Base64 Encoder

Base64 carries arbitrary bytes through channels that only accept text — a mail body, a JSON string, a URL, an HTTP header. Three bytes become four characters drawn from a 64-character alphabet, which is where the name and the roughly 33% size increase both come from.

Which alphabet

Standard Base64 uses + and /. Both mean something else in a URL, so Base64URL substitutes - and _ and the value survives unescaped in a path, a query or a filename. JWTs use Base64URL for exactly this reason.

Padding

Trailing = rounds the output to a multiple of four. It carries no information, which is why Base64URL omits it by specification and why removing it is lossless — though some strict parsers still want it.

Non-ASCII

Text is encoded to UTF-8 bytes before those bytes are Base64-encoded. That is what every standard expects, and it is why an accented character or an emoji round-trips through the Base64 Decoder unchanged rather than arriving mangled.

abc encodes to YWJj — three bytes, four characters, no padding needed. Drop a character and ab becomes YWI=; drop another and a becomes YQ==. The padding count is a direct readout of how the input length divided by three.