How is Base91 different from Base64?

Base64 maps 3 bytes to 4 characters, a 33% size increase. basE91 squeezes about 13 bits into 2 characters, so the overhead drops to roughly 23%. It uses 91 printable ASCII characters instead of 64.

Which characters does the alphabet use?

It uses 91 printable ASCII characters and deliberately excludes the dash, backslash and apostrophe-style quoting hazards, making the output safe to embed in source code and many text formats.

Is Base91 a standard like Base64?

It is a de-facto standard defined by Joachim Henke's reference implementation rather than an IETF RFC. This tool follows that canonical bit-packing algorithm so output is interoperable with other basE91 libraries.

Can I encode binary files?

This in-browser tool operates on text input encoded as UTF-8 bytes. The underlying algorithm handles arbitrary bytes, but the input box is text, so use it for strings and UTF-8 payloads.

Why does my decoded output look garbled?

Decoding produces raw bytes. If those bytes are not valid UTF-8 text — for example, they came from a non-text binary — the tool cannot render them as readable characters and reports an error.

What is the Base91 Encoder/Decoder?

Encode and decode text or binary data with basE91, the printable-ASCII scheme that is more compact than Base64. Runs entirely in your browser using the canonical Joachim Henke 91-character alphabet. Free and instant. It runs free in your browser on Gera Tools, with nothing uploaded.

Base91 Encoder/Decoder

Name: Base91 Encoder/Decoder
Creator: Gera Tools
License: https://creativecommons.org/licenses/by/4.0/

What basE91 does

basE91 (sometimes written Base91) encodes arbitrary bytes into a string of printable ASCII characters so binary data can travel safely through text-only channels. Its appeal over Base64 is efficiency: it carries more information per output character, so encoded payloads are noticeably shorter.

How it works

The encoder treats the input as a stream of bits and reads them into a bit buffer. Whenever the buffer holds enough bits it pulls a value of either 13 or 14 bits — choosing 14 bits when the low 13 bits would exceed 88, which keeps every value below 8281 (that is 91 * 91). Each value is then written as two characters: the low digit is value mod 91 and the high digit is value / 91, both indexed into the 91-character alphabet.

v = buffer & 0x1FFF        // try 13 bits
if v > 88:  shift out 13 bits
else:       v = buffer & 0x3FFF, shift out 14 bits
emit alphabet[v % 91], alphabet[v / 91]

Decoding reverses this: characters are read in pairs, recombined as low + high * 91, and the bits are streamed back out as bytes.

Why the 13/14 bit trick gives better density

The clever part of basE91 is the adaptive bit-width. If it always used 13 bits per two-character pair, each pair would carry log₂(91²) ≈ 13.15 bits of entropy, slightly better than Base64’s 6 bits per character. But by sometimes consuming 14 bits when the 13-bit value is small enough (≤88), the encoder extracts a bit more data from each character pair without exceeding the 91² symbol space. The result is an average payload density of roughly 13.3 bits per two output characters — about 16% better overhead than Base64.

In practical terms: a 100-byte binary encoded with Base64 produces about 136 characters. With basE91 the same 100 bytes typically produces around 123 characters — roughly 10% shorter.

The alphabet and its exclusions

basE91 uses 91 printable ASCII characters. The characters excluded from the 95 printable ASCII set are the 4 most likely to cause problems in text contexts:

Single quote ' — breaks SQL and YAML string delimiters
Double quote " — breaks JSON and HTML attribute values
Backslash \ — escape character in virtually all string literals
The soft-hyphen / certain extended characters

The remaining 91 printable characters are safe in most text environments without additional escaping, making basE91 practical for embedding in JSON values, source code comments, and configuration files.

When to use basE91 vs Base64

Use basE91 when payload size matters (email attachments with strict size limits, embedded blobs in constrained bandwidth contexts) and you control both the encoder and decoder.
Use Base64 when you need universal compatibility. Base64 is supported natively in every major language’s standard library; basE91 is not.
Use basE91url (if you need URL safety) — there is no standard URL-safe basE91 variant, so for URLs stay with Base64url.

Remember that decoding yields raw bytes; this tool then interprets them as UTF-8, so feeding it a non-text binary will produce a decode error rather than readable text.