Base64 Encoder and Decoder

How to encode or decode

1. Choose Encode or Decode. Encoding accepts literal text or exact file bytes; decoding accepts pasted Base64 only.
2. Choose standard or URL-safe alphabet. Standard output always uses padding; URL-safe padding is optional and defaults off.
3. Convert explicitly after changing settings. Review byte counts and the selectable preview, then copy or download complete output. Clear offers Undo; Cancel discards the current job.

Worked example: f, fo and foo

Standard Base64 encodes f as Zg==, fo as Zm8= and foo as Zm9v. UTF-8 Swedish text and emoji round-trip correctly. Bytes FB FF encode as +/8= in standard mode and -_8 in unpadded URL-safe mode.

Strict validation rules

Standard accepts A–Z, a–z, 0–9, + and /; URL-safe substitutes - and _. The other alphabet is rejected rather than guessed. Standard requires canonical padding for partial groups; URL-safe permits correct padded or unpadded input. Internal or excess padding, length modulo four equal to one, invalid characters and nonzero unused trailing bits are rejected. Zh== is invalid; Zg== is canonical.

Whitespace is invalid unless you enable ignoring ASCII spaces, tabs, CR and LF. The removed count is shown. Unicode whitespace is still invalid; input is never trimmed or percent-decoded. Data URLs are rejected: paste only their encoded portion after the comma.

Text versus binary

Text encodes as UTF-8, with unpaired UTF-16 surrogates rejected rather than replaced. A deliberate leading U+FEFF remains in text and file mode preserves every byte. Decoded UTF-8 uses strict validation and preserves a leading BOM. Invalid UTF-8 is not invalid Base64: bytes remain available as decoded.bin with application/octet-stream, never opened or executed. Valid UTF-8 can additionally download as decoded.txt. Encoded output downloads as encoded.txt.

Limits and privacy

Maximum 5 MiB UTF-8 text or file input, 8 MiB encoded decode input and 5 MiB decoded output. Predicted decoded size is checked before allocation. A dedicated worker handles conversion; changing tabs, settings or inputs invalidates output without automatically reinterpreting it. Preview is limited to 20,000 characters; downloads always include all bytes. Copy is limited to 2 MiB of text characters; a denied clipboard leaves selectable output and download guidance.

Base64 is reversible encoding, not encryption and not a way to secure passwords or documents. No compression, remote URLs, image preview, data-URL import, content-type guessing, HTML interpretation or automatic file execution. Inputs and filenames are never placed in URLs, telemetry or browser storage by this tool.