lanshu-create-ai-presenter-video

This is an Agent Skill by 岚叔 (lanshu) that walks an AI coding agent through making an explainer video from start to finish. There are two routes: a digital-presenter route that uses an authorized real person's likeness and voice to generate a lip-synced presenter, and a no-real-person route that produces a styled animated explainer in one of 12 styles. The project has a single SKILL.md, and all 12 styles live in that one Skill. It's aimed at science, knowledge and course explainers, and works with terminal-capable coding agents such as Claude Code and Codex.

Who it's for:Creators of science, knowledge or course videos who want an AI assistant to help produce the whole piece.

Before you start: You need Python, FFmpeg, bash and jq, and the styled-animation route also needs Node.js, rsync and numpy. The default voiceover sends each line of text to MiniMax's speech service, which needs your own API key and is billed by usage; if you bring your own narration, you can skip it. The digital-presenter route uploads a real person's likeness and voice to the generation service you choose, and the Skill spells out the requirements: the person must be an adult and must agree to the use of their likeness, voice cloning and the upload. Remote generation may be billed, and the author suggests making a small test first and setting a retry limit. Fonts are downloaded from the web. The code is MIT-licensed, and fonts, models and services have their own terms.

Based on the project's public documentation. Not tested by AI Inspiration World.