How to use Upscale image
- 1
Pick an image (JPG, PNG, WebP, AVIF, GIF or BMP).
- 2
Choose 2× for images up to 1600 pixels on their longest side, or 4× for images up to 1000 pixels.
- 3
Press "Upscale". The image goes to our server, where the AI enlarges it piece by piece while the progress bar fills.
- 4
Compare the before and after detail, then download the upscaled image.
What this tool does
Upscale image enlarges a picture two or four times with an AI model, so a 600 × 400 photo becomes 1200 × 800 or 2400 × 1600. Unlike normal resizing, it sharpens edges and text instead of smearing them.
Features
- 2× or 4× AI upscaling with Real-ESRGAN, an open model (BSD 3-Clause).
- Before and after comparison of a zoomed-in detail.
- Works on any device: the upscaling runs on our server, so it doesn't drain a phone's battery.
- Keeps transparency in PNG and WebP images.
- Keeps the format: a JPG stays a JPG and a WebP stays a WebP. AVIF, GIF and BMP images are saved as PNG.
Limits
- 2× takes images up to 1600 pixels on their longest side, and 4× up to 1000 pixels.
- 20 images a day without an account and 60 with a free one, shared with Remove background.
- Text in a tiny, blurry picture may come out sharp but with slightly different letter shapes, so check names and numbers.
- Animated GIFs are saved as a still image, and EXIF data such as GPS location is not kept.
Where your image is processed
The upscaling needs far more computing than a phone should do, so it runs on our own server. Your image travels there over an encrypted connection and is never handed to a third party. We delete the upload once the bigger image is made, and the result when its 15-minute download window ends, which means within an hour at most. Press "Delete from our server now" to remove both straight away.
Frequently asked questions
How is this different from resizing?
Ordinary resizing stretches the pixels, so edges and text get soft and blurry. An AI upscaler has learned what sharp images look like and fills in plausible detail, so lines, edges and text stay crisp at the larger size.
Which AI model is used?
Real-ESRGAN, the compact general model (realesr-general-x4v3) by Xintao Wang and others, released under the BSD 3-Clause licence. It runs on our own server, not through an outside AI service.
How long does it take?
A few seconds for small pictures. The time grows with the number of pixels, so the largest images allowed take up to about a minute. The progress bar shows how far it is.
Can it fix any blurry photo?
It works best on small images that are sharp for their size, such as small product photos, icons, screenshots and old web pictures. It also smooths away some noise and compression blocks. It can't recover detail from a badly out-of-focus or motion-blurred photo, and faces may look smoother than in a real higher-resolution photo.
Where is my image processed?
On our server. The image travels over an encrypted connection and is deleted automatically, the upload as soon as it is upscaled and everything within 1 hour. You can also delete it yourself straight away.