cclank/lanshu-awesome-ai-video-kit

This is a Chinese-language reference kit for writing prompts for AI video generation tools, built and shared by someone who spent time on commercial AI video projects. The repository collects 543 tested prompts across 15 different AI video models, along with 7 Claude Code Skills, 21 methodology articles, and 3 small web tools, all organized into a browsable collection. The README is primarily in Chinese, with an English version linked separately. The 15 models covered include well-known commercial systems like Kling, Veo, Runway, Pika, Sora, and Hailuo, plus several Chinese-origin models like Seedance, HappyHorse, Wan, and Hunyuan Video, as well as four open-source models including CogVideoX and LTX-Video. For each model the repository documents the specific prompt formula or structure that model responds to, since different models have different conventions for how to describe camera movement, timing, lighting, and subject behavior. The README includes a quick comparison table covering things like maximum clip length, Chinese-language support, audio generation support, and physics simulation quality. The 7 Claude Code Skills are installable add-ons for Claude Code (the AI coding assistant) that automate common prompt tasks. Two of the skills are described as core: one that recommends which video model to use for a given goal, and one that translates a prompt written for one model's format into the correct format for a different model. Other skills handle structured prompt generation and debugging for the Seedance model specifically. A GitHub Actions workflow runs every week to check 32 official API endpoints across the covered models and opens an issue automatically when a version changes. This is meant to catch cases where a model update makes older prompts stop working as expected. To run the project locally you clone the repository and start a small Python server. The web interface is a single-file HTML application with no external dependencies. The project is MIT-licensed.