Zlib and Deflate Compression to Base64

How to compress with zlib and encode as Base64

1. Type or paste the text, or drop a file.
2. Format starts on zlib (deflate). Switch to Raw deflate when the receiver wants the bare stream, or to gzip.
3. Optionally tick URL-safe (- _) or Wrap at 76.
4. Copy output, or download the Base64 .txt or the binary .zz / .deflate file.

Swap sends the output to the Decompress side to confirm the round trip.

The same text in three wrappers

Hello, world! (13 bytes), compressed at the default level:

The deflate stream is the same 15 bytes in all three, yet the Base64 looks different because the 2-byte and 10-byte headers shift where each byte falls in the 3-byte groups. zlib’s header is 78 9C at the default level, which is why zlib Base64 begins eJ. At level 9 it is 78 DA, Base64 eN.

Which call produces which format

Most “deflate” bugs come from two sides using the same word for different wrappers:

Note the traps: .NET’s DeflateStream and Go’s compress/flate are raw, while Node’s and the browser’s “deflate” are zlib. A C# service and a Node client that both “use deflate” will not understand each other without a change on one side.

Where each wrapper appears

• zlib: PDF FlateDecode streams, PNG image data, Git objects, many binary protocols and game save formats.
• raw deflate: the SAML HTTP-Redirect binding (the SAML decoder handles it), entries inside ZIP files, WebSocket permessage-deflate.
• gzip: files, HTTP, logs. The gzip page covers it.

Zlib + Base64 in code

import base64, zlib

packed = base64.b64encode(zlib.compress(b"Hello, world!")).decode()
# 'eJzzSM3JyddRKM8vyklRBAAgXgSK'
text = zlib.decompress(base64.b64decode(packed))

// Node.js: raw deflate for a SAML-style payload
const zlib = require("zlib");
const packed = zlib.deflateRawSync(Buffer.from("Hello, world!")).toString("base64");
// '80jNycnXUSjPL8pJUQQA'

To read such strings back, use the Base64 gzip and zlib decoder, which detects all three wrappers. More in Python and Node.js.

Common problems

• incorrect header check: zlib decoder, raw or gzip data. Match the wrapper.
• invalid stored block lengths or invalid block type: raw-deflate decoder fed zlib data; the two header bytes are read as deflate blocks.
• The result is longer than the input: short strings do not compress. Below roughly 50–100 bytes, plain Base64 is shorter.
• Output does not match another tool byte for byte: the compression level and the library both change the bytes, and any correct decompressor returns the same original from either.

FAQ