Text-to-3D Generator

Converts text descriptions into 3D models using:

• AWS Bedrock Stability AI Stable Image Core for text → image generation
• Hi3DGen (hosted on EC2 GPU instance) for image → 3D model generation

Architecture

[User Text Prompt] → [Bedrock Stable Image Core] → [Generated Image] → [Hi3DGen on EC2] → [3D Model (.glb)]

Components

• backend/bedrock_image_gen.py - Bedrock Stable Image Core integration (text → image)
• backend/hi3dgen_server.py - FastAPI server wrapping Hi3DGen (runs on EC2)
• frontend/app.py - Gradio UI with both Text-to-3D and Image-to-3D tabs
• infra/deploy.py - EC2 deployment (g5.2xlarge with NVIDIA A10G GPU)
• infra/teardown.py - Clean up all AWS resources

Quick Start

1. Install local dependencies

pip install -r requirements.txt

2. Deploy Hi3DGen to EC2 (if not already deployed)

python infra/deploy.py

3. Launch the frontend

python frontend/app.py

The Gradio UI opens at http://localhost:7860 with two tabs:

• Text to 3D: Enter a text prompt → generates image → converts to 3D
• Image to 3D: Upload an image → converts to 3D (original Hi3DGen flow)

Teardown

python infra/teardown.py

Configuration

The config.json file (auto-generated by deploy.py) stores the EC2 instance details. You can also set HI3DGEN_API_URL environment variable to override.