Simplify your AI media generation workflows with Visionatrix—an intuitive interface built on top of ComfyUI

• 🔧 Easy Setup & Updates: Quick setup with simple installation and seamless version updates.
• 🖥️ Minimalistic UI: Clean, user-friendly interface designed for daily workflow usage.
• 🌐 Prompt Translation Support: Automatically translate prompts for media generation.
• 🛠️ Stable Workflows: Versioned and upgradable workflows.
• 📈 Scalability: Run multiple instances with simultaneous task workers for increased productivity.
• 👥 Multi-User Support: Configure for multiple users with ease and integrate different user backends.
• 🤖 LLM Integration: Effortlessly incorporate Ollama/Gemini as your LLM for ComfyUI workflows.
• 🔌 Seamless Integration: Run as a service with backend endpoints for smooth project integration.
• 😎 LoRA Integration: Easy integrate LoRAs from CivitAI into your flows.
• 🐳 Docker Compose: Official Docker images and a pre-configured Docker Compose file.

Access the Visionatrix UI at http://localhost:8288 (default).

Note: Starting from version 1.10 Visionatrix launches ComfyUI webserver at http://127.0.0.1:8188

We provide public template for RunPOD to help you quickly see if this project fits your needs.

• RunPOD template

• Python 3.10 or higher. (3.12 recommended)
• GPU with at least 8 GB of memory (12GB recommended)

sudo apt install wget curl python3-venv python3-pip build-essential git

sudo apt install ffmpeg libsm6 libxext6

Download and run the easy_install.py script:

Note: This script will clone the Visionatrix repository into your current folder and perform the installation. After installation, you can always run easy_install from the "scripts" folder.

Using wget:

wget -O easy_install.py https://raw.githubusercontent.com/Visionatrix/Visionatrix/main/scripts/easy_install.py && python3 easy_install.py

Using curl:

curl -o easy_install.py https://raw.githubusercontent.com/Visionatrix/Visionatrix/main/scripts/easy_install.py && python3 easy_install.py

Follow the prompts during installation. In most cases, everything should work smoothly.

To launch Visionatrix from the activated virtual environment:

python -m visionatrix run --ui

We offer a portable version to simplify installation (no need for Git or Visual Studio compilers).

Currently, we provide versions for CUDA/CPU. If there's demand, we can add a DirectML version.

1. Install VC++ Redistributable: vc_redist.x64.exe from this Microsoft page.
2. Download: Visit our Releases page.
3. Get the Portable Archive: Download vix_portable_cuda.7z.
4. Unpack and Run: Extract the archive and run run_nvidia_gpu.bat or run_cpu.bat.

For manual installation steps, please refer to our detailed documentation.

The easiest way to set up paths is through the user interface, by going to Settings->ComfyUI.

In most cases, the easiest way is to set ComfyUI base data folder to some absolute path where you want to store models, task results, and settings.

This will allow you to freely reinstall everything from scratch without losing data or models.

Note: For easy Windows portable upgrades, we assume you have ComfyUI base data folder parameter set.

We highly recommend filling in both the CivitAI token and the HuggingFace token in the settings.

Many models cannot be downloaded by public users without a token.

Run the easy_install script and select the "Update" option.

python3 easy_install.py

Updating the portable version involves:

1. Unpacking the new portable version.
2. Moving visionatrix.db from the old version to the new one.

`sqlite+aiosqlite:///C:/Users/alex/visionatrix.db`

We provide official Docker images along with a pre-configured docker-compose.yml file, making deployment faster and easier. The file is located at the root of the Visionatrix repository.

Our Docker images are primarily hosted on GitHub Container Registry (GHCR): ghcr.io/visionatrix/visionatrix. This is the default used by the docker-compose.yml file.

For users who experience slow download speeds from GHCR (e.g., on certain cloud providers), we also provide a mirror on Docker Hub: docker.io/bigcat88/visionatrix.

• visionatrix_nvidia: Visionatrix with NVIDIA GPU support.
• visionatrix_amd: Visionatrix with AMD GPU support.
• visionatrix_cpu: Visionatrix running on CPU only.
• pgsql: A PostgreSQL 17 container for the database.

Choose the service appropriate for your hardware:

• For NVIDIA GPU support:

  docker compose up -d visionatrix_nvidia

• For AMD GPU support:

  docker compose up -d visionatrix_amd

• For CPU mode:

  docker compose up -d visionatrix_cpu

By default, these commands will pull images from GHCR. A visionatrix-data directory will be created in the current directory in the host and used for the models, user, input and output files.

You can easily customize the configuration by modifying environment variables or volume mounts in the docker-compose.yml file.

If you prefer to pull images from Docker Hub instead of GHCR, you can set the VIX_IMAGE_BASE environment variable before running docker compose up.

Method 1: Using a .env file

1. Create a file named .env in the same directory as your docker-compose.yml file.

2. Add the following line to the .env file:

   VIX_IMAGE_BASE=docker.io/bigcat88/visionatrix

3. Now, run docker compose up as usual. Compose will automatically read the .env file and use the Docker Hub images.

   # Example: Start NVIDIA service using images from Docker Hub defined in .env
   docker compose up -d visionatrix_nvidia

Method 2: Setting the variable temporarily

You can set the environment variable directly on the command line for a single command execution:

VIX_IMAGE_BASE=docker.io/bigcat88/visionatrix docker compose up -d visionatrix_nvidia

If you have any questions or need assistance, we're here to help! Feel free to start a discussion or explore our resources:

• Documentation
  • Available Flows
  • Admin Manual
    • Installation
    • Working Modes
    • Command Line Options
    • Environment Variables
  • Flows Developing
    • Vix Workflows
    • Technical Information
    • Creating Workflows
    • Models Catalog
    • Gated Models
  • Common Information
    • FAQ
    • Hardware FAQ
    • Hardware Results
    • OpenAPI