This application uses GFPGAN (Generative Facial Prior-GAN) to restore old, damaged, or low-quality photos, particularly focusing on facial restoration and enhancement.
| Before | After |
|---|---|
 |
 |
| Key Improvements: |
- Enhanced facial features clarity
- Improved skin texture
- Better detail in eyes and facial features
- Preserved original expression and identity
| Before | After |
|---|---|
 |
 |
| Key Improvements: |
- Removed scratches and damage
- Enhanced color balance
- Improved overall sharpness
- Restored lost facial details
| Before | After |
|---|---|
 |
 |
| Key Improvements: |
- Increased resolution
- Enhanced facial details
- Improved image clarity
- Better color reproduction
-
Face Restoration
- Enhances facial features
- Restores missing details
- Maintains natural look
- Preserves identity
-
Image Enhancement
- Improves resolution
- Removes noise and artifacts
- Enhances color balance
- Sharpens details
-
Damage Repair
- Removes scratches
- Fixes color fading
- Repairs tears and marks
- Restores damaged areas
The application uses state-of-the-art AI models:
- GFPGAN: For facial restoration
- Real-ESRGAN: For background enhancement
- Processing resolution: Up to 2K
- Average processing time: 2-5 seconds per image
For best results:
- Input images should be clear enough to identify faces
- Works best with front-facing portraits
- Can handle multiple faces in one image
- Supports common image formats (JPG, PNG)
- Very severe damage might not be fully restored
- Extreme angles might affect face restoration quality
- Processing time increases with image size
- Results may vary based on input image quality
git clone https://github.com/TencentARC/GFPGAN.git
cd GFPGAN
pip install -r requirements.txt
python setup.py develop
uvicorn main:app --host 0.0.0.0 --port 8000