Improve PDF Quality
Upscale the pictures inside a PDF with AI on your device. Text, links and layout stay exactly as they are.
Improve the quality of a PDF for free by upscaling the pictures inside it with an AI model (Real-ESRGAN) that runs in your browser. Each low-resolution picture is replaced in place with a sharper one, two or four times its size, while the text, links and layout are left untouched. Nothing is uploaded. The file gets bigger, and very small text inside a picture can be reshaped, so the page shows a preview first.
Drop a PDF here, or browse
A PDF with pictures in it: a scan, a guide, a catalogue. Nothing is uploaded.
①Select file→②Scan→③Download
Private
Stays on Device
Instant
How it works
- Choose a PDF. It is opened on your device and never uploaded.
- See which pictures will be upscaled, and why the others are left alone.
- Compare one picture before and after, and pick 2x or 4x.
- Upscale and download. The text, links and layout stay as they were.
When it helps
- Game guides, manuals and hobby books whose pictures look blocky when you zoom in
- Documents scanned at low resolution, where each page is stored as one picture
- Catalogues and brochures with small product photos
- Getting a PDF ready for printing when its pictures are too coarse
- Slides and reports with pixelated screenshots
Key Benefits:
- ✓Text and Links Kept - Only the pictures are replaced, in place. The text stays selectable and every link still works
- ✓AI on Your Device - The model runs in your browser, and the PDF is never uploaded
- ✓No Credits, No Watermark - No sign-up, no daily limit and nothing added to your pages
- ✓Honest About the Result - A preview first, a time estimate for your device, and the new file size shown before you download
Privacy First:
The PDF, its pictures and the AI model all stay in your browser. The model is downloaded once from this site, and your document is never sent anywhere.
Frequently Asked Questions
Will the text in my PDF stay selectable and searchable?
Yes. Only the picture objects are replaced; the text, fonts, links and layout are not touched. Before the download is offered, the new file is checked against the original: the same pages and the same text.
Can it fix a blurry scanned PDF?
It can make a low-resolution scan sharper, because each scanned page is stored as a picture. It cannot bring back detail from a page that was badly out of focus, and it does not make the scan searchable. For that, run OCR on the result.
Can the AI change letters, numbers or prices?
It can, in very small text inside a picture. In our tests, letters around 10 pixels tall were sometimes reshaped, while text of normal size stayed correct. Check codes, prices and small numbers in the preview and in the result before relying on them.
Is my PDF uploaded anywhere?
No. The PDF is opened in your browser and the AI model runs there too. The model files, about 10 MB, are downloaded once from this site. Your document never leaves your device.
Why is my PDF bigger afterwards?
A picture upscaled 2x has four times as many pixels, and 4x has sixteen times as many. The new size is shown before you download. If it is too big, choose 2x, or run the result through Compress.
How long does it take?
It depends on your device. With a graphics card, a picture usually takes about a second. On the processor alone, a full scanned page can take a minute or more. After the preview, the page estimates the time for your whole PDF on your device.
Does it make the pages bigger, for example A4 to A3?
No. The pages keep their size. Only the resolution of the pictures on them goes up, so they stay sharp when you zoom in or print.
Why were some pictures left as they are?
Pictures already sharp enough to print, tiny icons, black-and-white scans and some kinds of transparency are skipped, each with its reason shown, because upscaling them would only make the file bigger.
Is it free? Are there limits?
It is free, with no sign-up, no credits and no watermark. The only limit is your device's memory: a picture that would grow past about 16 megapixels is upscaled less, or skipped.
Which AI model does it use?
Real-ESRGAN, the general x4v3 model by Xintao Wang, under the BSD licence. It runs with ONNX Runtime Web on your graphics card where the browser allows it, and on the processor otherwise.
💡
If the file got too big, compress it. To make a scanned page searchable, run OCR on the result.