zsxkib / cog-aura-sr

AuraSR: GAN-based Super-Resolution for real-world

AuraSR: GAN Super-Resolution for Images 🖼️

AuraSR is a powerful tool that makes images bigger and clearer. It's based on the GigaGAN idea and works great for certain types of images.

Quick Start 🚀

Use Cog to make your image bigger:

cog predict -i image=@your_image.png -i scale_factor=4

What It Does 🎨

• Makes PNG, lossless WebP, and high-quality JPEG XL (90+) images bigger and clearer
• Can make images 2, 4, 8, 16, or 32 times bigger
• Works fast and can handle different sized jobs

Things to Know ⚠️

AuraSR is great, but it has some limits:

1. It works best with PNG, lossless WebP, and high-quality JPEG XL (90+) images.
2. It doesn't like images that have been squeezed too much (compressed).
3. It can't fix mistakes in images.
4. It's best for making AI-generated images or very high-quality photos bigger.

What You Need 📋

• Python 3.7 or newer
• Cog

How to Use It 🛠️

Simple Way

cog predict -i image=@your_image.png -i scale_factor=4

Advanced Way

cog predict -i image=@your_image.png -i scale_factor=8 -i max_batch_size=4

Options 🔧

• image: The picture you want to make bigger
• scale_factor: How much bigger you want to make it (2, 4, 8, 16, or 32)
• max_batch_size: Controls how fast it works (default is 1, increase if you have a powerful computer)

Thank You 🙌

• GAN-based Super-Resolution for real-world images, a variation of the GigaGAN paper for image-conditioned upscaling. PyTorch implementation is based on the unofficial lucidrains/gigagan-pytorch repository.

Let's Talk 🐦

Have questions? Follow me on Twitter @zsakib_ and let's chat!