File Size Analyzer
Drop a folder and see where the gigabytes actually went — with duplicates found by content, not by name.
Drop files or a folder here
Read in this page. Nothing is uploaded, moved or deleted.
- Total
- —
- Files
- —
- Extensions
- —
- Reclaimable
- —
This tool reports, it does not delete
Nothing is written to disk and nothing is uploaded. When you find the 4 GB export you forgot about, delete it in your own file manager — that is where a destructive action belongs.
Everything here runs on your device. Nothing you enter is uploaded or stored.
Drop a folder in and the page tells you where the space went: a treemap sized by bytes, the twenty biggest files with their share of the total, and — if you ask for it — every set of genuinely identical files found by hashing their contents. Everything is read locally. Nothing is uploaded, and nothing is deleted.
How to use it
- Drop files in, or press Choose a folder to walk a whole directory tree.
- Read the treemap. Switch between extension, family and folder.
- Press Find duplicates to hash the files that could possibly match.
- Export a CSV of the full inventory if you want to keep it.
Why the treemap is squarified
The obvious way to draw a treemap is to slice the rectangle: divide the width by the first value, divide the remainder by the second, alternate direction, repeat. It is about ten lines of code and it is unusable, because the long tail of any real folder is thousands of small files, and each one becomes a strip a fraction of a pixel wide. You cannot read its label, you cannot see its colour, and on a touchscreen you cannot hit it.
This tool uses the squarified layout instead, which greedily fills a row while doing so keeps the row’s worst aspect ratio improving, then lays the row down along the shorter edge and recurses into what is left. The result is rectangles close to square, which means a category holding two per cent of your data is a small square you can actually see and label rather than a hairline.
Only the ten largest groups get their own tile. Everything past that is merged into one grey more tile, because a legend with forty rows is not a legend.
Size is grouped by extension, on purpose
The browser reports a MIME type for a file, and it is frequently an empty string — for anything unusual, anything without a registered type, and quite a lot of ordinary files depending on the operating system. Extensions are what actually survive, and they are also what you think in. The family view maps around 120 extensions into Image, Video, Audio, Document, Spreadsheet, Slides, Archive, Code, Font and Program so you can see the shape of a folder at a glance, and anything unrecognised lands in Other rather than being silently dropped.
The sentence about small files
Under the treemap you will usually see something like 1,204 files under 8 kB account for 0.3% of the total. This is deliberately prominent, because the instinct when a disk fills is to go hunting for clutter — thousands of little files feel like the problem. They almost never are. Four video exports and a forgotten virtual machine image are the problem. Naming the tail explicitly is the fastest way to stop you spending an afternoon in the wrong folder.
What it does not do
It does not delete anything, move anything, or write to disk at all. It does not read inside archives — a 2 GB ZIP is 2 GB of ZIP here, and ZIP Extractor is the tool that looks inside. It cannot see files you have not handed it, so it cannot report on your whole drive the way a native utility can; the trade is that nothing you point it at is ever transmitted.
Which tool for which job
| You want | Use |
|---|---|
| To see where space went | This tool |
| To find identical files | This tool, Find duplicates |
| To look inside an archive | ZIP Extractor |
| To make a file small enough to send | File Splitter |
| To shrink photos rather than count them | Image Compressor |
Questions
Why find duplicates by hash instead of by name and size?
Because name and size both lie, in opposite directions. Two files called invoice.pdf in different folders are usually two different invoices, so matching on name produces false positives that would have you deleting real data. Matching on size alone is worse — every 4 kB stub, every 1×1 tracking pixel and every empty Office template shares a size with dozens of unrelated files. A SHA-256 digest over the actual bytes is the only comparison that cannot be wrong, because if two digests match the files are identical. The tool still uses size as a pre-filter, since two files of different sizes cannot possibly be identical, so only files whose size is shared with at least one other file ever get read. On a normal folder that skips well over ninety per cent of the bytes, which is the difference between a second and a minute.
Can it read a whole folder including subfolders?
Yes. Choose a folder uses the directory-picker mode of a file input, which hands the page every file in the tree along with its relative path — so the folder view groups by top-level directory and the largest-files table shows full paths. Drag and drop gives you the files you dropped but not their directory contents, which is a browser restriction rather than a choice here. If you want the tree, use the button.
Is a treemap actually better than a sorted list?
For finding the one big thing, no — that is what the largest-files table is for. A treemap answers a different question, which is what proportion of the whole each category holds, and whether you have many medium things or a few huge ones. The layout here is squarified, meaning it keeps every rectangle close to square instead of slicing the space into strips. That matters because the naive slice-and-dice layout renders a 2% item as a rectangle one pixel wide, which is unreadable, unlabelable and impossible to tap on a phone.
Does anything get uploaded or changed?
No, in both directions. Files are read in the page using the File API and hashed with Web Crypto, and no request carries their contents anywhere. Equally, nothing is written — this tool reports and does not delete. When you find a duplicate set or a forgotten 4 GB export, you delete it yourself in your own file manager, which is where a destructive action belongs.
Last updated