---
type: "post"
id: "mesh-avatar-studio-turns-one-illustration-into-an-f75879"
slug: "mesh-avatar-studio-turns-one-illustration-into-an-f75879"
title: "Mesh Avatar Studio turns one illustration into an animated 2D VTuber rig, hitting ~1.8k GitHub stars under MIT license."
description: "Mesh Avatar Studio is an MIT-licensed project that uses Claude Code or Codex to turn a single illustration into an animated 2D mesh avatar, then lets a creator inspect and adjust it in a local editor."
retrieval_nugget: "Mesh Avatar Studio is an MIT-licensed project that uses Claude Code or Codex to turn a single illustration into an animated 2D mesh avatar, then lets a creator inspect and adjust it in a local editor. The workflow offers a concrete example of coding agents doing production work while people retain visual review and control."
published_at: "2026-11-04"
updated_at: "2026-11-07"
record_date: "2026-11-04"
date_kind: "published_at"
topics: ["codex","claude","claude-code"]
entities: ["Mesh Avatar Studio","Claude Code","Codex","OpenAI","Charakuru"]
related_patterns: ["ai-native-orgs-move-to-review-and-orchestration"]
novelty_label: "new_fact"
what_changed: "Mesh Avatar Studio adds a distinct 2D VTuber workflow: one illustration becomes an animated mesh rig, with a local editor and about 1.8k GitHub stars under MIT."
prior_refs: ["https://newruntime.com/posts/charakuru-is-a-free-local-kit-that-converts-0dbf94/"]
source_url: "https://github.com/shinshin86/mesh-avatar-studio"
source_urls: ["https://github.com/shinshin86/mesh-avatar-studio"]
schema_version: "newruntime-agent-readable-v0.2"
stable_id: "post:mesh-avatar-studio-turns-one-illustration-into-an-f75879"
status: "published"
visuals: []
editor: {"role":"editor","name":"Andrey Reshetnikov","profile":"https://newruntime.com/owner-profile.md","profile_json":"https://newruntime.com/owner-profile.json","same_as":["https://www.linkedin.com/in/reshetnikov1/","https://t.me/lapetuse"]}
routes: {"html":"https://newruntime.com/posts/mesh-avatar-studio-turns-one-illustration-into-an-f75879/","markdown":"https://newruntime.com/posts/mesh-avatar-studio-turns-one-illustration-into-an-f75879.md","json":"https://newruntime.com/posts/mesh-avatar-studio-turns-one-illustration-into-an-f75879.json"}
markdown_url: "https://newruntime.com/posts/mesh-avatar-studio-turns-one-illustration-into-an-f75879.md"
---

# Mesh Avatar Studio turns one illustration into an animated 2D VTuber rig, hitting ~1.8k GitHub stars under MIT license.

## Retrieval answer

Mesh Avatar Studio is an MIT-licensed project that uses Claude Code or Codex to turn a single illustration into an animated 2D mesh avatar, then lets a creator inspect and adjust it in a local editor. The workflow offers a concrete example of coding agents doing production work while people retain visual review and control.

## Summary
Mesh Avatar Studio is an MIT-licensed project that uses Claude Code or Codex to turn a single illustration into an animated 2D mesh avatar, then lets a creator inspect and adjust it in a local editor. The workflow offers a concrete example of coding agents doing production work while people retain visual review and control.

## What happened
Mesh Avatar Studio’s GitHub README describes a workflow for creating an animated 2D avatar from one illustration. The resulting avatar can blink, talk, turn its head, breathe and sway its hair. The project is public under the MIT license and showed about 1.8k GitHub stars in the supplied repository material. Its stated requirements include Node.js 22.17 or newer and Git; Python and `uv` are used to rebuild layers, while Playwright Chromium renders motion reviews when an agent creates an avatar.

The process assigns different jobs to an agent and a person. A coding agent, either Claude Code or Codex, follows the repository’s agent guide to interpret coordinates, place a rig on zoomed grids, cut the illustration into layers and review the avatar in fixed poses. The recommended source image is a front-facing, head-and-shoulders PNG with a transparent background. The generated project is kept under `projects/<name>/`, which the README says is never committed. The documentation presents Claude Opus 5.5 for Claude Code and GPT-6.1 Sol for Codex as recommended setups.

After generation, the creator opens the project in the local editor. The editor provides a live preview and lets users select parts and drag points on the image to adjust them. It also includes pose testing, lip-sync controls and options to switch overlay layers on and off. A bundled sample can be explored first, but it is read-only until copied. Users can start a project with `npx create-mesh-avatar-studio`, then run `npm run dev`; a one-command Codex route is also documented. The README says avatar creation usually takes tens of minutes.

