Find duplicate files without installing anything
Photos, videos, audio and documents. Point this page at a folder and it tells you what is repeated, by comparing the contents of every file — never its name.
100% private. Your files are processed on your device and never uploaded to any server. You can disconnect from the internet after this page loads and watch it keep working. Handles folders with tens of thousands of files. Files over 2 GB cannot be compared, because the browser cannot load them into memory in one piece.
What makes FileDupes different: you install nothing
Every duplicate file finder on the market is a program you have to download, install and often pay for. This one is not. It is a web page that works on your files using your own computer's power.
Nothing to download
No installer, no adware, no toolbars that appear out of nowhere. You open the page and it is already working.
Your files never travel
The folder is read from your disk and the analysis runs on your machine. Disconnect from the internet and see for yourself: it keeps working.
No permanent permissions
You grant access to one folder, and that permission dies when you close the tab. Nothing stays installed watching your disk.
Works on any machine
Windows, macOS, Linux, Android, iOS or iPadOS. If the browser opens, the tool runs. No version to update and no device left out.
Which systems it runs on
There is no Windows build and no Mac build: there is one web page, and any device with a browser opens it. What does change from system to system is whether the browser can also move the files for you, or only read them.
| System | Finds duplicates | Sets the extras aside |
|---|---|---|
| Windows | Yes | Yes, automatic |
| macOS | Yes | Yes, automatic |
| Linux | Yes | Yes, automatic |
| Android | Yes | Yes, automatic |
| iOS · iPhone | Yes | You move them |
| iPadOS | Yes | You move them |
On a computer, full mode needs Chrome, Edge, Brave or Opera. Firefox and Safari can read the folder but not modify it: you get exactly the same report and the same CSV, and you move the files yourself.
On Android full mode works: the browser asks for permission on the folder you pick and sets the duplicates aside exactly as on a computer.
On iPhone and iPad, the system will not let a web page write inside your folders. There you pick the files, get the same report with the path of every copy, and do the cleanup in the Files app or on a computer.
The Downloads folder is a special case, and on every device. The browser will not let any web page open it: desktop Chrome shows "this folder cannot be opened", and Android has greyed out the confirm button since version 11. This is not a fault in this tool. Downloads is where everything from the internet lands, installers included, which is why it is protected. The way round it, and it works in both cases: create a subfolder inside Downloads, move the files there and pick that subfolder — subfolders are allowed.
We say this up front rather than in the small print, because you should know it before you start and not after.
The file name does not matter
This is the most common confusion, so let us be blunt: the tool never looks at what a file is called. You could have the same photo saved as holiday.jpg, IMG_0042.jpg and untitled (3).jpg, in three different folders, and all three would come out grouped together.
What gets compared is the content. Every file gets a SHA-256 fingerprint, a 64-character number summarising its bytes. If two files produce the same fingerprint, they are the same file. Change a single bit and the fingerprint changes completely.
That fingerprint is indifferent to format, which is why it works just as well on a video, a song, a PDF or a spreadsheet as it does on a photo.
Full guide: how to find duplicate files and which folders you should never touch →
One detail that keeps it fast
Reading a 4 GB file end to end just to fingerprint it takes time. So before anything else, files are grouped by exact size: if no other file in the folder weighs exactly the same, it cannot have a twin and there is no need to read it. In a folder full of video, that single trick is the difference between seconds and half an hour.
What it detects for each file type
It is worth being honest about the limits, because not every format allows the same thing:
- Any file — exact, byte-for-byte copies. Video, audio, documents, archives, executables. No margin for error.
- Images — all of the above, plus the same photo across different files: re-compressed, resized, or back from a WhatsApp round trip.
- Video and audio — exact copies only, for now. Recognising the same song at two different qualities means analysing the sound itself, which is a far heavier engine. We would rather not promise it before we have it.
- Documents — exact copies only. Two PDFs with the same text but produced by different programs are different files, and the tool treats them as such.
How it decides two photos are the same
The visual comparison happens in two stages, and that is what avoids the false positives almost every tool suffers from.
First, a perceptual hash proposes candidate pairs. It is fast but imprecise — a folder of screenshots fools it easily, since they are all pale with thin text — so it never decides anything on its own.
Then comes the real comparison: each candidate is reduced to a 32×32 grayscale grid and the correlation between the two is measured. That number, as a percentage, is what you see. Because it is normalised by contrast, it ignores brightness and compression changes but collapses the moment the content differs.
Two more guards: images with different aspect ratios are never grouped, because a crop is not a duplicate; and images without enough detail are only grouped if they are literally the same file.
