What it is
Mesh Avatar Studio is an open-source avatar tool from Yuki Shindo (GitHub user shinshin86). You hand an illustration to Claude Code or Codex, and the coding agent follows the project's guide to cut the image into layers and place the eyes, mouth and hair; in the README's words, "A coding agent prepares the avatar from your image." You then check and fine-tune it by dragging in a local editor, and the avatar can blink, talk, turn its head, breathe and sway its hair. For streaming, it can follow your face through a webcam or lip-sync from your microphone alone, and plugs into OBS as a transparent overlay, or into Zoom and Google Meet through OBS Virtual Camera. What sets it apart is that one picture is enough, with no extra sets of expressions to draw; if you want drawn expressions, AI can add those too.
Who it's for:People who want an animated avatar for streaming or video calls but don't know how to build layered Live2D-style models.
Before you start: You install Node.js 22.17+ yourself and run it with npm, and you need Claude Code or Codex to prepare the avatar. The best source image is a front-facing, head-and-shoulders PNG with a transparent background. The README says webcam video and microphone audio stay on your machine and only motion values reach the stream view; if you use a cloud AI to generate expressions, your image goes to that provider. The code is MIT-licensed; the bundled sample character Miko has its own terms, which allow using and modifying her inside your own work but not redistributing her on her own.
Based on the project's public documentation. Not tested by AI Inspiration World.