ai-video-remix

What This Skill Does

The ai-video-remix skill is a sophisticated AI-driven video composition engine designed to transform raw footage from your local library into professional-grade video content. By leveraging a high-performance stack—integrating ShotAI for semantic searching, LLM-based planning, and Remotion for programmatic rendering—it automates the laborious process of manual video editing. The skill understands the context and theme of your request, selects the best-matching clips from your library through semantic similarity, and orchestrates a cinematic sequence with appropriate visual effects, captions, and background music.

Installation

Installation is managed via the OpenClaw package manager. First, ensure you have the necessary system-level dependencies: ffmpeg and yt-dlp. To install, run: clawhub install openclaw/skills/skills/abu-shotai/ai-video-remix. Following installation, navigate to the skill directory, initialize your environment variables by copying .env.example to .env, and populate the required ShotAI credentials. The skill is highly modular and supports custom LLM providers to ensure you get the best creative results based on your preferred model.

Use Cases

• Travel Vlogging: Transform raw vacation clips into an organized travel vlog with location-based captions.
• Sports Highlights: Quickly edit exciting match moments into high-energy highlight reels with ESPN-style presentation.
• Cinematic Montages: Create nature documentaries or moody artistic montages based on specific aesthetic themes like Cyberpunk or Scenic.
• Memory Preservation: Generate "Best Of" reels from large, unorganized video libraries using the automated --probe mode.

Example Prompts

1. "Make a travel vlog from my Tokyo trip footage last summer, keep it upbeat."
2. "帮我做一个体育集锦，使用我硬盘里的足球赛录像，加点进球特效。"
3. "Create a nature montage using my camping videos, use a cinematic and calm style."

Tips & Limitations

• Probe Mode is Key: When you aren't sure what specific clips you have, always use the --probe flag. It scans your library and allows the LLM to design a story around what you actually possess, rather than forcing a template that might not fit your content.
• Performance: Rendering is CPU/GPU intensive. Ensure you have sufficient disk space for the temporary clip extraction phase.
• Quality Tuning: If results are generic, provide specific BGM files via the --bgm flag to anchor the pacing of the edit. This is often more effective than standard auto-searched audio.