The screenshot of a photo
This is the most common case and the hardest one. You have the photo, and you also have the screenshot you took of it: to you they are the same image, but as files they look nothing alike. The screenshot has the shape of the screen, not of the photo, and the photo sits in the middle between black or white bands filling half the height. On top of that come the clock and the phone icons. Compared as they are, they match 66%.
What the tool now does is strip those bands before comparing. Telling padding apart from content has a catch, because the sky in a photo is also a flat band: the difference is that padding keeps the same tone row after row, while a sky shifts. The status bar, which interrupts the black band, is stepped over.
With the padding gone, those two images match 99.7%. And the part that matters: the signature of the whole image is kept unchanged. The uncropped comparison still happens, so nothing that was detected before has stopped being detected.
It is tested against the eight ways a phone screenshot actually looks: with the status bar over the black band or over the photo itself, with a notification pulled down on top, with rounded corners and a notch, in light-background galleries, with the photo off-centre, and as the recompressed copy WhatsApp forwards. All eight match above 99%.
Two cases still do not group, deliberately: when the screenshot is full-screen and the photo has been cropped at the sides to fill it. Part of the image is missing there, and that is a reframe, not a copy.
Nothing gets deleted
The tool never calls a delete operation. Everything you select is moved into a subfolder called _duplicates inside the folder you scanned. You open it, look through it, and delete it yourself when you are sure. If something went wrong, dragging the files back takes seconds.
Guides on duplicate files
Finding duplicates is the easy part. The hard part is knowing which ones are worth removing, which must never be touched, and why they appear on their own. These guides come out of the work of building this tool, measurements included:
- How to find duplicate files without installing anything — why searching by name fails and how they are actually found.
- Exact duplicates and visual duplicates — one kind can be cleared without looking and the other cannot. The difference, and the failure that ruins most finders.
- Screenshots of your photos — the same picture matching 66%, and how it gets to 99.7%.
- Why WhatsApp turns one photo into five files — the most efficient duplicate generator you own.
- Folders you should never scan — some repetition holds entire programs together.
- Duplicates on your phone — more duplicates, fewer tools. What each system allows.
- Why you have three copies of the same video — one minute of video weighs what four hundred photos weigh.
Frequently asked questions
Does it work for more than photos?
Yes. Exact-copy detection works with any format: video, audio, PDF, Word, Excel, archives, anything. The visual comparison, which recognises the same photo at another size, is images only.
Does it find duplicates with different names?
Yes, and the name is never used. Contents are compared through a SHA-256 fingerprint of the bytes. Two files with the same fingerprint are the same file even if they are named differently and sit in different folders.
Do I have to install anything?
No, and that is the whole point of this tool. It runs inside the browser you already have. No program is downloaded, nothing is left on your machine, and there is no version to update.
What exactly does the percentage mean?
It is how closely two images match when compared pixel by pixel on a 32×32 grid. 100% with the "same file" label means identical byte for byte. 99% means the same photo at another compression or size.
What if it groups things that are not the same?
Set the sensitivity to "Strict". Groups are recalculated instantly without re-reading the folder. And the percentage is visible on every file: if you see a 94% on something that does not belong, uncheck it.
Are my files uploaded anywhere?
No. The browser reads the folder from your disk and all the work happens on your machine. Disconnect from the internet after the page loads and the scan still finishes.
Why can't I select the Downloads folder?
The browser blocks it, not this tool, and it happens on computers as well as phones. Chrome keeps a list of folders no web page may open — system folders, installed-program folders, and Downloads too, because that is where everything from the internet lands. On a computer you will see "this folder cannot be opened"; on Android, since version 11, the confirm button is greyed out. The fix is the same either way: create a subfolder inside Downloads, move the files there and pick that subfolder. Subfolders are not blocked.
Which systems does it work on?
All of them: Windows, macOS, Linux, Android, iOS and iPadOS. Being a web page, any device with a browser opens it. The difference is in setting files aside: on a computer with Chrome, Edge, Brave or Opera it does that for you; on phones and tablets the system will not let a web page write into your folders, so there it scans and hands you the list, and you do the cleanup.
Which browsers get the full version?
Desktop Chrome, Edge, Brave and Opera can read and write the folder you choose. Firefox and Safari can read it, so you still get the full report and the CSV export, but you move the files yourself.
How many files can it handle?
Tens of thousands. The analysis is spread across several parallel workers, and the size pre-filter avoids reading files that cannot possibly have a twin.