Devotd Github
NewAI agents

alchaincyf/huashu-motion v0.3.1 — Code-based promo video production skill

alchaincyf/huashu-motion is a skill that lets coding agents create promo videos and music with code, and supports scene-by-scene editing in the browser.

Repositoryalchaincyf/huashu-motion ↗

alchaincyf/huashu-motion is a skill that lets coding agents create promo videos and music with code, and supports scene-by-scene editing in the browser.

GitHub description: 让你的agent一键直出商业级产品宣传片:会作曲,能指着画面改。Commercial-grade launch films by your coding agent — point at the frame to change it.

stars
★ 1.9K
Forks
177
Language
JavaScript
License
MIT
Latest release
v0.3.1
Read original source ↗View on GitHub ↗

alchaincyf/huashu-motion is an Agent Skills skill that proposes a direction for promo videos based on a product or dataset and creates animation and music with code. The latest release published on GitHub is v0.3.1. The provided records list no changes for this version, so the specific changes cannot be confirmed. The video is implemented as index.html, and in the browser editor you can click a scene to leave revision comments.

This skill uses code that draws in the browser instead of a video generation model. You point to the part of the screen you want changed, and the corresponding scene is fixed; the result is left as a file that can be reopened and edited. According to the README, it inspects and exports the video with local Chrome and FFmpeg, and also supports narration, music and sound effects. Its configuration includes music composition and sound effects as part of the video production process. It is written in JavaScript and is licensed under the MIT license.

That said, the provided material contains neither a changelog for v0.3.1 nor comparison data for judging the quality of the output. The repository's quality score is 5.0, and this score alone makes it hard to judge the completeness of the actual production results. Therefore, the strengths that can be confirmed at this point are the editing flow of specifying a scene to revise and the breadth of the production features, while output quality has to be verified separately.