Base64 Decoder: Decode Base64 to Text Online

How to decode Base64

1. Paste the Base64 into the Input · Base64 box. Line breaks, spaces, surrounding quotes and a data:...;base64, prefix can stay; the decoder removes them.
2. The decoded text appears in Output · text. A status line reports Valid Base64, the variant (Standard or Base64URL), the number of decoded bytes and, under Looks like, the detected content type.
3. If the text is garbled, change Charset, or choose Auto-detect.
4. If the content is JSON, Pretty-print JSON indents it for reading.
5. Press Copy output, or Download to save the decoded bytes as a file. The extension follows the detected type, so JSON is saved as decoded.json and a PNG as decoded.png.

Swap sends decoded text back to the input and switches to encoding, which is a quick way to confirm the string round-trips.

What the decoder accepts

Base64 rarely arrives clean. It comes wrapped at 76 characters from an email, quoted from a JSON file, percent-encoded from a URL or without its padding from a token. The decoder accepts all of these without any setting:

The decoder is lenient about this packaging and strict about everything else: a character outside the alphabet stops it, with an error naming the character and its position. Some libraries silently skip bad characters and hand you corrupted output (see the code section below).

For URL-safe tokens specifically, the Base64URL decoder explains the alphabet and Base64 vs Base64URL the differences.

When the decoded Base64 is a file

Base64 carries any bytes, and many Base64 strings are images or documents. The tool checks the first bytes of the decoded data against known file signatures (magic numbers) and shows the result as Looks like: PNG, JPEG, PDF, ZIP, gzip, JSON, XML and many more. For binary formats it shows a printable preview, says the data is binary and links to the right tool, such as Base64 to image or Base64 to PDF. Download saves the original bytes either way.

The start of a Base64 string often gives the format away before you decode it:

More signatures are listed in common Base64 prefixes. For a file of unknown type, Base64 to file detects the format and previews it.

Charsets: fixing garbled text

Base64 stores bytes and says nothing about which charset turned characters into those bytes. The decoder assumes UTF-8, which is right for most modern data. When it is wrong, the symptoms are recognizable:

• � characters and the note “Some bytes are not valid UTF-8”. The text came from a legacy charset. Y2Fm6Q== reads as caf� in UTF-8 and as café in ISO-8859-1 or Windows-1252. Russian text such as z/Do4uXy needs Windows-1251 to show Привет.
• A dot or gap after every letter. That is UTF-16, where ASCII letters are followed by a zero byte. SABlAGwAbABvAA== is Hello in UTF-16LE (Base64 to UTF-16).
• é, ü or ’ in the output. The decoding is correct; the text was double-encoded before it was turned into Base64, because some earlier step misread UTF-8 as Latin-1 and stored the result as UTF-8 again. Y2Fmw4PCqQ== decodes to exactly café. Fix it at the source.

Auto-detect checks for a byte order mark, then whether the bytes are valid UTF-8, then for the zero-byte pattern of UTF-16. Single-byte code pages such as Windows-1251 and KOI8-R look alike to any detector, so pick those by hand when you know the source.

Error messages and what they mean

The Base64 errors guide goes through each case in more depth. To find the exact character or padding problem in a string, use the Base64 validator.

Base64 decode in JavaScript, Python and the command line

atob() returns a “binary string” with one character per byte, so UTF-8 text needs a TextDecoder step. atob accepts missing padding but throws InvalidCharacterError on URL-safe - and _.

const s = "R3LDvMOfZSDwn5GL";
new TextDecoder().decode(Uint8Array.from(atob(s), c => c.charCodeAt(0)));
// "Grüße 👋"

// Node.js accepts both alphabets
Buffer.from(s, "base64").toString("utf8");          // "Grüße 👋"
Buffer.from("8J-Riw", "base64url").toString("utf8"); // "👋"

Python raises binascii.Error: Incorrect padding on unpadded input, so add the padding yourself:

import base64

s = "SGVsbG8"
base64.b64decode(s + "=" * (-len(s) % 4)).decode("utf-8")
# 'Hello'
base64.b64decode("SGVs*bG8=", validate=True)
# binascii.Error: Only base64 data is allowed

Without validate=True, Python discards the * and returns b'Hello'. Node’s Buffer.from behaves the same way, so a damaged string can decode without any error.

On the command line:

printf '%s' 'R3LDvMOfZSDwn5GL' | base64 -d
base64 -d message.b64 > message.bin

The first line prints Grüße 👋. Older macOS versions spell the flag -D. More detail per language: JavaScript, Node.js, Python, PHP.

FAQ