VisionFlow is a Streamlit-powered web application that allows you to generate images easily using the power of ComfyUI workflows. Simplify AI image creation with an intuitive and accessible interface.
- Intuitive web interface built with Streamlit
- Integration with ComfyUI workflows
- AI-powered image generation
- Automatic ComfyUI startup (if not already running)
- Flexible parameter configuration
- Real-time result visualization
- Python 3.12 or higher
- ComfyUI installed and configured
If GIT is installed, open a terminal where you want to install VisionFlow and type:
git clone https://github.com/Franck-Demongin/VisionFlow.git
cd VisionFlowIf GIT is not installed, download the ZIP file, unzip it to your desired directory and rename it to VisionFlow.
Open a terminal in the VisionFlow folder.
Create a virtual environment to isolate dependencies:
python -m venv .venvpython should be replaced by the appropriate command according to your installation. On Linux, it could be python3.12, on Windows python.exe
Activate the virtual environment:
# Windows
.venv\Scripts\activate
# Linux/macOS
source .venv/bin/activateInstall dependencies:
pip install -r requirements.txtComfyUI will be automatically launched by VisionFlow if it's not already running. You can also start ComfyUI manually if you want to access logs in a separate terminal (logs are not accessible when ComfyUI is launched by VisionFlow).
Create a config.ini file (copy and rename config.ini.example) or modify the existing configuration file.
The configuration file uses INI format with a [comfyui] section:
[comfyui]
comfyui_path = /path/to/your/ComfyUI
python_path = /path/to/your/ComfyUI/venv/bin/python
url = localhost:8188
; uncoment the next ligne to add some parameters to start ComfyUI server
; params = --use-sage-attentionAvailable options include:
comfyui_path: path to your ComfyUI installation directorypython_path: path to Python executable in ComfyUI's virtual environmenturl: ComfyUI server address and port. Default: localhost:8188params: additional parameters for ComfyUI startup
To add uncommented parameters, separate them with a space. eg. params = --use-sage-attention --auto-launch
To start VisionFlow, run:
streamlit run app.pyThe application will be accessible at http://localhost:8501
- Workflow Selection: Choose your desired ComfyUI workflow
- Parameter Configuration:
- Seed: seed for random generation
- Size: output image dimensions
- Batch size: number of images to generate
- Prompt: Enter your text description
- Generation: Click the button to start generation
- Visualization: Generated images display directly in the interface
- Download: Ability to download images individually
- History: Access to previous generations
- Parameters: View parameters used for each generation
VisionFlow currently supports the following ComfyUI workflows:
- Text-to-Image: Generate images from text descriptions
Coming soon:
- Inpainting and outpainting
- Image-to-image transformation
VisionFlow allows you to integrate your own ComfyUI workflows. Here's how to add a new workflow:
Prerequisites
- Your workflow must be fully functional in ComfyUI
- All required nodes and models must be already installed
- Test your workflow in ComfyUI before integration
- Export your workflow: In ComfyUI, go to Workflow > Export (API) to export your workflow as JSON
- Copy workflow file: Place the exported JSON file in the workflows/ directory of your VisionFlow installation
- Create configuration file: Create a configuration file with the same name but with _config suffix
Example: workflowSDXL.json → workflowSDXL_config.json
The configuration file must contain general information about the workflow and parameter mappings:
{
"name": "Display Name for UI",
"description": "Brief description of what this workflow does",
"global": {
"seed": {
"type": "int",
"node": "25",
"input": "noise_seed",
"default": 42
},
"width": {
"type": "int",
"node": "144",
"input": "value",
"default": 1024
},
"height": {
"type": "int",
"node": "142",
"input": "value",
"default": 1024
},
"batch_size": {
"type": "int",
"node": "27",
"input": "batch_size",
"default": 1
},
"clip_l": {
"type": "string",
"node": "63",
"input": "clip_l",
"default": "{prompt}"
},
"t5xxl": {
"type": "string",
"node": "63",
"input": "t5xxl",
"default": "{prompt}"
}
}
}- name: Display name shown in VisionFlow UI
- description: Brief description of the workflow functionality
- global: Dictionary containing all workflow parameters
All the parameters (name, description and global) are required.
The global section must include these standard parameters:
- seed: Random seed for generation
- width: Image width
- height: Image height
- batch_size: Number of images to generate
Each parameter contains:
- type: Variable type (int, float, string, etc.)
- node: Node ID as string (identifies the corresponding node in ComfyUI workflow)
- input: Field name in the ComfyUI workflow
- default: Default value for the parameter
For fields that should receive the user's prompt:
- The parameter name can be anything descriptive (eg. prompt, clip_l, t5xxl)
- Set default to "{prompt}"
Multiple fields can receive the same prompt by using "{prompt}" in their default values
This allows VisionFlow to automatically inject the user's text prompt into the appropriate nodes of your workflow.
Check that:
- ComfyUI is started and accessible
- Connection parameters are correct in configuration
- No firewall is blocking the connection
- Reduce image size or batch size
- Check available system resources
- Ensure your GPU is properly configured
Contributions are welcome! Feel free to:
- Report bugs
- Suggest new features
- Submit pull requests
This project is licensed under the MIT License. See the LICENSE file for details.
Added:
- Intuitive Streamlit interface
- ComfyUI integration
- AI image generation
- Flexible configuration
- Custom workflow management
For help or to report issues, please open an issue on the GitHub repository.
Built with ❤️ using Streamlit and ComfyUI