Do Online File Converters Keep Your Files?

Last updated: 2026-11-10 · Checked and published by Leo, who runs Hanji

How upload-based and in-browser file tools differ, what can happen to an uploaded file, and two simple tests that show where any tool sends your data.

When you search for “HEIC to JPG” or “compress PDF”, the results look alike: a box to drop a file in and a download button. What happens between those two steps can be completely different. Some sites send your file to their servers and process it there. Others do all the work inside your browser, and the file never leaves your device. From the outside you usually cannot tell which is which, and the difference matters most for exactly the files people tend to convert in a hurry: ID scans, contracts, medical letters.

This guide explains the two designs, what can happen to a file once it is uploaded, and how to check any tool yourself in a couple of minutes. It ends with what Hanji does and does not do, including the parts that are not about files.

Two ways a file tool can work

Upload-based tools send your file to a server. The server converts it, stores the result for a while, and gives you a link to download it. The heavy work, such as reading PDFs or decoding HEIC photos, happens on the service’s machines.

In-browser tools send the program to you instead. The page loads code that runs on your own phone or computer, reads the file from your disk, does the conversion in memory and hands you the result. There is no copy on anyone else’s machine because the file was never sent.

Neither design is automatically good or bad. Upload-based services can run heavier processing, and a reputable one may handle files carefully. But with an in-browser tool, the question of what happens to your file on a server does not arise, and you can verify that yourself instead of relying on a promise.

What can happen to an uploaded file

An upload-based service may state that files are deleted after a period of time, and it may well do so. Still, once a file is on a server, several things are outside your control and usually outside your view:

  • Retention. “Deleted after a few hours” describes the main copy. Backups, caches and failed jobs can keep copies longer, depending on how the service is built.
  • Logs. Servers record requests. File names alone can say a lot (“passport_scan_jane_doe.pdf”), and some systems log more than names.
  • Other companies. A service may run on rented cloud servers or pass files to other providers for processing or storage. Each one is another place the file exists.
  • People. Staff with access to the systems can, in principle, see what is stored there.
  • Breaches. Any server can be broken into. A service that holds thousands of ID cards and contracts at any moment is a more interesting target than one that holds none.
  • Change. Privacy policies get rewritten, and companies get sold. The promise that applied when you uploaded may not be the one that applies later.

None of this means a given service is misusing files. It means that uploading turns “my file” into “my file, plus a copy I have to trust someone else with”.

How to check a tool yourself

You do not need to read the code or trust the marketing. Two tests answer the question directly.

Test 1: watch the Network tab

This works in Chrome, Edge and Firefox on a computer.

  1. Open the tool’s page. Press F12 (or Cmd+Option+I on a Mac) to open developer tools and choose the Network tab.
  2. Click the clear button so the list is empty, and keep the tab open.
  3. Pick a harmless test file of a few megabytes and run the conversion.
  4. Look at the new rows. Right-click the column headers and turn on Method if it is not shown. Requests that send data are usually POST or PUT. Click one and open its Payload (or Request) section.

If you see a request whose payload is your file, often marked multipart/form-data with your file name in it, or one that takes a long time to send while the progress bar moves, the file is being uploaded. If the only requests are the page loading its own scripts, styles and images, the work is happening on your device. On Hanji you may also see one small POST to /cdn-cgi/rum: that is Cloudflare’s visitor statistics, carrying the page address and load timing, not your file. When you run a tool you may also see a small POST to /e: anonymous usage statistics (the tool, whether it worked, file types such as HEIC to JPG, the number of files, a size range and the time it took). It has no file name and no file content. Open its payload and check both.

On many sites, including this one, you will also see requests to advertising services. Those are worth knowing about; click them and check their size and payload to see that none of them carries your file.

Test 2: go offline after the page loads

This one also works on a phone.

  1. Open the tool and let the page load completely.
  2. Run it once with a test file. Some tools load extra code the first time a feature is used; on Hanji, for example, the HEIC decoder and the PDF engine load on first use.
  3. Turn on airplane mode or switch off Wi-Fi.
  4. Run the tool on another file.

