The Oracle documentation for java.util.Base64 refers to three different kinds of Base64, but only by reference to RFCs. Here it is in one place.

RFC 4648 Table 1: Normal base64

Both RFC 2045 “Multipurpose Internet Mail Extensions” (November 1996) and RFC 4648 Table 1 “The Base16, Base32, and Base64 Data Encodings” (October 2006) define this same alphabet:

ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789+/ padded with =

This is the alphabet used by Java’s Base64.getEncoder(), and also by Base64.getMimeEncoder(). The only difference is that getMimeEncoder also inserts \r\n every 76 characters, and ignores non-base64 characters (including any \r and \n) during decoding.

In Java, new String(Base64.getEncoder().encode("Hello??".getBytes())) should give you "SGVsbG8/Pw==".

In Kotlin, it’s String(Base64.getEncoder().encode("Hello??".toByteArray())).

RFC 4648 Table 2: URL-safe base64

RFC 4648 Table 2 “The Base16, Base32, and Base64 Data Encodings” (October 2006) defines this “URL- and filename-safe” alphabet:

ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789-_ padded with =

This is the alphabet used by Java’s Base64.getUrlEncoder().

In Java, new String(Base64.getUrlEncoder().encode("Hello??".getBytes())) should give you "SGVsbG8_Pw==".

In Kotlin, it’s String(Base64.getUrlEncoder().encode("Hello??".toByteArray())).