Generate 3D models, video, and audio with AI agents
Imagine a coding agent directly manipulating professional software like Blender or Maya to complete game assets, video, and audio. This repository is a collection of skills that helps AI agents select the correct model, verify costs, and generate content while maintaining character and brand consistency. It is provided in a format ready for use in tools like Claude Code or Cursor.
Difference between professional tools and general skills
The skills in this repository are divided into two main types. The first is general skills that generate images, video, and 3D assets through the Scenario MCP server. The second is expert tools that directly control DCC (Digital Content Creation) software and game engines installed by the user, such as ZBrush, Blender, Maya, Unreal Engine, and Unity. Expert tools require the specific application to be installed rather than an MCP server to function, and the verification level is recorded in the README of each folder.
Installation and update methods
To install all Scenario skills, run the npx skills add scenario-labs/skills command. To select only skills suitable for a specific role or purpose, you can use the group-specific commands in INSTALL.md. Expert tools are installed in family units; select the corresponding group in the installer menu or run the commands in the family README. To keep up to date, run npx skills update or re-run the original add command. To pin to a specific reviewed version, install by specifying a commit obtained from a release tag.
Main skill categories
Skills are organized into various categories including generation direction setting, image, game art and 3D, video and audio, social media, consistency and custom models, output review and cleanup, format and batch, workflow, troubleshooting, team management, and model families. For example, scenario-3d generates 3D meshes from text or images, scenario-blender controls Blender scenes, and scenario-video handles generation and editing from text to video. Each skill is designed for specific tasks, allowing you to selectively use only the necessary features.
Pre-implementation checks
To use this repository, the Scenario MCP server must be connected, and you must log in with OAuth or an API key. To use expert tools, the corresponding DCC software must be installed locally. For ZBrush tools, live testing is complete, but full live testing for all skills is still pending. Blender tools have been confirmed to pass 12 test suites headlessly on Blender 5.2.1. Before use, you must always check the requirements and verification status in the README of each skill.
