Frequently asked questions
Is it really free and unlimited?
Yes. Inference runs on your own hardware, so each image costs us nothing. Ads are the only revenue - no credits, no paywall, no watermark.
What are the current limits?
Beta limit is 4 megapixels per image (about 2600x1500) and PNG output. Larger inputs and batch mode are planned. Very large upscales need more device memory.
Which AI model is used?
Real-ESRGAN general x4v3 (BSD-3-Clause), a compact 4x super-resolution model - the same family as popular desktop upscalers. The 4.6 MB file downloads once and is cached by your browser.
Are my images uploaded?
No. Processing is 100% local JavaScript/WebAssembly. You can verify in DevTools: after the model loads, no image data leaves the page.
Why does a big image take longer?
The image is enlarged in overlapping tiles; time grows with pixel count. A 400x300 photo finishes in seconds on most laptops; a 12-megapixel photo would take minutes (and is above the beta limit).
Does it work on phones?
Yes, on modern iOS and Android browsers. Keep images small on older phones - memory is the limiting factor.
Can it fix a blurry photo?
It recovers detail from small images but cannot invent information that was never captured. For heavily blurred sources, expect a cleaner enlargement, not a miracle.
Is commercial use allowed?
Yes - the model license (BSD-3-Clause) permits commercial use. The model is credited on the About page as its license requires.
Ready to try? Open the upscaler.