Bomx/super-video-maker-skill

What it solves

This project provides a comprehensive "skill" for AI agents (like Cursor or Claude Code) to automate the end-to-end production of high-quality videos. It bridges the gap between fragmented AI video tools by giving an agent a structured pipeline to handle everything from scriptwriting and asset gathering to final technical quality control, avoiding the "AI slop" often found in generic generated content.

How it works

The repository acts as a reusable instruction set and toolkit. It provides a main instruction file (SKILL.md) and a suite of Python and JavaScript tools that the agent uses to orchestrate various third-party services:

• Visuals: Uses HeyGen for synthetic avatars, fal.ai/Replicate for b-roll, and OpenAI for image generation.
• Composition: Employs FFmpeg for basic editing, and Remotion or HyperFrames for React/HTML-based timeline composition.
• Audio & Captions: Uses ElevenLabs for music/voice and OpenAI Whisper for word-level timestamping to ensure captions and visuals are perfectly synced.
• Quality Control: Includes specific scripts (broll_layout_qc.py, ffmpeg_qc.py) to check for visual collisions and technical stream errors before final export.

Who it’s for

Content creators, marketers, and developers who use agentic coding frameworks to automate the creation of professional-looking avatar explainers, product demos, and social media shorts.

Highlights

• Agent-Ready: Designed to be dropped directly into AI agent projects as a modular skill.
• Anti-Slop Philosophy: Prioritizes real screenshots, source receipts, and UI recordings over generic AI-generated imagery.
• Full Pipeline: Covers the entire workflow: story $\rightarrow$ script $\rightarrow$ storyboard $\rightarrow$ generation $\rightarrow$ composition $\rightarrow$ QC.
• Multi-Format Support: Capable of producing avatar explainers, faceless videos, and repurposed social shorts.