ZipForge · ZIP · GZIP · 0 uploads

A folder in.
A .zip out.
Nothing uploaded.

Pick a handful of files and get a real .zip — every CRC32 and Deflate stream written by this page. Drop a .zip back in and read its tree: original and packed size per file, a preview for images and text, one file out or all of them. GZIP gets its own tab. Nothing is sent anywhere, so the request counter stays at zero.

ZIP · GZIP·checked 2026-11-19 0 bytes uploaded 0 external requests Local header · Central directory · EOCD

Files in, one archive out

read on the device only

What actually happens

no server in the loop

01Read, never send

Choosing a file gives the page a File object, not an upload. It is read into an ArrayBuffer with the File API, and that buffer is all the page ever handles.

02Pack

Each entry gets its CRC32, then a Deflate attempt; the shorter of the two wins. Local headers, the central directory and the EOCD are assembled into one blob and offered as a download.

03Unpack

Opening an archive means finding the EOCD from the end of the file, walking the central directory, and slicing each entry out by its offset. Deflate entries go back through DecompressionStream.

04Preview without extracting

An image becomes a blob URL the page draws; a text file is decoded as UTF-8 and shown in a pre block. Nothing is written to disk until you press download.

Questions people actually ask

five answers
Is this a real .zip, or only this page can read it?
A real one. Local file headers, the compressed bytes, a central directory and the EOCD — the same four parts every archiver writes. Windows Explorer, macOS Archive Utility, 7-Zip and unzip open the download without a warning.
What if the browser has no deflate-raw?
The archive is still valid, just bigger: entries are stored and the method byte says 0 instead of 8. The result panel names the method, so the number you see is the one you got.
Can it open a .zip with Korean filenames?
Yes. A name is decoded as UTF-8 when the UTF-8 flag is set; when it is not, the page tries UTF-8 first and falls back to EUC-KR, which is what Korean Windows archivers wrote for years. A name that survives neither is shown as bytes rather than silently mangled.
Is anything uploaded, cached or logged?
No. One HTML file, no fetch, no XHR, no WebSocket, no analytics, no external font or stylesheet. The footer counts outbound requests and keeps the number it sees, which is zero.
Will my photos get smaller?
Barely. A JPEG, MP4, PNG or PDF has already had its redundancy squeezed out, so Deflate usually returns it a few percent larger and the page then stores it unchanged. Text, JSON, CSV, SVG, logs and source code are where a .zip earns its keep.