CozyGen: A Mobile-Friendly ComfyUI Controller

DISCLAIMER

This project was 100% "vibe-coded" using Gemini 2.5 Pro/Flash. I dont code, but wanted to share a working LLM assisted projected. Everything AFTER this disclaimer section is 99% made by an LLM. I just wanted to make dumb cat pictures with my desktop ComfyUI from my phone, so now this exists. Thanks to @acly and the comfyui-tooling-nodes for the inspiration.

Known Issues:

• Not all dropdowns populate in the webpage.
• Connecting a node to multiple fields does not work (for example: an int type for "Height" going to two height nodes does not work).
• If you use a batch higher than 1 they will generate and show in the gallery, but the preview only displays the first image of the batch.
• Currently only t2i workflows have been tested. I plan to add i2i, t2v, i2v, and other types, but no promises (this is a hobby project).

Changelog:

• 8/29/2025 - Initial release

✨ Overview

CozyGen is a custom node for ComfyUI that provides a sleek, mobile-friendly web interface to remotely control your ComfyUI server. Designed for ease of use, it allows you to load pre-defined workflows, dynamically adjust parameters, and generate stunning images from any device with a web browser. Say goodbye to the desktop interface and hello to on-the-go creativity!

🚀 Features

• Modern & Intuitive UI: A beautiful, mobile-first interface built with React, Vite, and Tailwind CSS, featuring a stylish dark theme.
• Dynamic Controls: The user interface automatically generates input controls (text fields, sliders, dropdowns, toggles) based on the CozyGenDynamicInput nodes in your ComfyUI workflows.
• Priority Sorting: A "priority" field that determines how the webpage is ordered. A 0 priority will push the field towards the top of the page.
• Real-time Previews: Get instant visual feedback with real-time previews of your generated images directly in the web interface.
• Persistent Sessions: Your selected workflow, input values, and even the last generated image are remembered across browser sessions.
• Image Gallery: Browse, view, and manage all your previously generated images, complete with extracted prompt and seed metadata.
• Randomization: Easily randomize numerical inputs like seeds with a dedicated toggle.
• Seamless Integration: Works directly with your existing ComfyUI setup, leveraging its core functionalities.

📸 Screenshots / Demos

Mobile-first design:

<p align="center"> <img width="744" height="1267" alt="Image" src="https://github.com/user-attachments/assets/6f121edd-5cc0-4b25-b173-2cbc97761bec" /> </p>

Adapts to browser size:

<p align="center"> <img width="1514" height="865" alt="Image" src="https://github.com/user-attachments/assets/c656b73b-8773-4e5f-af9a-83249701f384" /> </p>

Custom Node adapts to the string/int/float/dropdown they are connected to:

<p align="center"> <img width="1745" height="920" alt="Image" src="https://github.com/user-attachments/assets/52d00ee5-42ef-4a5e-a39e-52b6ff80f852" /> </p>

A gallery tab that can navigate your ComfyUI output folder. Click on the path in the top left to go back to the base output folder.

<p align="center"> <img width="1532" height="692" alt="Image" src="https://github.com/user-attachments/assets/1951a027-bf49-48f2-b1d5-e9e85c3351a8" /> </p>

📦 Installation

Node can be installed with the ComfyUI Manager. Search for "CozyGen" to install.

Follow these steps to get CozyGen up and running with your ComfyUI instance.

1. Clone the Repository

Navigate to your ComfyUI custom_nodes directory and clone this repository:

cd /path/to/your/ComfyUI/custom_nodes
git clone https://github.com/gsusgg/ComfyUI_CozyGen.git

2. Install Python Dependencies

(This node only requires aiohttp, which should already be installed with ComfyUI.)

CozyGen requires the aiohttp Python package. Navigate into the ComfyUI_CozyGen directory and install them using pip.

cd custom_nodes/ComfyUI_CozyGen
pip install -r requirements.txt

3. Restart ComfyUI

After completing the above steps, restart your ComfyUI server to load the new custom node and its web interface.

4. (Optional) ComfyUI --listen

If you want to use this as a remote to your machine running ComfyUI on the local network, add the "--listen" flag to your ComfyUI startup.

🚀 Usage

1. Prepare Your Workflow

In ComfyUI, create or open a workflow that you want to control remotely. For each parameter you wish to expose to the web UI:

• Add a CozyGenDynamicInput node and connect its output to the desired input on another node.
• Configure the CozyGenDynamicInput node's properties (e.g., param_name, param_type, default_value, min_value, max_value, add_randomize_toggle).
• Add a CozyGenOutput node at the end of your workflow to save the generated image and send real-time previews to the web UI.
• IMPORTANT When exporting your workflow, export with API into the ComfyUI_CozyGen/workflows/ directory.

2. Access the Web UI

Open your web browser and navigate to:

http://<your-comfyui-ip>:8188/cozygen

(Replace <your-comfyui-ip> with the IP address or hostname where your ComfyUI server is running, e.g., 127.0.0.1 for local access).

3. Generate Images

1. Select your prepared workflow from the dropdown menu.
2. Adjust the dynamically generated parameters as needed. Your settings will be saved automatically.
3. Click the "Generate" button.
4. The generated image will appear in the preview area. You can click it to expand it or use the "Clear" button to reset the panel.
5. Click the "Gallery" link in the header to browse all your generated images.

🤝 Contributing

I do not plan to update this further, but wanted to share what I had. Feel free to take it and update it on your own!

📄 License

This project is licensed under the GPL-3.0 license - see the LICENSE file for details.