Base64 Encode / Decode

Which alphabet btoa writes

btoa walks the JavaScript string one code unit at a time and emits the standard Base64 alphabet. Digits, letters, plus, and slash appear in the body. Padding is zero, one, or two equals signs so the length is a multiple of four. hello world is eleven bytes, so the result is aGVsbG8gd29ybGQ= with one padding character. An empty field encodes to an empty string and still counts as success. The button does not insert line breaks at 76 characters, and it does not add a data: prefix or a MIME type.

Decode is atob on the field as typed. The browser's forgiving decoder ignores ASCII whitespace, so aGVs bG8= and a broken line inside the blob still return hello. Missing padding can also succeed: aGVsbG8gd29ybGQ without the equals sign returns hello world. Characters outside the standard alphabet fail. Underscore is outside that alphabet, which is why a URL-safe blob is not silently repaired here.

The two mode buttons

The group is labeled Base64 mode and contains Encode and Decode. Encode is pressed when the page opens. The buttons only change which function Run will call and which labels you see. They do not clear the field and they do not clear a previous result. If you decode, then press Encode, the old decoded text stays in the result until you press Run again.

Copy output reads the string last published by a successful or failed Run. Before the first Run, the shell has nothing to copy and reports that no output exists yet. After a failure, the copied text is the error sentence, because that sentence is written into the result. There is no second button for download and no filename.

Where standard Base64 shows up

HTTP Basic credentials are the standard alphabet of the ASCII string user:password, including padding. You can encode that string here or decode a copied Authorization value after you strip the Basic prefix yourself. The page does not add or remove the word Basic. A query string is a different encoding: percent-escapes belong on the URL encode/decode page, where a space becomes %20. Bytes shown as spaced hexadecimal pairs belong on Hex encode/decode.

A JWT segment is Base64URL. Minus and underscore are not rewritten on this page. Use the JWT decoder when the job is to read header and payload JSON. That decoder calls atob only after it maps the URL-safe characters and only on the first two segments. The guide Turn a browser request into Python or JavaScript is the place to look when the Base64 sits inside a captured request you are about to turn into code.

Latin-1 limits and the shared error sentence

btoa accepts code points 0 through 255 and throws InvalidCharacterError above that. é is U+00E9, so it encodes (héllo becomes aOlsbG8=) and atob returns é again. 你, a check mark, and an emoji are above 255, so Run fails. The catch block does not show InvalidCharacterError. Encode and decode failures both replace the result with Invalid Base64 input. Please check the input and try again. The status line repeats that sentence.

That means a UTF-8 encode of 你 is not available here, and a failed encode looks like a failed decode. If the text you care about is not Latin-1, this page is the wrong tool. It also will not decode a hex digest, and it will not guess whether a string is Base64 or a hash. Paste one value, choose the mode, and press Run once per attempt.

Rules and limitations

Encode calls btoa. Decode calls atob. The alphabet is standard Base64 with plus, slash, and optional padding. Whitespace is ignored on decode. URL-safe minus and underscore are not mapped.

Code points above 255 fail to encode. Both failures surface as the same sentence. Mode changes do not rerun. There is no download control and no data-URI wrapper.

Common problems

If a JWT segment fails, it likely still contains minus or underscore. Decode it on the JWT page. This page will not translate that alphabet.

If a non-Latin character fails, the status text still says Invalid Base64 input even though you were encoding. The character is outside the 0–255 range btoa accepts.

Frequently asked questions

About this tool

Maintained by Toolbench. For large or deeply nested files, split the input into smaller batches and verify representative results before using them in an import or application. Report reproducible problems through Contact.