Downloads · 30 days
581
4% of all-time downloads
dog-god/texture-synthesis-sdxl-lora
texture-synthesis-sdxl-lora is a text-to-image model from dog-god. Use it when you need an image from a text prompt. It is set up for diffusers. The card lists the license as apache-2.0.
NOTE: The Inference API uses the 3D model by default, so if you want to generate 2D textures you may want to clone to a space or run the model yourself.
Downloads · 30 days
581
4% of all-time downloads
All-time downloads
13.4K
Public
Repo size
711 MB
Likes
45
Public
Click a slice to open those files.
.safetensors24.5 MB · 85%
From the Hugging Face model README
NOTE: The Inference API uses the 3D model by default, so if you want to generate 2D textures you may want to clone to a space or run the model yourself.
This is a SDXL LoRA specifically designed for generating material textures, as well as images of 3D objects with the textures already applied to them.
It's recommended to use controlnet for maximum consistency between images. The model is already in a standard safetensors format, and can be readily used in A1111, ComfyUI, or any other model inference API of your choosing. This model is part of the "Material synthesis with diffusion models" project by Artemiy Zhukov.
To represent different texture types, the model uses a two-token system for the type of texture.
heighmap - height texture
colormap - albedo texture
roughmap - roughness texture
normalmap - normal texture
specmap - specular texture
ambmap - ambient occlusion texture
metalmap - metallic texture
Here are some examples of the 3D model in action (although you may want to use the base if you just want top-down textures) <Gallery />