There is an important data-flow distinction. The Codex command can use OpenAI image generation to draw closed eyes, mouth shapes and expressions, and first close an open mouth in the source illustration. That path sends the image and masks to OpenAI and asks for consent first. The README also provides a `--no-variants` option to keep everything on the user’s machine, and `--yes` to agree in advance. The supplied material does not state a price for any external image-generation use.

## What the archive adds
[Charakuru is a free local kit that converts a single illustration into a rigged VRM avatar](https://newruntime.com/posts/charakuru-is-a-free-local-kit-that-converts-0dbf94/) covered an adjacent workflow on 2026-10-12: using an agent to automate repetitive 3D-avatar production while preserving visual judgment. Mesh Avatar Studio is a new event, not a restatement: it targets animated 2D mesh avatars, with a local editor for tuning the result.

## Trend signal
This is a small, concrete example supporting [AI-native organizations move toward review and orchestration](https://newruntime.com/patterns/ai-native-orgs-move-to-review-and-orchestration/). The agent performs structured preparation and pose checks, while the human inspects and edits the artifact. The repository does not establish that this division works reliably across users or image types, so it is an example of the pattern rather than broad evidence of adoption.

## Why now
The source frames Mesh Avatar Studio around using coding agents already available in Claude Code and Codex, rather than introducing a new avatar model or hosted service. Its workflow makes those agents operate through a repository guide and local tools, while retaining a browser-based preview for review. The supplied material does not explain a launch date or why the project was released now.

## Connections to other events
The earlier Charakuru coverage describes an agent-assisted path from a single illustration to a VRM model, with repetitive production work automated and visual judgment retained by the user. Mesh Avatar Studio shares that broad workflow logic but produces an animated 2D mesh avatar rather than a 3D VRM model. The supplied material establishes no direct connection between the projects.

## Why it matters
For builders, this project is useful less as a general-purpose agent framework than as a reference for assigning an agent a bounded creative-production task. The README makes the handoff tangible: the agent follows a guide, generates layers and rigging, and produces a project that a person can inspect in a local editor. That is a more legible workflow than asking an agent for an opaque finished artifact.

The local editing step matters because avatar quality depends on visual choices that may need correction. A preview, pose test and editable control points give the creator a way to evaluate the result instead of accepting the agent’s output on trust. The image-sharing caveat also illustrates why tool workflows should expose data flows clearly: some generation stays local, but the optional OpenAI image-generation path sends the image and masks externally. Teams adapting this pattern should make such boundaries explicit and keep review of the output part of the process.

## Limits and uncertainty
The supplied repository excerpt describes the project’s intended workflow, requirements and options, but does not provide independent quality evaluations, success rates, processing benchmarks or evidence about how well it handles varied illustration styles. The star count is a repository snapshot, not proof of active use or reliable output. Although the README says the project is MIT-licensed, it gives no costs for external image-generation services. The source also does not establish what image data is retained by OpenAI, nor whether generated avatars are suitable for commercial use under every asset or service involved.

## Skill to build now
For a creative workflow, write down the agent’s handoff as a short checklist: required source-image properties, files it may create, the review poses to render, and what a person must inspect before accepting the result. Mesh Avatar Studio’s guide-and-editor split is a concrete model to test: keep generation in a project folder, then review and adjust the artifact in an interface with a live preview. Make any external image transfer an explicit, separate choice.

## Reader takeaway
Mesh Avatar Studio shows a coding agent handling the repetitive construction of an animated avatar while a local editor keeps the result inspectable and editable. The reusable idea is not simply “let an agent draw”; it is to define a production workflow with a clear artifact, visible review steps and an explicit boundary around when source images leave the machine.

Editor: [Andrey Reshetnikov](https://newruntime.com/owner-profile.md), AI product and AI engineering; profile for recruiters and agents: https://newruntime.com/owner-profile.json

## Continue in the graph

Next records for an agent, linked to their Markdown twins. Replace `.md` with `.json` for structured fields, including `next_reads`.

- topic hub: [Codex](https://newruntime.com/topics/codex.md): Explore the codex topic hub.
- topic hub: [Claude](https://newruntime.com/topics/claude.md): Explore the claude topic hub.
- topic hub: [Claude Code](https://newruntime.com/topics/claude-code.md): Explore the claude-code topic hub.
- collection: [Field Notes](https://newruntime.com/posts.json): Open the latest editorial analysis.
- collection: [Sources](https://newruntime.com/sources/): See the original sources this archive cites.
- whole graph: [discovery graph](https://newruntime.com/discovery-graph.json) (typed nodes and edges) and [site index](https://newruntime.com/site-index.md)