How to scan a document
- Open ScanBench on your phone and press Take photos. Allow the camera: the back camera opens at the highest resolution it offers.
- Hold the page in the frame. A yellow outline shows where ScanBench sees the page; press the round button for each page. The counter in the corner shows how many pages you have, and a torch button appears if the phone has a torch.
- Press Done. Every page is already cropped and straightened; tap one to fix its corners, filter or rotation.
- Choose the format in Save and press Save PDF, or share the file straight to a messenger or mail.
On a computer, or for photos you already have, press Add photos or PDF, drag files onto the page or paste a picture with Ctrl+V. Each page of a PDF becomes a page of the scan. If the camera cannot start (no permission, an old browser or a page opened without HTTPS), ScanBench offers the phone’s own camera app instead; its photos work just as well.
How to take a good photo of a page
- Contrast with the background. A white sheet on a dark table is found instantly; white on white is hard for any scanner. Put a dark folder or a book under the page.
- Light without shadow. Daylight from a window is best. Do not stand between the lamp and the page: your own shadow is the most common reason for a grey scan. The Colour and B&W filters even out soft shadows, but not a sharp one across the text.
- From above, the whole page in the frame. A slight angle is fine because the perspective is corrected, but the closer to straight down, the sharper the far edge of the page.
- Hold still and focus. Tap the screen on the text if the phone focuses on the table, and wait until the picture is sharp.
- The torch helps in a dark room, but makes a bright spot on glossy paper: try with and without it.
Edges, corners and perspective
For every photo ScanBench looks for the outline of the page: it finds sharp changes of brightness (the Canny method), joins them into shapes and picks the largest four-cornered shape whose sides lie on real edges. Then it fits a straight line along each side, so the corners land precisely, and stretches the page into a flat rectangle, as if it had been on a flatbed scanner (a perspective transform, or homography).
If the page was not found, its thumbnail gets a yellow ! and the whole photo is kept. Tap the page and drag the four yellow corners to the corners of the sheet: a magnifier above your finger shows the exact spot. Find edges searches again, Whole photo takes the frame as it is. With the keyboard, arrows move a selected corner by one pixel (with Shift, by ten).
The proportions of the result come from the longest sides of the outline. A page shot at a steep angle may come out a little shorter than it really is; on an A4 or Letter page the picture is fitted without stretching.
Filters, turning and trimming
| Filter | What it does | Good for |
|---|---|---|
| Original | the photo as it is, only straightened | photos, drawings, colours you want to keep exactly |
| Colour | evens out shadows and the colour of the light, makes the paper white and colours a little brighter | forms with stamps, colour documents (the default) |
| Grey | the same cleaning, without colour | text with photos, smaller files |
| B&W | black text on white, like a scanner or a fax (adaptive threshold) | text documents, contracts, receipts |
Shadows are removed by estimating the colour of the paper all over the page and dividing every pixel by it, so a shadowed corner becomes as white as the middle. The black-and-white filter compares every pixel with its own neighbourhood rather than with one level for the whole page, so it copes with uneven light too.
The filter buttons above the pages change all of them at once; inside a page, Use these settings for all pages copies its filter and trimming to the others. The turn buttons rotate by 90°, and Auto-rotate looks at the direction of the text lines and their straight left margin; pages that are clearly sideways or upside down are turned automatically when you add them. Trim edges cuts up to 10% off every side, for a dark border or a bit of table left in the frame.
Saving: PDF, JPG, PNG
- PDF: all pages in one file. Page size A4 or Letter (turned to match each page), or As photo, where the page has the proportions of the scan.
- JPG or PNG: one picture per page. Several pictures are shared together where the phone allows it, or come in one ZIP.
- Quality: Smaller (about 1600 px on the long side, for mail and messengers), Balanced (about 2500 px, roughly 300 dpi for A4, sharp to print) or Best (the full resolution of your photos, up to 16 megapixels, with the least compression).
Previews on the screen are made small to save memory; the saved file is always made again from the original photo in the chosen quality. The share button sends the file through the phone’s share menu (messengers, mail, cloud drives) where the browser can share files. Location and camera data of your photos never get into the result.
Text recognition (OCR)
Recognise text reads the pages with Tesseract.js, the browser version of the open-source Tesseract engine. Choose Ukrainian, English or both, for documents that mix the two. The first time, the language files are downloaded from fof.lol, with a progress bar: about 1.6 MB for Ukrainian and 2 MB for English. Your browser keeps them, so recognition works offline afterwards; their size and a Delete button are shown under the card.
The text appears in a field you can copy or save as .txt. With Add the recognised text to the PDF, the PDF gets an invisible text layer exactly under the words of the picture: you can search the document, select and copy text in any PDF viewer, and it still looks like a scan. A page you change after recognition (corners, filter, rotation) is read again next time.
Recognition is best on straight, sharp pages of printed text. Handwriting, very small print, decorative fonts and blurry photos give mistakes, so check important numbers and names.
Keeping a document for later
By default the pages live only in the open tab and disappear when you close it; the browser even asks before you leave. Turn on Save this document on this device to keep the photos, settings and recognised text in this browser (IndexedDB): the next time you open ScanBench, the document is there. Delete saved copy removes it. Nothing is kept without your choice.
Privacy
Photos, pages and recognised text are processed by your browser and never uploaded to fof.lol or anywhere else, so ScanBench suits passports, contracts and medical papers. Only code and data are downloaded, and only when you use the feature that needs them: PDF.js to open PDF files and pdf-lib to write the PDF (from cdnjs), Tesseract.js (from jsDelivr) and the language files (from fof.lol). Page detection, straightening and filters are our own code and need no download at all. The camera picture is analysed on the device, and the camera stops as soon as you close it or switch to another tab.
Limitations
- The camera in the browser needs HTTPS and your permission. Some browsers inside apps (messengers, social networks) do not give the camera to sites: open the page in Chrome or Safari, or use the system camera button.
- Resolution. Where the browser can take photos (Chrome on Android), ScanBench uses the full photo resolution; elsewhere it takes a frame from the video, which is often smaller. The camera screen shows how many megapixels you get. For the sharpest result, take photos with the camera app and add them.
- Edges are hard to find for a white page on a white table, a crumpled or curled page, or when a hand covers a corner: set the corners by hand. Curled pages of thick books are not flattened.
- Memory. The original photos stay in memory while you work; 20–30 pages from a 12 MP camera are fine on a modern phone. On an old phone, save in parts or choose Smaller.
- OCR knows only Ukrainian and English and does not read handwriting. A page takes a few seconds on a computer and longer on a phone.
- HEIC photos open only in browsers that can show them (Safari); iPhones normally convert them when you pick a photo.
Frequently asked questions
Do I need to install an app?
No. ScanBench is a web page. You can add it to the home screen, where it opens like an app and works without the internet, but you do not have to.
Is it safe to scan a passport or a contract here?
Yes. The photos are not sent anywhere: there is no server that could receive them. You can check it in the browser’s developer tools: no request carries your pictures.
Why is my scan grey or yellowish?
The Original filter is on. Choose Colour or B&W: they make the paper white and remove soft shadows.
Can I scan a multi-page document into one PDF?
Yes. Photograph all the pages in a row, put them in order by dragging the thumbnails and save one PDF.
Can I search the text in the PDF?
Yes, if you recognise the text first and leave Add the recognised text to the PDF on. Then Ctrl+F finds words, and you can select and copy them.
Does it work without the internet?
Scanning, filters and saving pictures work offline once the page has been opened. PDF and OCR need their libraries and language files once; after that they work offline too.