How to use it
- 1
Add a small image. Drop a JPG, PNG or WebP. It works best on images under about 1000 pixels wide.
- 2
Let the AI work. The model reconstructs detail at twice the width and height. Bigger images take longer.
- 3
Compare and download. Look at both versions side by side, then download the 2× PNG.
[ADD YOUR OWN TEST HERE: what you tried, screenshots of the result, what worked and what didn't, and the date.]
What an AI upscaler does differently
Normal resizing just stretches pixels, so edges go soft and blurry. An AI super-resolution model has learned what sharp edges and textures look like, so it fills in plausible detail as it enlarges. This tool uses Swin2SR, an open-source model, running in your browser.
Getting the best result
- Start from the best copy you have. Heavy JPEG blockiness gets enlarged too.
- Upscale before you crop or add text.
- Need 4×? Run the result through a second time (on a strong computer).
Limits
An upscaler can't recover detail that isn't there: a tiny face won't become a recognizable portrait. To keep your browser responsive, images larger than 1024 pixels on the long side are reduced first. Large images can take a minute or more on older devices.
Questions
Is it free with no sign-up?
Yes, free, unlimited and with no account.
Will it add a watermark?
No.
Why is it slow on my phone?
Upscaling is heavy work. Phones without WebGPU support run it more slowly. A laptop with Chrome or Edge is fastest.
Are my images uploaded?
No, the model runs on your device.