Generate and texture 3D assets inside Blender using AI backends

What it does

StableGen is a Blender addon that bridges your scene to generative AI backends. It can generate fully textured 3D meshes from a single image or text prompt using Microsoft’s TRELLIS.2 model, then texture or refine existing models via SDXL, FLUX.1-dev, or Qwen Image Edit through a ComfyUI server. The addon handles scene-wide texturing, applying materials to all mesh objects simultaneously from multiple camera viewpoints.

The interesting bit

The addon uses seven camera placement strategies—including orbit rings, hemisphere coverage, and greedy occlusion passes—to capture unlimited viewpoints with per-camera aspect ratios, then stitches the results together with weighted blending and color matching. This turns environment-wide texturing from a manual export-and-paint workflow into a single automated pass inside Blender.

Key highlights

• Generate 3D assets from images or prompts via TRELLIS.2 (4B parameters) with resolution modes up to 1536 Cascade.
• Texture entire scenes at once using multi-view consistency with sequential or grid processing modes.
• Seven camera placement strategies with per-camera optimal aspect ratios and no camera limit.
• Local Edit mode for targeted touch-ups and UV Inpaint mode to fill untextured areas using surrounding context.
• ControlNet (Depth, Canny, Normal) and IPAdapter support for geometric precision and style guidance.
• Optional PBR material map generation via Marigold decomposition.

Caveats

• Requires a separate ComfyUI server; it is not a standalone tool.
• FLUX.1-dev support is explicitly marked as experimental.
• TRELLIS.2 is a 4B-parameter model, so VRAM and disk offloading considerations apply even with the addon’s optimizations.

Verdict

A solid pick for Blender look-dev and concept artists who want generative AI inside their viewport. Less useful if you need a fully self-contained package or lack the GPU resources to run ComfyUI and large diffusion models locally.