Charakuru Kit Builds VRM Avatars From One Illustration With an AI Agent
The kit gives avatar production a defined checkpoint structure where appearance decisions stay with the user and the agent does the file work, which is the part of AI-assisted 3D that has been least standardized.
Reporting from 1 source: GIGAZINE.
A GitHub kit called Charakuru turns a single full-body character illustration into a VRM 1.0 avatar by splitting the work between the user and an AI agent. Developer Nano's procedure runs through a local confirmation app and combines the 3D service Tripo with Blender. The agent generates eight directional head and body images, the user picks from four model candidates each, and the final avatar is exported after a walking-motion check.
The app runs locally, but the generation steps do not: image and 3D work go through outside services, so each service's terms, fees, and credits apply on their own. Setup needs Node.js 22 or later, a desktop Chromium-based browser such as Chrome, Codex with image generation and browser control, a Tripo account, plus Blender 5.2, VRM Add-on for Blender, and Blender MCP.
Expressions for newly created characters are limited to a normal face, with no expression switching or eye movement. Redistribution requires prior consultation, and if a company's annual revenue or cumulative funding reaches 100 million yen or more, commercial and business use needs a separate written license agreement.
Synthesized by Yomimono from the 1 cited source below, including Japanese-language reporting where cited, then editorially reviewed before publishing.