Traditional upscaling just stretches pixels — you get blur. AI super-resolution *invents* plausible detail.
How AI upscaling works
The model has seen millions of (low-res, high-res) image pairs. When it sees a small image, it predicts what the high-resolution version should look like — edge by edge, texture by texture.
What it's good at
- **Faces** — portrait upscaling (great for old photos)
- **Product photos** — e-commerce, print
- **Artwork** — clean lines, sharp colors
- **Landscapes** — natural textures
What it's bad at
- Heavily compressed JPEGs (artifacts get amplified)
- Text-heavy images (letters get weird)
- Pixel art (defeats the purpose)
Scale options
- 2× — quick, safe
- 4× — best balance
- 8× — max, slower
Output
PNG or JPG at your chosen scale. Original + upscaled side-by-side for comparison.