KI-Foto-Restaurierung

How It Works

What This Tool Does

Our AI Old Photo Restoration tool combines two powerful neural networks in a single pipeline:

• AI Colorization — The Siggraph 2017 deep learning model by Richard Zhang et al. analyzes the grayscale content of your photo and predicts natural, historically plausible colors. It works in Lab color space, preserving the original luminance while adding color channels.
• AI Enhancement — Real-ESRGAN (Enhanced Super-Resolution GAN) upscales the image by 2× while simultaneously removing noise, compression artifacts, and blurriness. The result is a crisp, detailed photo at higher resolution than the original.

You can disable colorization if your photo is already in color and you only want enhancement and denoising.

Best Practices

• Scan at high DPI — For printed photos, scan at 300+ DPI before uploading. Higher input quality gives better results.
• Crop before uploading — Remove borders, tape marks, or album edges. The AI works best when the photo fills the frame.
• B&W vs Color — If your photo is already in color but faded, select “No” for colorization. The enhancement step alone will sharpen and denoise the image.
• Very damaged photos — For photos with tears, stains, or missing areas, use our AI Object Removal tool first to clean up damage, then restore.