visual methods is required; no real people, brands, lines, or copyrighted music are replicated. - : Short videos, product showcases, Vlog clips, output "Seven Elements Breakdown + Finished Product Prompt + Negative Prompt". - : Advertisements, short dramas, MVs, multi-shot replicas, output 18-dimensional per-shot timeline + finished storyboard prompts + global style anchoring. - : On top of standard breakdown, rewrite finished product prompts according to Seedance's official syntax: - @引用 Multimodal references (each reference annotation uses + exclusions) - Time-segmented descriptions (each segment includes end status, ensuring shot transitions) - Audio parenthesis syntax: music ( ), sound effects, dialogue { }, subtitles 【 】 - 2.0 complexity matched at 4–15 seconds; 2.5 Supports up to 30 seconds of continuous take, up to 50 reference materials - characters/storylines, products/advertisements, scenery/travel, music/MVs, food/crafts/tutorials. 1. (Such as tools compatible with the Agent Skills specification like ZCode / Claude Code) Place this directory into the skills directory (such as ~/.zcode/skills/ or the project .codex/skills/), or directly clone: - "Help me reverse the prompt for this video" (General Edition) - "Break down this ad shot by shot with the full Pro version" (Professional Edition) - "Reverse the prompt for Seedance 2.5" (Seedance native format) - Video overview + seven elements/per-shot breakdown - reference material list (@图片1 character references, exclusions, etc.) - generation settings (mode/duration/scale, do not include prompt body) - Seedance native prompt body text (time segment + audio parentheses + consistency constraint) - negative prompt + original boundary declaration - LINUX DO — This project is shared and exchanged through the LINUX DO community. Thanks to the community and friends for their support and feedback. A Skill for AI video generation workflows: reverse-engineers a reference video into concrete, ready-to-use AI video prompts.
Overview (machine translated)
Chinese | English
Chinese Description
A skill for AI video generation workflows: Breaks down reference videos into concrete, directly usable AI video prompts, supporting Sora, Kling, Runway, Vidu, Pika, and Seedance 2.0 / 2.5 (Dream / Doubao / Volcano Engine) native formats.
This skill only analyzes and prompt text and does not directly submit generation tasks. Only learns camera logic and visual methods, without replicating real people, brands, lines, or copyrighted music.
What can be done?
- Quick Universal Version: Short videos, product showcases, Vlog clips, output “Seven Elements Breakdown + Finished Product Prompts + Negative Prompts”.
- Full Professional Edition: Advertisements, short dramas, music videos, multi-shot replicas, output 18-dimensional frame-by-shot timeline + finished storyboard prompts + global style anchoring.
- Seedance 2.0 / 2.5 Native Format: Based on the standard breakdown, the finished product prompt is rewritten according to Seedance’s official syntax:
@引用Multimodal references (each reference annotated for purpose + exclusion)- Segment description (each segment includes the ending status, ensuring shot transitions)
- Audio parenthesis syntax: Music
( ), Sound effects{ }, Subtitles【 】 - 2.0 Complexity matches in 4–15 seconds; 2.5 supports up to 30 seconds in a single take, with up to 50 reference materials
- Five Categories of Themes: Characters/Plot, Products/Advertisements, Scenery/Travel, Music/MV, Food/Handicrafts/Tutorials.
File Structure
How to use it
-
Install on a client that supports Skill (such as ZCode / Claude Code or tools compatible with the Agent Skills specification), place this directory into the skills directory (such as
~/.zcode/skills/or Project.codex/skills/), or directly clone: -
Trigger Skill: Upload or provide a reference video, then say:
- “Help me reverse the prompt for this video” (general version)
- “Breaking This Ad Down Shot by Camera with the Full Pro Edition” (Pro Edition)
- “Reverse Seedance 2.5 Prompt” (native Seedance format)
- Output Content (Seedance Objectives as an Example):
- Video overview + seven elements / shot by shot
- Reference material list (
@图片1character image references, exclusions, etc.) - Generation settings (mode/duration/scale, not included in the prompt text)
- Seedance native prompt body text (time segments + audio parentheses + consistency constraints)
- Negative prompts + original boundary statements
Seedance prompt example skeleton
Friend Links
- LINUX DO — This project is shared and exchanged through the LINUX DO community. Thanks to the community and friends for their support and feedback.
English
This page was translated automatically. Technical details may differ from the original.
Recommended Tools
Try a different keyword or remove a filter.
Install
npx skillfish add coolxxxx/video-prompt-reverse-engineer