Summary

StableGen is an open-source Blender plugin that integrates generative AI for 3D texturing. It empowers 3D artists by leveraging advanced diffusion models (SDXL, FLUX.1-dev) via a ComfyUI backend, enabling complex, controllable, and scene-wide texture generation directly within Blender. The plugin streamlines the texturing workflow, offering sophisticated control over geometry and style.

How It Works

The plugin functions as an intuitive Blender interface that communicates with a ComfyUI backend for heavy AI computation. It utilizes diffusion models, integrating ControlNet for precise geometric adherence and IPAdapter for style guidance. Key features include scene-wide multi-mesh texturing, multi-view consistency achieved through inpainting and sophisticated blending techniques, and advanced inpainting/refinement tools for detailed control.

Quick Start & Requirements

Installation involves setting up a ComfyUI instance, running the provided installer.py script to fetch necessary ComfyUI custom nodes and AI models, and then installing the StableGen Blender plugin (StableGen.zip). Configuration requires specifying paths to the ComfyUI installation and desired output directories within Blender's add-on preferences.

• Primary Install: Install ComfyUI, run installer.py from the repository, install StableGen.zip plugin in Blender.
• Prerequisites: Blender 4.2+, Windows 10/11 or Linux, NVIDIA GPU with CUDA (recommended for ComfyUI backend performance), Python 3.x (for installer script), Git, significant disk space (10GB to 50GB+ for AI models).
• Links: ComfyUI GitHub: https://github.com/comfyanonymous/ComfyUI.
• Setup Footprint: Requires substantial disk space for ComfyUI, AI models, and generated textures.

Highlighted Details

• Scene-Wide Multi-Mesh Texturing: Apply textures to all visible scene meshes or selected objects simultaneously from defined camera viewpoints, ideal for cohesive environments or asset libraries.
• Multi-View Consistency: Achieved via Sequential Mode (using inpainting between views) or Grid Mode (faster previews with an optional refinement step), ensuring coherence across different angles.
• Geometric Control: Leverages multiple ControlNet units (Depth, Canny, Normal) simultaneously for precise texture adherence to model geometry, with adjustable strength and steps.
• Style Guidance: Utilizes IPAdapter for style transfer from reference images or for enhanced multi-view consistency without explicit reference images.
• Advanced Inpainting: Supports Refine Mode (Img2Img) for re-styling and detailing existing textures, and UV Inpaint Mode for intelligently filling untextured areas on UV maps.
• Integrated Workflow Tools: Includes utilities for camera setup, view-specific prompts, texture baking, HDRI lighting, applying modifiers, converting curves to mesh, and exporting animations (GIF/MP4).
• Preset System: Offers built-in presets for common scenarios and allows users to save and manage custom parameter configurations for repeatable workflows.

Maintenance & Community

The project originated as a Bachelor's Thesis by Ondřej Sakala. A "List of planned features" indicates ongoing development. Contact information (email, X/Twitter) is provided. No explicit community channels like Discord or Slack are listed in the README.

Licensing & Compatibility

• License: GNU General Public License v3.0 (GPLv3).
• Compatibility: GPLv3 is a strong copyleft license, requiring derivative works to be shared under the same terms. This may impose restrictions on integration into closed-source commercial products.

Limitations & Caveats

Experimental support is available for the FLUX.1-dev model architecture. Remote ComfyUI server connections are not currently supported but are listed as a planned future feature. The plugin requires Blender version 4.2 or newer. An NVIDIA GPU with CUDA is recommended for optimal performance of the ComfyUI backend.