Base85 / Ascii85 Encoder and Decoder

How to encode and decode Ascii85

1. Type or paste text into Input · Text. The Ascii85 appears in Output · Ascii85 (Adobe).
2. Under the output, tick Add <~ ~> delimiters for the PostScript/PDF framing, and untick z for zero groups if your decoder does not support the shortcut.
3. To use another Base85 alphabet, change To to Base85 (Git / Python b85), Z85 (ZeroMQ) or RFC 1924 (IPv6 Base85).
4. To decode, click Ascii85 (Adobe) → Text in the direction toggle. The decoder accepts the <~ ~> delimiters, z, y and any whitespace, so line-wrapped data from a file pastes straight in.
5. For binary results, set To to Hex or click Save bytes.

How Ascii85 works

Ascii85 takes 4 bytes at a time, reads them as one 32-bit big-endian number and writes it as 5 base-85 digits. Each digit is a character from ! (value 0) to u (value 84). Five characters for four bytes is a 25% overhead, against 33% for Base64.

The classic example is Man (with the trailing space): the bytes 4d 61 6e 20 become 9jqo^. When the last group is short, the encoder pads it with zero bytes and writes only n + 1 characters for n bytes, so Hello (5 bytes) becomes 87cURDZ: five characters for Hell and two for o.

Two shortcuts shorten common patterns. z replaces a group of four zero bytes, which is !!!!! written out. y replaces four spaces in the btoa variant. This decoder reads both, and the encoder writes z (switchable) but never y.

Base85 variants compared

The Git and RFC 1924 alphabets are the same, but RFC 1924 treats a whole 16-byte address as one 128-bit number instead of four 4-byte groups. RFC 1924 was published on 1 April 1996 and proposes writing 1080::8:800:200c:417a as 4)+k&C#VzJ4br>0wv%Yp. The converter reproduces it with From set to IPv6 address and To set to RFC 1924 (IPv6 Base85).

Ascii85 in PDF and PostScript

PostScript Level 2 introduced Ascii85 for embedding binary data, such as images and fonts, in a text file. The data is framed as <~ … ~>. In PDF, a stream with /Filter /ASCII85Decode holds the same encoding, terminated by ~>. The decoder here finds the end marker with or without the opening <~.

Ascii85 is often the outer filter of a chain, for example /Filter [/ASCII85Decode /FlateDecode]. Decoding the Ascii85 then gives zlib-compressed bytes rather than readable content.

Git uses Base85 differently. In a GIT binary patch, each line starts with one length character (A–Z for 1–26 bytes, a–z for 27–52), followed by Base85 of the zlib-compressed data. Remove that first character from every line before decoding, then inflate the result.

Ascii85 and Base85 in Python

The standard library has both alphabets:

import base64

base64.a85encode(b"Hello")                 # b'87cURDZ'
base64.a85encode(b"Hello", adobe=True)     # b'<~87cURDZ~>'
base64.a85encode(b"    ", foldspaces=True) # b'y'
base64.b85encode(b"Hello")                 # b'NM&qnZv'
base64.a85decode(b"<~9jqo^~>", adobe=True) # b'Man '

a85decode needs adobe=True to accept the delimiters. See the Python page for more encoding recipes.

Common problems

• "z" at position … is only allowed between 5-character groups. A z appears inside a group. The data is damaged, or it is not Ascii85; Z85 and Git’s Base85 use z as an ordinary digit.
• the group starting at position … is out of range. Five characters encode a value above 2^32 − 1. This happens when Git Base85 or Z85 is decoded as Ascii85; switch From.
• Unexpected character such as v, { or |. Everything after u except the y and z shortcuts is outside the Ascii85 alphabet, another sign of a different variant.
• The output has extra zero bytes at the end. Your encoder padded the last group and wrote all 5 characters. Ascii85 writes n + 1 characters for a final group of n bytes.

FAQ