Skip to content

🟠 Generate 3D models using Gradio API directly from Blender

License

Notifications You must be signed in to change notification settings

gsbm/autosculptor

Folders and files

NameName
Last commit message
Last commit date

Latest commit

Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 

Repository files navigation

🟠 Generate 3D models using Gradio API directly from Blender


Features

  • βš™οΈ Multiples Models: Access different model pipelines for 3D model generation.
  • ✨ Prompt Enhancer: Enhance the prompt for better results adapted to 3D generation.
  • 🏭 Batch Generation: Generate multiples differents models in a row (if random seed enabled).
  • πŸ”— Threading: Run the model generation in a separate thread (to avoid the UI to freeze).

Get started

Installation

Download latest release in releases page. Then, in Blender, go to Preferences > Add-ons > Install and select autosculptor.zip. To activate it, check it in the add-on list.

This add-on integrate a built-in dependencies installer. At first use, it will ask the user the consent to install complementary dependencies. This step will require internet to download and install required packages. In case of error, or for manual dependencies installation, please refer to troubleshooting dependencies section.

Usage

After installation, Autosculptor is available in Blender on View3D > Sidebar > Autosculptor. Please note that this add-on require internet to execute API requests.

Input Type Description
Prompt String The text prompt describing the 3D model to generate
Model Select Model pipeline to use for generation
Prompt Enhancer Boolean Enhance the prompt for better results
Seed Integer Seed for generation
Random Seed Boolean Use a random seed for each generation
Guidance Scale Integer Scale for the guidance during generation
Inference Steps Integer Number of inference steps for generation
Apply Material Boolean Apply material to the generated model
Run in Thread (experimental) Boolean Run the model generation in a separate thread
Batch Count Integer Number of 3D models to generate
HF Token String User Access Token for Hugging Face to get a higher priority in queues

Troubleshooting

Dependencies

This addon the use third-party Gradio Python module. If a dependency error still occurs after complete installation, please try to reinstall Gradio module.

In Blender Python console:

>>> import sys
>>> sys.exec_prefix
'/path/to/blender/python'

Then in a shell (running as administrator on Windows):

# change the working directory to Blender python bin path
$ cd /path/to/blender/python/bin

# check if pip is installed
$ ./python -m ensurepip

# uninstall Gradio module (if applicable)
$ ./python -m pip uninstall gradio_client

# install Gradio module
$ ./python -m pip install gradio_client

If you still have trouble installing Gradio to Blender python, please check this StackExchange thread.

APIs

Please note that services availability cannot be guaranteed at all times. This add-on relies entirely on community APIs, operated through Gradio clients. These APIs are hosted on Hugging Face and are therefore subject to the vagaries of (rare) outages. You can find out more about the status of services on their own page.

In addition, some Spaces on which APIs are hosted can also be paused or put on standby at any time. To manually check the status of an individual service, please refer to the list of API hosts.

Available models

Model Pipeline Details API Host(s) Average generation time
Shap-E Shap-E (mesh) hysts/Shap-E ~13s
SDXL + Shap-E SDXL (image) > One-2-3-45 (preprocess) > Shap-E (mesh) hysts/SDXL, hysts/Shap-E, One-2-3-45/One-2-3-45 ~30s
SDXL + DreamGaussian SDXL (image) > One-2-3-45 (elevation estimation) > DreamGaussian (mesh) hysts/SDXL, jiawei011/dreamgaussian, One-2-3-45/One-2-3-45 ~600s
SDXL + InstantMesh SDXL (image) > InstantMesh (mesh) hysts/SDXL, TencentARC/InstantMesh ~60s
SDXL + TripoSR SDXL (image) > TripoSR (mesh) hysts/SDXL, stabilityai/TripoSR ~30s

Examples

Shap-E SDXL + Shap-E SDXL + DreamGaussian SDXL + InstantMesh SDXL + TripoSR
A pinguin, 3d model
A hamburger, 3d model
A chair, 3d model

Prompt Enhancer

Prompt enhancer requests gustavosta/magicprompt-stable-diffusion for generating an extended prompt more adapted to 3D model generation. Please note that Prompt Enhancer can sometimes deviate slightly from its original subject.

Prompt Model Standard Prompt Enhanced
A orange cat SDXL + InstantMesh
A orange cat, photorealistic SDXL + InstantMesh

Implementations

Currently working on other implementations like :

  • Adding image to Mesh option
  • Updating displayed parameters on differents models
  • Thinking about self hosting/using more stable hosted models
  • Adding preset system
  • Adding generation progressbar/estimated generation time (through Gradio Status ETA)
  • Making pipelines more modulable

License

This project is licensed under GPL-3.0 License.

Warning

Before using this addon, please be aware that different models used in this project may have their own specific licenses and usage terms. It is your responsibility to review and comply with the licenses of each model to ensure that your use case is permitted. Failure to do so may result in legal consequences. Always verify the licensing terms of the models you intend to use. This program is provided for educational and experimental purposes.