About

Base64 encoding inflates payload size by approximately 33% and obscures the underlying data structure. When the encoded payload is tabular data (CSV, TSV), a decoding error or charset mismatch silently corrupts field values. Truncated padding characters (=) cause NULL byte injection. Misidentified delimiters collapse multiple columns into one. This tool decodes Base64 input using the native atob function with full UTF-8 reconstruction via TextDecoder, then parses the result as structured CSV with automatic delimiter detection by frequency analysis. It handles RFC 4648 standard and URL-safe alphabets. Note: this tool assumes the encoded content is plaintext CSV. Binary formats (XLSX, ODS) embedded in Base64 require a different decoder.

Formulas

Base64 encoding maps every 3 input bytes (24 bits) to 4 printable ASCII characters (6 bits each). The decoded byte length is computed as:

L_decoded = 34 × L_encoded − P

where L_encoded = length of the Base64 string (excluding whitespace), and P = number of padding = characters (0, 1, or 2).

Delimiter auto-detection uses a frequency consistency score. For each candidate delimiter d, the tool counts occurrences per line and computes variance:

S(d) = 1n n∑i=1 (c_i − c)²

where c_i = count of delimiter d on line i, and c = mean count across all lines. The delimiter with the lowest non-zero variance and highest mean count is selected. A variance of 0 with c ≥ 1 indicates a perfectly consistent structure.

URL-safe Base64 substitution rule: + → - and / → _. The tool normalizes URL-safe input before decoding.

Reference Data

Base64 Character	Index	Binary	Notes
A - Z	0 - 25	000000 - 011001	Uppercase Latin
a - z	26 - 51	011010 - 110011	Lowercase Latin
0 - 9	52 - 61	110100 - 111101	Digits
+	62	111110	Standard alphabet
/	63	111111	Standard alphabet
-	62	111110	URL-safe variant
_	63	111111	URL-safe variant
=	-	-	Padding (mod 4 alignment)
Delimiter	Name	Common Use	Detection Priority
,	Comma	RFC 4180 CSV standard	1 (highest)
;	Semicolon	European locale CSV exports	2
\t	Tab	TSV files, database exports	3
\|	Pipe	Legacy systems, mainframes	4
Input Size	Encoded (Base64)	Decoded (Bytes)	Overhead
Small	1 KB	768 B	33.3%
Medium	100 KB	75 KB	33.3%
Large	1 MB	768 KB	33.3%
Max recommended	10 MB	7.5 MB	33.3%
Padding	Input Length mod 3	Pad Characters	Example
No padding	0	0	QUJD
Single pad	2	1	QUI=
Double pad	1	2	QQ==

Frequently Asked Questions

RFC 4648 requires padding with = to make the encoded length a multiple of 4. However, many systems strip padding. This tool auto-restores missing padding by calculating L mod 4 and appending the required 0, 1, or 2 pad characters before decoding. No manual correction needed.

The tool decodes using TextDecoder with the fatal flag set to FALSE. Malformed byte sequences are replaced with the Unicode replacement character U+FFFD (◊). If the output contains more than 5% replacement characters, the tool warns that the source data is likely binary (e.g., XLSX, images) rather than plaintext CSV.

The detector samples the first 20 non-empty lines and computes variance for each candidate delimiter (, ; \t |). A delimiter with zero variance and a mean occurrence ≥ 1 is ideal. If all candidates show high variance, the tool defaults to comma per RFC 4180 and flags a warning about inconsistent structure.

Yes. URL-safe Base64 replaces + with - and / with _ to avoid issues in URLs and filenames. The tool detects the presence of - or _ characters and automatically converts them back to + and / before decoding.

The tool accepts up to 10 MB of Base64 input, which decodes to approximately 7.5 MB of CSV. The table preview is limited to the first 100 rows for rendering performance. The full decoded CSV is available for download regardless of size. Processing time scales linearly: 1 MB typically decodes in under 200 ms.

The CSV parser implements RFC 4180 compliant tokenization. Fields enclosed in double quotes (") preserve embedded delimiters, newlines, and literal quote characters (escaped as ""). For example, the field "Smith, John" is parsed as a single column value Smith, John rather than being split.