Initializing... drag & drop files here
Supports: TAR
.tar archives from your computer. Batch is supported — process several archives in one pass..tgz, or Individual Archives to produce one .tgz per input .tar. For converting an existing TAR straight to TGZ, Individual Archives is the typical pick..tgz is returned for download. No sign-up, no watermark, no software install.A plain .tar archive bundles files but performs no compression — tar literally stands for tape archive, designed in 1979 for Seventh Edition Unix to write streams to magnetic tape. Adding gzip on top (producing .tar.gz or its short form .tgz) typically shrinks the archive 50-80% for source code, text, logs, and other compressible payloads. The two extensions are byte-identical: archive.tgz is just archive.tar.gz renamed so it fit MS-DOS 8.3 filename limits, and the convention stuck.
.tgz..tar.gz/.tgz: kernel.org tarballs, GNU project releases, and the contents of npm packages (which are .tgz under the hood) all use this format.tar -xzf archive.tgz unpacks gzipped tar in a single step on Linux, macOS, BSD, and WSL. No separate decompress-then-untar step required.npm pack and npm publish produce .tgz files; pip wheels and Python sdists, RubyGems, and Composer all rely on tar+gzip conventions. Converting a plain .tar to .tgz is often the last step before publishing.If you need the opposite direction (strip gzip back to plain tar), use TGZ to TAR. For a heavier compression ratio than gzip, see TAR to 7Z or TAR to ZIP.
| Property | TAR | TGZ (TAR.GZ) |
|---|---|---|
| Compression | None — archive only | Gzip (DEFLATE, RFC 1952) |
| Typical size vs source | ~100% (sum of inputs + headers) | ~20-50% for text/code, ~90-100% for already-compressed media |
| Extraction command | tar -xf file.tar |
tar -xzf file.tgz |
| Random access | Sequential only | Sequential only (must decompress stream) |
| Stores Unix permissions, owners, symlinks | Yes | Yes |
| Stores Windows ACLs | No | No |
| Streamable over a pipe | Yes | Yes (gunzip | tar -x) |
| Standardized in | POSIX ustar (1988), pax (2001) | RFC 1952 (gzip) over POSIX tar |
| File signature (magic) | ustar at offset 257 |
1f 8b at offset 0 (gzip header) |
| Common uses | Tape backup, intermediate archive | Linux source releases, npm packages, log bundles |
| Workload | Expected gzip ratio | TGZ a good choice? |
|---|---|---|
| Source code, configs, plain text | 3-5x smaller | Yes — universal default |
| HTML, JSON, XML logs | 5-10x smaller | Yes |
| Database dumps (SQL text) | 4-8x smaller | Yes |
| JPEG / PNG / WebP images | Roughly the same size | No real benefit — already compressed |
| MP4 / H.264 video, MP3 audio | Roughly the same size | No — use plain .tar instead |
| Mixed: code + binaries + media | 1.5-3x smaller overall | Usually yes |
| Need maximum compression | — | Consider .tar.xz or .tar.zst instead |
Yes — byte-for-byte identical when produced from the same inputs with the same gzip settings. The .tgz extension is a contraction created so the name would fit the old MS-DOS 8.3 filename limit (eight-character base, three-character extension). On any modern OS you can rename archive.tar.gz to archive.tgz (or vice versa) and every tool will keep working. We use .tgz as the output extension here for the shorter, more portable name.
No. The tar payload is wrapped in a gzip stream, not repacked. Inner file names, permissions (mode bits), owner/group IDs, modification times, symbolic links, and hard links are preserved exactly as they were in the source .tar. Extracting with tar -xzf yields the same tree you started with.
It depends entirely on what's inside. Text-heavy content (source code, JSON, XML, SQL dumps) typically compresses 3-10x — a 100 MB tar of code might land at 10-30 MB as tgz. Already-compressed content (JPEGs, MP3s, MP4s, PDFs with image-only pages) compresses by only a few percent at best, since gzip can't find redundancy that the underlying codec has already removed. Mixed archives usually end up in the 50-70% range.
Gzip is specified in RFC 1952 and uses the DEFLATE algorithm from RFC 1951. DEFLATE combines LZ77 (a sliding-window dictionary that finds repeated byte sequences) with Huffman coding (variable-length symbol encoding that assigns shorter codes to more frequent values). The same algorithm powers .zip, PNG image compression, and HTTP Content-Encoding: gzip.
Yes. Windows 10 (build 17063+) and Windows 11 include a built-in tar.exe on the command line — tar -xzf archive.tgz works in PowerShell or Command Prompt. For a GUI, 7-Zip, WinRAR, and PeaZip all handle .tgz natively. Note that 7-Zip and some other tools may show the archive as two nested layers — the outer gzip and the inner tar — requiring two extract steps if you click through manually.
.tgz — best default for Linux/Unix audiences, npm, and source releases. Universal tooling..zip — best for Windows and cross-platform users who want one double-click extract. Loses Unix permissions..7z — smaller than tgz (often 20-40% better) but requires 7-Zip-compatible software; not built into most OSes..tar.xz — slower to compress but smaller than tgz; common for kernel and large source releases..tar.zst — modern Zstandard; faster and often smaller than gzip, but tooling is still catching up on older systems.No. This page is for plain .tar (uncompressed) input. If you already have a .tgz and want to repack at a different compression level or switch to a different compressor, run TGZ to TAR first to strip gzip, then convert the resulting .tar to your target format.
Free conversions are sized for typical source bundles and backups — most users won't hit the cap. If you have a multi-gigabyte archive (full disk image, large media library), upload time becomes the practical constraint rather than the conversion itself. For very large jobs, consider running tar czf locally with gzip --best or pigz for parallel gzip.
Uploads are processed and then deleted on a short retention window. We don't index, share, or scan your archive contents. For sensitive backups (production database dumps, customer data) we recommend gzipping locally with the tar binary on your own machine — no upload, no network exposure.