عن المشروع

scroll-scrub-hero is an MIT-licensed, zero-dependency Node.js skill for creating scroll-driven website heroes from a single photograph. It is designed to run inside SKILL.md-compatible AI coding agents such as Claude Code, Codex, and Cursor, and it uses the kie.ai platform to run the underlying image and video models. The core workflow begins with a photograph and a short description of the desired transformation, such as turning a bare yard into a finished pool. The pipeline then writes a storyboard, renders keyframe stills from the original photo, presents a contact sheet for approval, generates video tweens between approved stills using Kling first/last-frame models, and finally cuts the videos into numbered WebP frames plus a small config.js file. The result is a lightweight, canvas-based hero that scrubs smoothly with scroll, works offline once deployed, and avoids runtime AI or video streaming. The project includes helper scripts for storyboarding, keyframe generation, tweening, frame building, diagnostics, and pricing checks, along with documentation, references, examples, and tests. It requires Node 18+, Git, ffmpeg/ffprobe, and a kie.ai account with credits. A typical six-step hero is estimated to cost about $2.55, with video generation making up most of the expense. The README emphasizes cost transparency, approval before expensive steps, and practical troubleshooting for common issues such as subject drift, hard cuts, and heavy mobile performance.