cclank/lanshu-awesome-ai-video-kit

What it solves

This project addresses the lack of structured, verified, and up-to-date prompt engineering resources for AI video generation. It solves three primary issues: the lack of source attribution for prompts, the rapid obsolescence of prompting techniques as models evolve, and the fragmented nature of prompt formulas across different AI video platforms.

How it works

The kit provides a comprehensive library of 543 tested prompts across 15 different AI video models (including commercial giants like Sora, Kling, and Veo, as well as open-source options like Mochi and CogVideoX). It organizes these prompts according to the official formulas of each model. To ensure the content remains current, it uses a GitHub Action to automatically monitor 32 official endpoints for version changes, triggering issues when updates are detected. Additionally, it includes "Claude Code Skills"—specialized configuration files that allow Claude to act as a model selector or prompt translator between different video AI platforms.

Who it’s for

It is designed for AI video creators, prompt engineers, and developers who need a reliable, model-specific framework for generating high-quality video content and want to avoid the trial-and-error process of using outdated prompts.

Highlights

• Extensive Model Coverage: Supports 15 models, including 11 commercial and 4 open-source/friendly options.
• Automated Monitoring: Weekly automated checks of 32 official endpoints to prevent prompt obsolescence.
• Prompt Translation: Tools and skills to translate prompts from one model's specific formula to another.
• Structured Methodology: 21 SOPs and methodology documents covering everything from basic formulas to advanced character consistency and FPV flight paths.
• Zero-Dependency Tools: Includes a single-file HTML prompt browser and a cross-model comparison matrix.