Base64 to File Converter

How to decode Base64 to a file

1. Paste the string into Base64 or data URI. Standard and URL-safe Base64 are both accepted, with or without padding, line breaks or quotes. A whole data URI works, and so does text that contains one: a CSS url("data:…"), an HTML src="data:…" or a Markdown image. For a long string in a file, use Open .txt file or drop it on the box.
2. Read the status line under the input: whether the input is valid, which variant it is (Standard or Base64URL), the decoded size and the Detected type. A data URI also shows its Declared type.
3. Look at the preview and the three fields below it. File name, Extension and MIME type are filled in from the detection. Anything you type there stays until you press Clear.
4. Press Download file, or Copy as data URI to get the bytes back as data:<MIME type>;base64,….

Sample loads a different test file on each click: PNG, PDF, ZIP, WAV, DOCX, JSON, gzip and SVG. If you drop a file that is already binary, say a PDF instead of a text file holding its Base64, the tool tells you so and links to File to Base64, which goes the other way.

Why the file shows up as application/octet-stream

application/octet-stream means “some bytes, type unknown”. It ends up in front of Base64 for ordinary reasons. Chrome’s FileReader.readAsDataURL() produces data:application/octet-stream;base64, for any file whose type the browser does not know. Storage and export systems that keep the bytes but not the type label everything that way. Some APIs declare every attachment as octet-stream. A viewer that relies on the label then refuses to show the file, and a download gets no useful extension.

The bytes themselves usually identify the format: most formats begin with a fixed signature, and the tool checks more than 60 of them. A few signatures are shared, so the tool looks further into the file:

The full list of signatures is on common Base64 prefixes.

When a data URI declares a type and the content disagrees, a note says so, for example “The data URI says application/octet-stream, but the content looks like PDF.”, and the detected type is used. The declared type wins only when detection is weak: when the bytes look like plain text or unknown binary.

Base64 inside Base64, URL encoding and gzip

Data that went through several systems is often wrapped more than once. After decoding, the tool checks whether the result is another layer and offers Decode again:

Decode again puts the inner layer into the input and decodes it; Back steps out again and remembers up to 20 steps. The gzip and zlib notes appear only after the tool has actually unpacked the data, for results up to 16 MB. The gzip preview also shows the original file name and modification time stored in the header, with Save unpacked for the content. For text payloads, the gzip Base64 decoder is the more direct tool.

Saving with the right extension

Windows and macOS pick the program by extension, so a PDF saved as decoded.bin looks broken even though every byte is right. The Extension and MIME type fields follow the detection until you type in them:

• The file name comes from a name= or filename= parameter in a data URI, or from the name of the text file you opened, without its .txt or .b64 ending: report.pdf.b64.txt is saved as report.pdf.
• For an OLE file you know is a spreadsheet, type xls in Extension; the tool cannot tell DOC, XLS and MSG apart.
• A MIME type you enter also re-renders the preview as that kind. If the tool says Binary but you expect audio, typing audio/ogg makes it try the audio player.
• The MIME type affects only the download and Copy as data URI; a file on disk has no MIME type, only its bytes and its name.

What the preview shows

Every result lists type, MIME type, extension, size and a SHA-256 hash, with Show hex dump for the first 512 bytes. The hex dump is open from the start for unknown binary data. Depending on the kind:

Detect the MIME type of Base64 in code

Twenty-four characters of Base64 decode to 18 bytes, enough for the common signatures. This works in browsers and in Node.js 16 and later:

function sniffMime(b64) {
  const head = b64.replace(/^data:[^,]*,/, "").replace(/\s+/g, "").slice(0, 24)
    .replace(/-/g, "+").replace(/_/g, "/");
  const b = Uint8Array.from(atob(head), (c) => c.charCodeAt(0));
  const str = (start, len) => String.fromCharCode(...b.subarray(start, start + len));

  if (str(0, 5) === "%PDF-") return "application/pdf";
  if (str(0, 8) === "\x89PNG\r\n\x1a\n") return "image/png";
  if (b[0] === 0xff && b[1] === 0xd8 && b[2] === 0xff) return "image/jpeg";
  if (str(0, 4) === "GIF8") return "image/gif";
  if (str(0, 4) === "RIFF" && str(8, 4) === "WEBP") return "image/webp";
  if (str(0, 4) === "RIFF" && str(8, 4) === "WAVE") return "audio/wav";
  if (str(4, 4) === "ftyp") return "video/mp4"; // also M4A, MOV, HEIC: see bytes 8-11
  if (str(0, 4) === "PK\x03\x04") return "application/zip"; // also DOCX, XLSX, PPTX
  if (b[0] === 0x1f && b[1] === 0x8b) return "application/gzip";
  if (str(0, 3) === "ID3" || (b[0] === 0xff && (b[1] & 0xe6) === 0xe2)) return "audio/mpeg";
  return "application/octet-stream";
}

sniffMime("data:application/octet-stream;base64,JVBERi0xLjcKJeLjz9MK"); // "application/pdf"

On the command line, file reads the decoded bytes from standard input:

base64 -d payload.b64 | file -b --mime-type -

That prints application/pdf for a PDF. If you decode with OpenSSL instead, add -A for a string on a single line. Without it, openssl base64 -d silently writes nothing when the line has no final newline or is too long (1,024 characters or more in OpenSSL 3.5). More on Base64 in JavaScript and Base64 in Bash.

FAQ