🧩 AI 3D Scene Generator

A browser-based Three.js playground that uses LLMs (OpenAI, OpenRouter, Anthropic) to generate, modify, and export 3D objects — directly inside the browser.
Built with asyncLLM streaming, Bootstrap UI, and a sandboxed iframe runtime that ensures safe execution of AI-generated code.

---

🚀 Features

• LLM Integration

  • Configure providers (OpenAI, OpenRouter, Anthropic) via API key
  • Supports GPT-4.1, GPT-5, and Anthropic Claude models
  • AI generates or modifies Three.js scenes from text prompts

• Interactive 3D Rendering

  • Real-time Three.js rendering inside an iframe sandbox
  • Auto-rotate, wireframe toggle, and camera reset controls
  • Screenshot capture + display for reference

• Reference Image Support

  • Upload a reference image to guide LLM-based geometry creation
  • Side-by-side comparison with generated output

• Export Options

  • Export generated 3D objects as .obj files
  • Session history with prompt, code, and screenshots

• UI/UX

  • Bootstrap 5 + Bootstrap Alerts for responsive UI
  • Highlight.js + Marked for syntax-highlighted code previews

---

⚡ How It Works

flowchart TD
    A[📝 User enters prompt] --> B[🖼️ Upload reference image]
    B --> C[⚡ Send prompt + reference to LLM]
    C --> D[💻 LLM generates Three.js code]
    D --> E[🎨 Render 3D object in viewer]
    E --> F{User actions}

    F -->|📦 Export OBJ| G[💾 Save/Download file]
    F -->|🔄 Modify object| H[📝 Enter new prompt]
    
    H --> I[📷 Capture current 3D screenshot]
    I --> J[⚡ Send new prompt + screenshot + reference image to LLM]
    J --> D

Loading

🔄 Workflow

1. Input – Enter prompt + (optional) reference image.
2. LLM – Prompt + image → LLM → Three.js code.
3. Render – Code runs in iframe → 3D object with controls.
4. Export – Save as OBJ or screenshot.
5. Iterate – New prompt + current screenshot + reference → updated code → repeat.

---

⚡ Setup & Usage

1. Clone the repository

git clone https://github.com/Nitin399-maker/3dobjectgen.git
cd 3dobjectgen