Base64 to UTF-8: Decode Base64 to Unicode Text

How to decode Base64 to UTF-8

1. Paste the Base64 into Input · Base64. Charset is preset to UTF-8.
2. Read the result in Output · text. The status line shows the byte count and the detected content type.
3. If the status line says “Some bytes are not valid UTF-8”, the text is in another charset. Try Auto-detect, Windows-1252 or UTF-16LE.
4. Copy output copies the text; Download saves the decoded bytes unchanged.

To encode in the other direction, use UTF-8 to Base64.

Examples

The last three rows are the problems covered below: a wrong charset, a cut-off string and text that was garbled before encoding.

Invalid UTF-8 and the replacement character

UTF-8 has strict rules: a byte above 7F must be part of a 2-, 3- or 4-byte sequence with a fixed pattern. Bytes that break the pattern cannot be read as UTF-8. The decoder replaces each broken sequence with U+FFFD (�) and adds the note “Some bytes are not valid UTF-8. Try another charset.”. Nothing is dropped silently, so each � marks the exact spot of a problem.

Two causes cover nearly every case:

• A different charset. Y2Fm6Q== holds 63 61 66 E9, which is café in ISO-8859-1 and Windows-1252. Switch Charset and the text appears. The Base64 decoder page lists the common symptoms per charset.
• A truncated string. If Base64 was cut, for example by a column limit in a database or a log line, the last character may lose some of its bytes. 8J+YgA== is a complete emoji; 8J+Y is the same emoji missing its last byte and decodes to a single �. Look for the full value at the source.

Double-encoded text (é and ’)

When output shows é for é, ü for ü or ’ for a curly apostrophe, the decoder is not at fault. Somewhere earlier, UTF-8 bytes were read as Windows-1252 and saved again as UTF-8, turning each special character into two or three wrong ones. The Base64 faithfully holds that damaged text: Y2Fmw4PCqQ== is valid UTF-8 for the seven-byte string café.

You can often repair such a string in code by reversing the mistake: encode the text as Windows-1252 and decode the bytes as UTF-8. The real fix is in the system that wrote it.

Byte order marks

Some editors, especially on Windows, start UTF-8 files with the bytes EF BB BF, a byte order mark. In Base64 the file then starts with 77u/. The decoder recognizes the mark and leaves it out of the text, while Download keeps the bytes exactly as they were. In code you may need to remove it yourself:

import base64

raw = base64.b64decode("77u/SGk=")
raw.decode("utf-8")       # '\ufeffHi'
raw.decode("utf-8-sig")   # 'Hi'

base64.b64decode("Y2Fm6Q==").decode("utf-8", errors="replace")   # 'caf�'

Without errors="replace", Python raises UnicodeDecodeError on invalid bytes. In JavaScript, new TextDecoder("utf-8", { fatal: true }) throws a TypeError instead of inserting �, and the default decoder strips the BOM like this tool does. More in Base64 in Python and Base64 in JavaScript.

FAQ