TEXT2VOX: 3D Voxel Model Generator

TEXT2VOX is a Cog service that generates MagicaVoxel VOX files from text prompts. It combines state-of-the-art AI models to create detailed voxel models that can be used in games, 3D visualization, and creative projects.

Overview

This pipeline uses two AI models in sequence:

1. Flux - A text-to-image model that generates a 3D-style reference image
2. Hunyuan3D-2 - A 3D model generator that creates a GLB file from the image
3. GLB2VOX Converter - Converts the GLB model to MagicaVoxel VOX format at different resolutions

Features

• Generate voxel models from simple text descriptions
• Control the detail level (high or low)
• Multiple resolutions output simultaneously
• Includes both VOX files and GLB/GLTF versions

Usage

Basic Parameters

Detail Level and Resolutions

The service provides different resolution outputs based on the chosen detail level, which drives the generation:

• High Detail: Optimized for detailed models

  • Large: 96×96×96 voxels
  • Medium: 80×80×80 voxels
  • Small: 64×64×64 voxels

• Low Detail: Optimized for blocky, Minecraft-style models

  • Large: 64×64×64 voxels
  • Medium: 48×48×48 voxels
  • Small: 32×32×32 voxels

Output Files

For each generation, you'll receive:

• Multiple VOX files at different resolutions
• A GLB file representation of the voxel model
• A GLTF file for web viewing

All files are named based on your prompt (converted to snake_case) with size suffixes.