anyCreature by Gobkit

Product Introduction

1. Definition: anyCreature by Gobkit is an open-source, AI-native 3D asset generation engine. It is a procedural content generation (PCG) tool that compiles a single JSON specification into a fully rigged, textured, and animated 3D creature model in the GLB format, ready for game engines and real-time applications.
2. Core Value Proposition: It exists to automate and accelerate the 3D character creation pipeline for game developers, digital artists, and AI agents. Its primary value is delivering game-ready 3D creatures from a simple text prompt in a single, automated session, eliminating the need for manual modeling, UV unwrapping, rigging, and animation.

Main Features

1. Text-to-3D Creature Pipeline: The system operates through a structured, five-stage AI session (cards/00_START to 04_SHIP). An AI agent takes a natural language order (e.g., "menacing mountain giant"), conducts a brief interview (≤2 questions), and autonomously executes the design, validation, and delivery process. The output is a production-ready GLB file with vertex colors, baked ambient occlusion (AO), and animations.
2. Context-Free Quality Gates: A core innovation is its objective quality control. Instead of self-assessment, the generated creature silhouette passes through two gates evaluated by a separate, context-free AI reader. Gate 1 (RECOGNISED) ensures the silhouette is universally identifiable from four views. Gate 2 (PUNCHIER) ensures iterative design changes only make the silhouette bolder, preventing over-correction and loss of character.
3. Zero-Dependency Compiler Engine: The core engine/cli.js is a standalone Node.js module with no external runtime dependencies. It ingests a structured JSON spec and deterministically outputs a skinned GLB. It enforces "engine floors" (BLOCK:, warn:, info:) that catch critical failures like faceted meshes, broken skinning from mirror distortion, or detached parts, ensuring technical integrity.
4. Automated Metric-Driven Design: The harness includes tools (harness/silmetrics.mjs, maskmetrics.py) that automatically render 4-view silhouettes, compute layout metrics (width/height ratio, mass distribution, silhouette complexity), and flag design dullness (excessive symmetry, straight lines). This provides quantitative feedback for the AI designer.
5. Integrated Delivery Packager: The harness/deliver.py script finalizes the asset by stamping metadata into the GLB, generating promotional renders (hero.png, hero.jpg), and packaging an offline HTML showroom viewer, creating a complete delivery package from a single command.

Problems Solved

1. Pain Point: The 3D character modeling and rigging process is notoriously time-consuming, technically complex, and requires specialized artistic and technical skills (modeling, topology, UV mapping, rigging, skinning). This creates a major bottleneck in game development and digital content creation.
2. Target Audience: Indie Game Developers & Small Studios needing rapid creature prototyping; Technical Artists & Vibe Coders looking to integrate 3D generation into automated pipelines; AI Agent Developers requiring a reliable, API-like 3D asset generation module for their autonomous systems.
3. Use Cases: Rapid generation of enemy variants (minions, NPCs, bosses) for game jams or production; populating virtual worlds with diverse fauna; providing a reliable 3D asset generation component within a larger AI agent workflow; creating concept models for character design pre-visualization.

Unique Advantages

1. Differentiation: Unlike general-purpose 3D AI models that output unrigged, untextured meshes or video clips, anyCreature is a specialized engine that guarantees a technically valid, game-engine-ready asset. It contrasts with traditional manual pipelines by being fully automated and with cloud-based generative services by being open-source and locally executable.
2. Key Innovation: The "clean painter, strict inspector" doctrine. The AI designer operates with minimal constraints (only the order and a trap map), fostering bold, creative designs. A separate, rigorous, metric-driven inspection system (the gates and engine floors) ensures quality and technical compliance. This separation of creation and validation is fundamental to its reliability.

Frequently Asked Questions (FAQ)

1. What 3D file format does anyCreature output? anyCreature outputs industry-standard GLB files (binary glTF), which are universally supported by modern game engines (Unity, Unreal Engine, Godot) and web-based 3D viewers, complete with skeletal animations, vertex colors, and materials.
2. Do I need 3D modeling or coding skills to use anyCreature? The primary interface is designed for AI agents via the structured card system. Users need basic command-line skills to run the setup and harness scripts. While creating custom JSON specs requires understanding its schema, the tool is aimed at integration into automated, agent-driven pipelines rather than direct manual use.
3. Can I use anyCreature commercially for my game? Yes. The anyCreature engine is released under the permissive MIT license, allowing for unrestricted commercial use, modification, and distribution. You own the assets you generate with it.
4. What are the system requirements to run anyCreature locally? You need Node.js 18+ and Python 3.9+. The core engine has zero dependencies, but the full tooling for measurement and rendering requires a local Chromium instance (installed via Playwright by setup.sh) for silhouette generation and metric calculation.
5. How does anyCreature ensure the generated creature is "game-ready"? It enforces technical validity through its compiler "floors" that block malformed geometry, ensures proper skeletal rigging and skinning weights, bakes ambient occlusion, and outputs optimized vertex counts. The provided example (wolf.json) produces a model with 2,211 vertices and 31 joints, demonstrating real-time readiness.