Fix the blurry images inside a PDF while the text stays searchable. A PDF looks blurry when its pictures are printed larger than their pixels allow — this lists every image with the resolution it is actually drawn at, ticks the ones below 150 DPI for you, and upscales just those with a real model on your own GPU. The image objects are replaced; headings, body text, links and vector drawings are never touched.
How it works
- Drop the PDF. Every embedded image is listed with its page, its pixel size and its effective DPI, and the low-resolution ones are already ticked.
- Upscale. Compare any image against the original before committing, and untick anything that was fine to begin with.
- Download. Same pages, same searchable text, sharper pictures — with the before-and-after file size shown so the growth is never a surprise.
Questions
- Can I fix a blurry PDF without losing selectable text?
- Yes, and that is what this page does. The tool swaps the image OBJECTS inside the document and rewrites nothing else, so the text layer, the fonts, the links and any vector drawing survive exactly as they were — Ctrl+F still works on the result. The alternative, which most "upscale a PDF" tools do by default, is to render each page to a picture and enlarge that: sharper-looking, and the document stops being a document. If your PDF is a scan with no text layer to lose, that path is on the image upscaler, and it asks you before taking it.
- Why does my PDF look blurry in the first place?
- Almost always because an image inside it has fewer pixels than the space it is drawn into: a 600×400 screenshot placed across a whole A4 page is printed at about 70 DPI, and 300 DPI is what looks sharp. A draft-quality scan is the other common cause. The tool measures the effective DPI of every image, so you can see which of yours it is instead of guessing.
- Does my file get uploaded?
- No. The models and all processing run inside your browser on your own device. You can verify this yourself: open your browser’s DevTools, watch the Network tab, and process a file: there are no uploads. After the first visit most tools also work fully offline.
- How is this different from upscaling the whole PDF?
- Dropping a PDF on the image upscaler rasterises every page — the text becomes pixels, and the file stops being searchable or selectable. This tool swaps the image objects only, so headings, body text, links and vector drawings are untouched and the document still behaves like a document.
- Which images can it not handle?
- Stencil masks (1-bit shapes painted in a fill colour) are listed but not offered: there is no photograph there to improve. Anything the browser and pdf.js together cannot decode is listed with the reason and left alone rather than silently skipped.
- Will the file get bigger?
- Yes, and usually by a lot: four times the pixels is roughly four times the image data, and the upscaled images are stored as PNG. The tool reports the finished size before you download so it is never a surprise.
- Does it need a fast machine?
- The best tier is Real-ESRGAN running tile by tile on your GPU — tens of seconds per image is normal, and a big scan can take longer. The two faster tiers are instant and still sharper than a plain stretch. WebGPU is required for all three.