Repair Base64

How to repair a Base64 string

1. Paste the broken value into Base64 to check. Leave the surrounding quotes, prefixes and line breaks in; the tool needs them to understand what happened.
2. Choose the Output format: standard padded, standard without padding, URL-safe with or without padding, MIME (76-character lines, CRLF) or PEM (64-character lines).
3. Read What was fixed. Each repair is listed separately, so you can tell harmless cleanup from changes that might have cost data.
4. Copy or download the result, press Validate the result to run a full check on it, or Download decoded file to save the bytes.

What the repair fixes

It also removes data:…;base64, prefixes and PEM -----BEGIN …----- lines, unwraps plain quotes, turns literal \n and \/ back into characters (PHP’s json_encode writes \/ by default), strips whitespace and line breaks, unifies mixed +/ and -_ alphabets, and corrects the number of =.

For concatenated values, each piece is decoded separately and the bytes are joined, which is correct when a program appended padded chunks one after another. Deleting the inner = instead would shift every following bit and produce garbage.

Repairs that may hide a real problem

Two fixes are reported as warnings because they guess:

• Illegal characters removed. If the string contained *, # or a stray Unicode character, the tool deletes it and lists which characters went. A character in the middle of the data might have been a real symbol that got mangled, in which case the decoded bytes will still be wrong from that point on.
• A dangling character dropped. A length with remainder 1 cannot be produced by any encoder, so one character is extra or the string was cut off. The tool drops the last character so the rest decodes, but if the value was truncated, the end of the file is gone. A truncated image usually shows a gray or missing bottom; a truncated ZIP will not open.

When either warning appears, try to get the original again: an untruncated log line, a full API response, or the attachment straight from the mail server.

Why a naive fix is not enough

The common snippet for “fixing” Base64 handles whitespace, URL-safe characters and padding:

def fix_base64(s):
    s = "".join(s.split())
    s = s.replace("-", "+").replace("_", "/").rstrip("=")
    return s + "=" * (-len(s) % 4)

It repairs SGVsbG8gd29ybGQ and PDw_Pz8-Pg correctly. Feed it PDw/Pz8 Pg==, where a + turned into a space, and it removes the space, leaving 9 data characters. Python then fails with “number of data characters (9) cannot be 1 more than a multiple of 4”. JSON escapes, %3D, glued chunks and prefixes slip through the same way. The repair here tries each fix in turn and only applies the space-to-plus change when the result has a valid length.

The safer long-term fix is at the source: percent-encode Base64 before putting it in a URL, or use Base64URL, and keep it out of places that rewrap or escape text. If the repaired string still decodes to something unexpected, the Base64 errors guide covers garbled text and double encoding, and the Base64 decoder shows the bytes.

FAQ