https://github.com/huggingface/meshgen
Use AI Agents directly in Blender.
https://github.com/huggingface/meshgen
Last synced: 3 months ago
JSON representation
Use AI Agents directly in Blender.
- Host: GitHub
- URL: https://github.com/huggingface/meshgen
- Owner: huggingface
- License: mit
- Created: 2024-10-14T17:28:53.000Z (about 1 year ago)
- Default Branch: main
- Last Pushed: 2025-04-14T20:00:02.000Z (9 months ago)
- Last Synced: 2025-09-30T18:02:29.612Z (3 months ago)
- Language: Python
- Homepage:
- Size: 15.3 MB
- Stars: 655
- Watchers: 17
- Forks: 40
- Open Issues: 12
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# MeshGen
Use AI Agents to control Blender with natural language.
## Features
- Focused on AI as a tool, not replacement
- Multiple backend options:
- Local inference with [llama.cpp](#local-backend) or [Ollama](#ollama)
- Remote inference with [Hugging Face](#hugging-face), [Anthropic](#anthropic), or [OpenAI](#openai)
- Optional [LLaMA-Mesh](#llama-mesh) integration for local mesh understanding and generation
- Optional [Hyper3D](#hyper3d) integration for high-fidelity 3D mesh generation
## Installation
1. Go to the [Latest Release](https://github.com/huggingface/meshgen/releases/latest) page
2. Download the addon ZIP file for your platform
3. In Blender, go to `Edit` -> `Preferences` -> `Add-ons` -> Top-Right Arrow -> `Install from Disk...`
4. Select the downloaded ZIP file
## Setup
1. In blender, go to `Edit` -> `Preferences` -> `Add-ons` -> `meshgen`
2. Select either [`Local`](#local-backend) or [`Remote`](#remote-backend) and follow instructions below
### Local Backend
Run models free locally directly in Blender.
Only select this option if you:
1. Have a powerful NVIDIA GPU with at least 8GB of VRAM
2. Installed a `cuda` version of the addon during [Installation](#installation)
3. Prefer running the model directly in Blender instead of a local [Ollama](#ollama) server
To set up the local backend, you can either:
1. Click `Download Recommended Model` to download [Meta-Llama-3.1-8B-Instruct-GGUF](https://huggingface.co/bartowski/Meta-Llama-3.1-8B-Instruct-GGUF)
2. Manually download a `.GGUF` model and put it in the models folder (located by clicking the folder icon)
### Remote Backend
MeshGen supports a variety of remote backends.
1. [Ollama](#ollama) to run models in a free local server
2. [Hugging Face](#hugging-face), [Anthropic](#anthropic), or [OpenAI](#openai) to run powerful models via API
[Hugging Face](#hugging-face) is recommended for most users, providing limited free use of powerful models.
#### Ollama
Run models free locally with an Ollama server.
1. Install [Ollama](https://ollama.com/)
2. Run `ollama serve` in the terminal
3. Select `Ollama` in the `Provider` dropdown
4. Enter your `Ollama` server endpoint and model name (the defaults should work for most users)
#### Hugging Face
Run a wide variety of models such as Llama, DeepSeek, Mistral, Qwen, and more via the Hugging Face API.
1. Create an account on [Hugging Face](https://huggingface.co/)
2. Go to [hf.co/settings/tokens](https://hf.co/settings/tokens) and create a new token
3. Select `Hugging Face` in the `Provider` dropdown
4. Enter your `Hugging Face` token in the `API Key` field
5. Optionally, change the `Model ID` to the model you want to use (e.g. `meta-llama/Llama-3.3-70B-Instruct`)
#### Anthropic
Run Anthropic models (i.e. Claude) with the Anthropic API.
1. Create an account on [Anthropic](https://console.anthropic.com/)
2. Go to [console.anthropic.com/settings/keys](https://console.anthropic.com/settings/keys) and create a new key
3. Select `Anthropic` in the `Provider` dropdown
4. Enter your `Anthropic` key in the `API Key` field
5. Optionally, change the `Model ID` to the model you want to use (e.g. `claude-3-5-sonnet-latest`)
#### OpenAI
Run OpenAI models (i.e. ChatGPT) with the OpenAI API.
1. Create an account on [OpenAI](https://platform.openai.com/)
2. Go to [platform.openai.com/api-keys](https://platform.openai.com/api-keys) and create a new secret key
3. Select `OpenAI` in the `Provider` dropdown
4. Enter your `OpenAI` secret key in the `API Key` field
5. Optionally, change the `Model ID` to the model you want to use (e.g. `gpt-4o-mini`)
## Optional Integrations
To enable optional integrations, go to `Edit` -> `Preferences` -> `Add-ons` -> `meshgen` -> `Integrations`.
When these are enabled, the agent will automatically be given access to these tools, depending on the context.
### LLaMA-Mesh
Use [LlamaMesh](https://github.com/nv-tlabs/LLaMA-Mesh) local mesh understanding and generation.
Only select this option if you:
1. Have a powerful NVIDIA GPU with at least 8GB of VRAM
2. Installed a `cuda` version of the addon during [Installation](#installation)
3. Are using a remote API backend (e.g. [Hugging Face](#hugging-face), [Anthropic](#anthropic), or [OpenAI](#openai)), as LLaMA-Mesh will load locally on your machine
To enable LLaMA-Mesh, click `Load LLama-Mesh` and wait for the model to load.
### Hyper3D
Use [Hyper3D](https://hyper3d.ai/) for high-fidelity 3D mesh generation.
To enable:
1. Check `Enable Hyper3D`
2. Enter your Hyper3D API key in the `API Key` field (free use is currently provided with the `awesomemcp` key)
This may take several minutes per mesh.
## Usage
- Press `N` -> `MeshGen` (or `View` -> `Sidebar` -> Select the `MeshGen` tab)
- Enter a prompt, for example: `Create a snowman`
- Click `Submit`
## Troubleshooting
- Find errors in the console:
- Windows: In Blender, go to `Window` -> `Toggle System Console`
- Mac/Linux: Launch Blender from the terminal
- Report errors in [Issues](https://github.com/huggingface/meshgen/issues)