A scan to PDF workflow has four steps: turn the images into a PDF, extract the text so it becomes searchable, tidy the page order and rotation, then lock the file with a password. Every one of those steps can run in the browser, so a sensitive document never has to leave the machine it was scanned on.
The value of doing it in one place is that nothing is handed to a stranger in the middle. A scanned passport, a signed contract or a medical form is exactly the file that cloud converters keep for as long as their policy allows, and it is exactly the file you should not upload. This workflow uses four separate browser tools, and the file moves between them through your own downloads folder rather than through a server. Each step below states what it does, what to set, and where the tool stops working, so you can decide how far to take it.
Most people solve this with a scanner app that uploads to a cloud drive, or a converter site that returns a PDF after a round trip. Both work until the document is one you are not allowed to send anywhere: a signed agreement, a benefits letter, a page of medical results. The four steps below do the same job with the file never leaving your device.
Image to PDF
Step one of the workflow. Combine JPG, PNG, WebP, AVIF, GIF and BMP scans into one PDF, with page size, orientation and margins under your control.
Why keep the whole workflow on your machine
The reason to care is not abstract. A cloud converter has to receive your file, keep it for some period, and decide when to delete it. For a utility bill that is a shrug; for an ID, a contract or a medical record it is a copy you never meant to create.
| This workflow | Upload-based converter | |
|---|---|---|
| Where the scan goes | Stays on your device | Sent to their servers |
| How the file moves between steps | Your downloads folder | Their session storage |
| Retention | You decide | Their policy |
| Account | Not needed | Often required |
| Works offline | After the first engine load | No |
Everything below assumes a current Chrome, Edge, Firefox or Safari, since the workflow needs the modern Canvas, WebAssembly and File APIs. The engines download once and are cached, so a second run of a step is faster.
Step 1: Turn the scans into a PDF
The image to PDF tool takes JPG, PNG, WebP, AVIF, GIF and BMP files and writes one PDF with a page per image. JPG and PNG are embedded as they are; the other formats are decoded through a canvas first, which keeps them working without a server.
- Add the page images. Drop the scans in the order you have them. Drag the cards, or use the up and down arrows on each card, to fix the order before exporting.
- Set the page geometry. Choose A4 or Letter, then pick Auto, Portrait or Landscape, and a margin of None, Small, Medium or Large. Choose Contain to fit the whole scan inside the margin, or Cover to fill the page and let the reader crop the overflow.
- Name the file and build it. The tool reports the combined source size and the finished PDF size, so you can see whether embedding the scans produced something reasonable before you download.
Settings that matter for a scan
Use Contain for documents, because it guarantees the whole page is visible and never clips a signature or a stamp. Cover is for photo-like scans where the edges do not carry information. A margin of Medium or Large gives a page a printable border and looks deliberate; None suits a scan you intend to view on screen only. If a page arrives sideways, leave it and fix the orientation in step three, where rotation does not re-encode anything.
Step 2: Make the text searchable
The PDF to text tool uses a two-stage strategy that decides itself. It first asks each page for its text layer, which is instant and exact for anything exported by a word processor, a browser or a design tool. Only for pages that have almost no real text, meaning they are really just a picture of a page, does it render the page at double resolution and run OCR on it.
What the tool actually gives you
Be clear about the output before you rely on it. The tool returns the words themselves, as text you can copy or download as a plain text file. It does not write an invisible OCR layer back into the PDF, so the PDF itself will not become searchable in a reader after this step. What you get is a searchable, quotable copy of the document's text sitting next to the PDF, which is what most archiving and quoting tasks actually need.
- Add the PDF. Drop the file you built in step one. If the PDF asks for a password, unlock it first with the unlock PDF tool, because this reader cannot open a locked file.
- Pick the OCR language if you need it. The choice only affects scanned pages; a page with a real text layer is read as it is, whatever language it is in. English runs from local data, other languages download once and are cached.
- Run it and read the summary. The result reports how many pages came from a text layer and how many needed recognition, so you know which part of the output deserves a proofread before you trust it.
