Base64 Encoder / Decoder

How Base64 encoding works

Base64 exists to move binary data through channels that only tolerate text: JSON fields, XML documents, email bodies, HTTP headers and data URLs. The algorithm takes the raw bytes, groups them into sets of three, and rewrites each 24-bit group as four 6-bit numbers. A 6-bit number has 64 possible values, so each one maps to a single printable character — the 62 letters and digits A–Z, a–z and 0–9, plus + and /.

That three-bytes-to-four-characters ratio is the whole trick, and it is also the whole cost: the encoded text is roughly 33% larger than the source. If the input length is not a multiple of three, the final group is short and is padded with = characters so the length always lands on a multiple of four. Those = signs carry no data — they are a checksum-free way of telling a decoder how many bytes the last group really contains.

The subtle part in a browser is text. JavaScript strings are sequences of UTF-16 code units, and the legacy btoa() function rejects anything above U+00FF with an InvalidCharacterError. So btoa("café") and any string containing an emoji both throw. The correct route, and the one this tool takes, is to turn the text into UTF-8 bytes first with new TextEncoder().encode(text), Base64-encode those bytes, and on the way back decode to bytes and run them through new TextDecoder().decode(bytes). That round-trip is lossless for every character the Unicode standard defines, accents and emoji included.

Accuracy, limits and honest caveats

Base64 is a bijection over byte strings, so for well-formed input the round-trip here is exact. The honest limitations are about ill-formed input, and about what Base64 is not for.

• Invalid input is reported, never guessed. If the string contains characters outside the Base64 alphabet, or its length is impossible once padding is considered, the tool says so instead of returning empty output. It deliberately does not invent characters to force a decode.
• Padding and whitespace are tolerated. Missing = at the end is fine — it is re-added automatically, because the length already tells the decoder what is missing. Line breaks from MIME-encoded data are stripped too, which matters for Base64 copied out of email.
• Bytes are not always text. Decoding a Base64 string that really contains an image or a compressed archive produces replacement characters, because those bytes were never valid UTF-8. Nothing is lost — the file input pane is the right way to look at binary data — but the text pane will look like noise.
• Base64 is not compression and not security. It adds about a third to the size, and it protects nothing: it is a public, keyless transform. Treating Base64 as obfuscation is a common and costly mistake.
• Large files stay in memory. Everything happens on the main thread, so a multi-megabyte file will pause the page briefly while the string is built. For very large payloads a chunked streaming encoder is a better fit than a single-page tool.

How to use this tool

1. Encode text. Type or paste into the left pane — the output updates as you type, so there is no need to press anything. Accents and emoji are handled through UTF-8 and never throw.
2. Decode a string. Paste the Base64 into the same left pane and press Decode. Both alphabets are accepted: a value containing - or _ is read as URL-safe automatically, and missing padding is repaired.
3. Switch to the URL-safe alphabet. Tick the box before encoding when the result will travel inside a URL, a filename or a JWT-style token. The encoded string updates immediately, with + and / replaced and the padding removed.
4. Encode a file. Drop a file onto the dashed area, or press Enter there to pick one. It is read as an ArrayBuffer and converted to Base64, and the readouts show the original byte size next to the encoded character count.
5. Preview images. When the file is an image, a data-URL preview appears so you can confirm you encoded the file you meant. The browser renders the preview itself; nothing is uploaded.
6. Take the result away. Copy result puts the string on the clipboard and Download result as .txt saves it to a file, ready to paste into a config value, an API payload or an img tag.

Because the page is a single static document with no network calls, it keeps working on a plane, on a locked-down machine, or anywhere an online encoder would be a bad idea for the data you are handling.