video-generation
Use this skill when the user requests to generate, create, or imagine videos. Supports structured prompts and reference image for guided generation.
Security Assessment
About video-generation
The Video Generation skill enables users to create high-quality AI-generated videos through structured prompts and automated workflows. This skill solves the challenge of translating creative video concepts into actionable generation parameters by providing a systematic approach that combines detailed scene descriptions, character definitions, camera specifications, dialogue, and audio elements into a cohesive JSON format that drives the video generation process.
The skill's main capabilities include creating structured JSON prompts that define every aspect of a video scene, supporting optional reference images to guide the visual style or serve as the first or last frame, and executing video generation through an automated Python script. Users can specify technical parameters such as aspect ratio, control camera movements and angles, define character interactions and dialogue, and layer audio elements including ambient sounds, music, and sound effects. The system handles the entire pipeline from prompt creation to final video output.
This skill is ideal for content creators, filmmakers, educators, and marketing professionals who need to quickly prototype video concepts, create storyboards, or generate video content without extensive production resources. It's particularly useful for visualizing narrative scenes, creating reference material for production teams, generating marketing content, or exploring creative concepts before committing to full-scale video production.
FAQ
What file format and location should I use for prompts?
Create JSON files in /mnt/user-data/workspace/ with descriptive names like 'scene-name.json'. The JSON should include title, background, characters, camera settings, dialogue, and audio specifications.
Can I use reference images to guide video generation?
Yes, you can provide reference images using the --reference-images parameter. These images can guide the overall visual style or serve as the first or last frame of the generated video. If only one image is provided, it will be used as the guided frame.
What aspect ratios are supported?
The skill supports custom aspect ratios through the --aspect-ratio parameter. The default is 16:9, but you can specify other ratios like 9:16 for vertical video or 1:1 for square format.
Where are the generated videos saved?
Generated videos are saved to /mnt/user-data/outputs/ with the filename specified in the --output-file parameter. Use descriptive names with .mp4 extension for easy identification.
Do I need to understand the Python script to use this skill?
No, you should not read or modify the Python script. Simply call it with the required parameters (--prompt-file and --output-file) and optional parameters (--reference-images and --aspect-ratio) as documented.
Install video-generation
Quick Setup:
- Copy the skill folder to
.claude/skills/ - Claude will automatically detect and use the skill
Repository
bytedance/deer-flow