If it still works, the processing cannot have happened on a server. If it fails with a network error, the tool needed to send something. For extra caution, close the tab before you reconnect.

What I measured on Hanji

On 10 November 2026, in Chrome 156 on Windows, I opened HEIC to JPG on the live hanji.page, chose one 11,320-byte HEIC file the way Test 1 describes, and wrote down every request that followed.

Request Data sent What it is
GET hanji.page/_astro/heic-to…js none the HEIC conversion code, downloaded on first use
POST hanji.page/e 114 bytes anonymous usage statistics
2 blob: addresses none temporary addresses made inside the browser, never on the network
POST hanji.page/e 119 bytes anonymous usage statistics

The file is 11,320 bytes and the largest request that sent any data is 119 bytes. The file, whole or in part, cannot fit in it.

This record has limits. It lists only the requests after the file was chosen, not the ones made when the page first opens (fonts, visit counting). The test file is a small sample Hanji made itself, and browser extensions can add requests. So do not just trust this table: run Test 1 above on your own device.

Which files deserve the most care

Some files are harmless to upload anywhere: a meme, a photo of your lunch. Others give a stranger most of what they need to pretend to be you:

  • Identity documents: passports, driving licences, residence cards, national ID cards, and photos of them.
  • Official records: birth or family certificates, residence records, tax statements.
  • Contracts and financial papers: leases, employment contracts, bank statements, anything with an account number and a signature.
  • Medical documents: prescriptions, test results, insurance claims.
  • Photos with location data. Phone photos often carry GPS coordinates in their metadata.

A few habits help no matter which tool you use:

  • Send only what was asked for. If a form wants the front of an ID card, do not attach the back. Cover or crop numbers the receiver does not need when that is allowed.
  • Mark the copy with its purpose. Writing “Copy for the rental application at X only” across a photocopy, before you photograph it, makes it less useful if it ends up elsewhere.
  • Clean up afterwards. Photos of documents sit in your camera roll and may be backed up to a cloud photo library automatically. Delete them, and the converted files in your Downloads folder, when you are done.
  • Watch browser extensions. An extension that Chrome describes as able to “Read and change all your data on all websites” can see pages you use, including file tools. Keep extensions to the ones you trust.

What Hanji does and does not do

Every tool on this site, including the image converter, Compress PDF, Merge PDF and the passport photo maker, runs in your browser. There is no server that receives files, so we have no way to see, keep or lose them. Both tests above work on every page, and we would rather you run them than take our word for it.

Some specifics:

  • No account, no file storage, no watermark. Nothing you process is saved anywhere except where you download it.
  • Images are re-drawn, not copied. Converting, compressing or resizing an image creates a new file from the pixels, so the result does not carry the original’s metadata such as GPS location. The one exception: when an image is already under the compression target, it is handed back unchanged, metadata included.
  • Text stays on the page too. What you type into the word counter or QR code generator is not sent anywhere.

And the parts that are not about files, stated plainly:

  • The site shows ads through Google AdSense. Google’s ad code loads on our pages, and Google and its partners use cookies to choose ads based on your visits to this and other sites. Visitors in the EEA, UK and Switzerland are asked for consent before personalised ads are shown, and anyone can turn off personalised ads in Google’s ad settings. Like any script on a page, the ad code runs alongside the tools; the Network tab lets you see that its requests do not carry your files.
  • Our hosting provider may keep standard access logs, such as IP address, browser type and time of visit, for security and operations.
  • The word counter remembers your target length and counting rules, and autosaves your text, in your browser’s local storage, on your device only. Nothing is sent to us. Switch the autosave off or press Clear to delete the saved text.
  • We cannot protect files after they leave the page. Once you download a result, it is in your Downloads folder, your cloud backup and wherever you send it, like any other file.

The full details are in the privacy policy, and the reasons behind the design are on the About page. If you are working with photos of documents, the guide to scanning documents with your phone shows the whole process, from taking the photo to a PDF that fits the upload limit.

Guides