Base64 encoder & decoder

How to use it

1. Paste text to encode it, or paste Base64 and switch to Decode.
2. For a file, pick it and copy the Base64 or the ready-made data: URI.
3. Hit Copy to put the result on your clipboard.

The 33% tax, and when it is worth paying

Base64 maps every three bytes onto four ASCII characters, so output is four-thirds the size of input — 33% before padding or line breaks. RFC 4648 codifies an encoding that exists because much infrastructure was built when only 7-bit ASCII survived transit: SMTP in particular, and gateways that would strip a high bit or reinterpret a stray 0x0D.

That overhead makes inlining a data URI a trade, not a free win. A 6 KB icon becomes about 8 KB inside a stylesheet that can no longer be cached separately, cannot be lazy-loaded, and is re-downloaded whenever any other rule in the file changes. Below a kilobyte or two the saved request usually wins; above it, rarely. Compression recovers part of the difference, never all of it.

Two alphabets, and the padding that goes missing

RFC 4648 defines two alphabets. Standard Base64 ends with plus and slash, both meaningful inside a URL, so a token encoded that way arrives truncated at a query boundary or mangled. base64url substitutes hyphen and underscore, and is what JWTs and WebAuthn payloads use. The 62 alphanumeric characters are identical, so a string containing neither final symbol decodes the same under both — which is how this bug clears a test suite.

Padding is the other half. The equals sign pads output to a multiple of four and carries no data; a strict standard decoder rejects a wrong-length string, while base64url conventionally omits it and JWT segments never carry it, so a decoder insisting on padding refuses a valid token. Input is accepted here either way and in either alphabet; if a string still fails, suspect a line break from a terminal.

Questions