metadata

title: My Face Enhancement Space
emoji: 😊
colorFrom: blue
colorTo: green
sdk: gradio
sdk_version: 5.25.2
app_file: app.py
pinned: false

Face Enhance

A tool for improving facial consistency and quality in AI-generated images. Dramatically enhance facial fidelity while preserving the original image's background, lighting, and composition.

Installation

Prerequisites

Python 3.11 or higher
1 GPU with 48GB VRAM
At least 50GB of free disk space

Setup

Set up your Hugging Face token:
- Create a token at Hugging Face
- Log into Hugging Face and accept their terms of service to download Flux
- Set the following environment variables:
```
export HUGGINGFACE_TOKEN=your_token_here
export HF_HOME=/path/to/your/huggingface_cache
```
- Models will be downloaded to $HF_HOME and symlinked to ./ComfyUI/models/

Create the virtual environment:

python -m venv venv
source venv/bin/activate
python -m pip install -r requirements.txt

Run the install script:
```
python install.py
```
This will
- Install ComfyUI, custom nodes, and required dependencies to your venv
- Download all required models (Flux.1-dev, ControlNet, text encoders, PuLID, and more)

Run inference on one example:

python main.py --input examples/dany_gpt_1.png --ref examples/dany_face.jpg --out examples/dany_enhanced.png

Running on ComfyUI

Using the ComfyUI workflows is the fastest way to get started. Run python run_comfy.py

./workflows/FaceEnhancementProd.json for face enhancement
./workflows/FaceEmbedDist.json for computing the face embedding distance

Gradio Demo

A simple web interface for the face enhancement workflow.

Run python gradio_demo.py
Go to http://localhost:7860. You may need to enable port forwarding.

Notes

The script and demo run a ComfyUI server ephemerally
Gradio demo is faster than the script because models remain loaded in memory
All images are saved in ./ComfyUI/input/scratch/
Temporary files are created during processing and cleaned up afterward

Troubleshooting

Out of memory errors: If your GPU has less than 48 GB VRAM, install Flux.1-dev at float8 precision.
Face detection issues: This method works for photorealistic images of people. It may not work on cartoons, anime characters, or non-human subjects.
Downloading models fails: Check your Hugging Face token has proper permissions.