Recognition is good on clean, straight scans at a sensible resolution and noticeably worse on skewed, faint or handwritten pages. Treat OCR output as a draft you check, not as a transcript you can file unread.
Step 3: Put the pages in the right order
The organize PDF tool renders a thumbnail for every page and lets you rebuild the document before exporting. It is where the fixes that step one cannot make, like a sideways page or a sheet scanned twice, get resolved without touching the original scan again.
- Reorder by dragging a thumbnail, or with the left and right arrows for keyboard and touch use.
- Rotate a single page or a whole selection by 90 degrees at a time. The rotation is stored as page geometry, so the image is not re-encoded.
- Delete blank or duplicate pages, with an undo button that steps back through your last changes if a page disappears by mistake.
- Add page numbers if you want them, choosing the starting number and whether they sit bottom right, top right or centred at the foot of the page.
Numbers are drawn in a standard Helvetica face, so they render the same everywhere. The tool writes a new file rather than editing in place, so your original PDF is untouched and a mistake here costs only a re-run.
Step 4: Lock it with a password
The protect PDF tool encrypts the finished document with AES-256 in the browser, using a WebAssembly build of qpdf. Encryption happens locally, so the password and the file never travel together over a network, which is the failure mode you are trying to avoid in the first place.
Open password versus owner password
There are two passwords, and they do different jobs. The user password is the one anyone must type to open the document at all. The owner password governs the permissions that travel with the file. If you leave the owner password blank, it defaults to the same value as the user password, which is fine for a personal archive and less flexible for a document you are distributing.
- Add the PDF and set the user password. Four characters is the minimum the tool accepts, and it checks that the confirmation field matches before it starts.
- Set the owner password and permissions. Decide separately whether printing is allowed, whether text can be copied out, and whether the document can be modified.
- Encrypt and download. The result is written with a protected suffix so it does not overwrite the original. Keep an unencrypted copy somewhere safe, because a forgotten password on an AES-256 file is not recoverable here.
Two limits apply. Files larger than 50 MB are refused, so split a very large scan first. A document that is already password protected must be unlocked before it can be encrypted again.
Two optional steps
Neither is required, and both are worth knowing about because their limits are easy to misunderstand.
Shrink a bloated scan first
The compress PDF tool is a lossless structural rewrite: it packs objects into streams, recompresses the streams it can, and rebuilds the cross reference table. What it does not do is re-encode the images inside a scan, so a file whose size is mostly pictures will not shrink much. On a small, already tight PDF the output can even grow, because the object stream and linearization carry their own overhead. The tool says so honestly when it happens. Run it before step four so you can compare sizes while the file is still unlocked.
Watermark the images before you build the PDF
If a draft copy has to be marked, add the watermark to the page images with the image watermark tool before step one. It supports text watermarks with position, size, colour, opacity, rotation and a tiled mode, and it can process a batch of images in one pass. Two limits matter: it places text, not an image or a logo, and the font stack is Latin, so non-Latin scripts may not render as you expect. Do the watermarking on the loose images, because the PDF tools here do not add a text or image overlay to a finished document.
Where this workflow stops
This is a private browser workflow, not a document management system, and a few jobs fall outside it.
- No searchable PDF output. Step two gives you searchable text as a sidecar file. If a system insists on a PDF whose own text layer is searchable, this workflow does not produce that file.
- OCR is only as good as the scan. Handwriting, skewed pages, staples and low-contrast photocopies produce errors. Clean, straight, 300 dpi scans are where recognition works well.
- Compression is not image compression. The PDF compressor cannot shrink a picture-heavy scan. Re-scanning at a lower resolution or compressing the images before step one is the way to cut that kind of size.
- Encryption has a size ceiling. The protect and compress tools both stop at 50 MB, so a very large scan needs splitting first.
- One document at a time. Each step works on a single file, so a batch of separate documents is a batch of runs rather than one job.
- Engines download on first use. The PDF and OCR engines are a few megabytes, fetched once and cached. That is the cost of doing the work on your machine.