Base64 has more dialects than people expect, and a strict decoder turns every one of them into a failure. Three kept breaking real payloads: URL-safe alphabets (JWTs use - and _ instead of + and /), MIME-style line breaks (76-character wrapping, as produced by most mail and certificate tooling), and omitted padding (the trailing = characters, which JWTs also drop).

What changed

• All three dialects decode — URL-safe characters are mapped back, whitespace is stripped before decoding, and missing padding is restored automatically.
• Precise errors — when input is genuinely invalid, the error names the exact offending character and its code point (“unexpected character * (code 42)”) instead of a generic failure.
• The JSON check — decoded output is validated, and the badge tells you whether you recovered a document or plain text.

How to use it

Paste anything base64-ish into the Base64 to JSON decoder — a JWT segment, a wrapped MIME block, a padded or unpadded string — and the JSON appears on the right when the payload is one. Nothing is uploaded; the decode happens in your tab like everything else here.

Forgiving where it counts, precise where it matters

Every tolerance on this list corresponds to a real producer: JWTs (RFC 7515) omit padding, MIME (RFC 2045) wraps lines, URL contexts swap the alphabet. The decoder accepts all three — and when input is genuinely wrong, it names the exact offending character, so the rare real mistake is as quick to fix as the common dialects are to accept.