ComfyUI & AUTOMATIC1111 on Modal

This repository sets up a working environment for ComfyUI, AUTOMATIC1111 Stable Diffusion WebUI, and Deforum on modal.com.

It includes pre-built images that handle downloading and saving various models — including checkpoints, LoRAs, VAEs, and ControlNets — from Huggingface and CivitAI.

🚀 Setup

1. Create an account on modal.com.
2. Install dependencies:

   pip install -r requirements.txt

3. Set up Modal CLI:

   modal setup

4. Download models and assets:

   python download_data.py

5. Launch the WebUI:

   python comfyui.py   # For ComfyUI  
   python a1111.py     # For AUTOMATIC1111  

⚙️ Configuration

The config.json file defines which models to download.

To download models from CivitAI, include your CivitAI token in this file — it's required for access to CivitAI models.

If you want to add other models just add them to the config file and launch python download_data.py.

💾 Modal Volumes

The scripts use two persistent volumes:

• data: Stores models and user inputs.
• cache: Stores outputs and temporary files.

🛠️ Scripts

• download_data.py: Downloads models from HuggingFace and CivitAI based on config.json. Data is stored on a Modal volume to avoid re-downloading.
• a1111.py: Sets up the AUTOMATIC1111 WebUI and installs the Deforum extension.
• comfyui.py: Sets up ComfyUI with custom nodes and Deforum Comfy Nodes.

🔧 Useful Commands

Here’s a quick reference for managing data on Modal volumes:

📥 Download from Modal Volume:

modal volume get [OPTIONS] VOLUME_NAME REMOTE_PATH [LOCAL_DESTINATION]

Example:
To download outputs to a local output/ folder:

modal volume get --force cache / output/

📤 Upload to Modal Volume:

modal volume put [OPTIONS] VOLUME_NAME LOCAL_PATH [REMOTE_PATH]

Example:
To upload local input/ data to /data/input:

modal volume put --force data input/ /input