Base64 Encode

Before encoding, the tool turns your text into UTF-8 bytes. Plain ASCII letters take one byte each, but accented letters, symbols and other scripts take two to four, and the Base64 output grows to match:

Any program that decodes this output has to read the bytes as UTF-8 as well. If it assumes a single-byte encoding such as Latin-1, café comes back as café. Emoji work too, since they are just four UTF-8 bytes.

Nothing is trimmed. hi encodes to aGk=, but hi followed by a space encodes to aGkg. A line break is encoded as a single newline byte, even if you pasted text with Windows line endings, because the text box stores line breaks that way.

This matters when you encode a password, an API key or a user:password pair for a header. A space or newline picked up while copying becomes part of the encoded value, and the server will reject it. user:pass should encode to exactly dXNlcjpwYXNz.

The output uses the standard Base64 alphabet, with + and /, and adds = padding whenever the byte count isn't a multiple of three. It comes out as a single line, with no line breaks every 76 characters.

If the value is going into a URL or a token that expects the URL-safe variant, convert it afterwards: replace + with -, / with _, and remove the trailing =. For example, >>>??? encodes to Pj4+Pz8/, which in URL-safe form is Pj4-Pz8_. If you need to put standard Base64 into a query string instead, run it through the URL Encoder so that +, / and = become %2B, %2F and %3D.

• Encoding twice. If your input already looks like Base64, encoding it again gives a longer string that has to be decoded twice. aGk= encoded again is YUdrPQ==.
• Pasting a file's contents. The box takes text. For an image, use the Image to Base64 converter, which reads the file's actual bytes.
• Expecting the result to be secret. Anyone can decode it. Base64 changes the format, not who can read it.