Practical overview
When it fits
Use it when you want to investigate or run a local research workflow for text-guided, localized 3D mesh texturing and can provide the documented GPU and software environment.
Who it is for
- 3D graphics researchers exploring text-guided mesh editing.
- Developers and technical artists who can run Python and CUDA research code locally.
- Users studying cascaded diffusion guidance for 3D texturing or the included CSD image-generation and editing demo.
Documented uses
- Apply a text-described appearance change to a semantic region of a 3D mesh.
- Generate a localized texture map intended to integrate with standard graphics pipelines.
- Reproduce the repository’s pre-trained mesh-editing examples.
- Inspect a notebook demo of Cascaded Score Distillation for image generation and editing.
Requirements & limits
The repository is public and lists an MIT license. The documented workflow is local and requires Python 3.10, CUDA 11.3 and GPU access; the README specifies 48 GB of GPU memory and refers users with less memory to its optimization guidance. DeepFloyd IF use with Diffusers requires a Hugging Face account, model-card license acceptance and local login.
- The method targets local semantic regions on meshes; the README describes a separate global-texturing implementation maintained in another repository.
- The published setup has substantial GPU-memory requirements, with additional memory-optimization guidance for lower-memory systems.
- Pre-trained reproduction uses downloaded trained_models and inverse_map_cache folders alongside mesh data and per-edit configuration files.
- The CSD demo notebook illustrates image generation and editing, while the main project applies the technique to mesh texturing.
Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.
Set up the local 3D Paintbrush environment
Create the documented Python environment and install the project’s required packages.
Before you start
- Python 3.10
- CUDA 11.3
- GPU access while running the install script; the README specifies 48 GB GPU memory and points to memory optimization for lower-memory systems.
Steps
Create a Conda environment with Python 3.10: conda create -n "3d-paintbrush" python=3.10.
Activate it: conda activate 3d-paintbrush.
Run the repository install script with GPU access: bash ./install_environment.sh.
Expected outcome
The project environment and required packages are set up; the README proceeds to model access and reproducing paper results.
Guide limitations
- This recipe covers setup only; it does not verify successful installation or model execution.
- The README lists a 48 GB GPU requirement and notes a separate memory-optimization section for systems with less memory.
Prepare Hugging Face access for DeepFloyd IF
Authenticate locally to use DeepFloyd IF with Diffusers.
Before you start
- A Hugging Face account and access token.
- Acceptance of the license on the DeepFloyd/IF-I-XL-v1.0 model card.
- The Python environment used with the project.
Steps
Accept the license on the DeepFloyd/IF-I-XL-v1.0 model card.
Install or upgrade huggingface_hub with pip install huggingface_hub --upgrade.
In a Python shell, run from huggingface_hub import login followed by login(), then enter the Hugging Face access token.
Expected outcome
The local Hugging Face login is configured for the documented DeepFloyd IF and Diffusers use.
Guide limitations
- This covers authentication prerequisites, not a complete texturing run.
- Model-card license acceptance and an account are required for this documented model-access path.
Sources & review
Source record details
- Source availability
- Active
- Sources checked
- 17 October 2026
- Original website title
- threedle/3d-paintbrush · Official homepage title · retained in its source language