Base64 to PDF, without sending the document anywhere.

JSON cannot carry raw bytes, so APIs that return documents — e-signature, invoicing, payroll, shipping label and reporting services — usually put the PDF in a string field as Base64. Email does the same: a PDF attachment travels as Base64 inside the MIME source, wrapped at 76 characters per line. Both decode here exactly as they are; you do not need to remove line breaks first.

Every PDF begins with %PDF-, and in Base64 that always becomes JVBERi0. If your string starts with those seven characters, it is a PDF. Other prefixes tell you what you have instead:

UEsDB means a ZIP container, which is also what Word and Excel files are. H4sI means the document was gzip-compressed before it was encoded — download the result as a .gz file and decompress it to reach the PDF.

• Copied from raw JSON. JSON may escape the slash as \/. Copy the value from a parsed view (or replace \/ with /) — a backslash is not valid Base64.
• Truncated. A complete PDF ends with %%EOF. If the Hex tab shows the document stopping short, the Base64 was cut off; the PDF reader will report it as damaged.
• Encoded twice. If the result starts with JVBERi0 as text, it was Base64 encoded twice. Press Send to input and decode again.
• Password protected. Decoding restores the file exactly; if it was encrypted, the PDF reader will still ask for its password.

# Python
import base64
with open("document.pdf", "wb") as f:
    f.write(base64.b64decode(response["pdf"]))

// Node.js
fs.writeFileSync("document.pdf", Buffer.from(json.pdf, "base64"));

# Shell, straight from an API response
curl -s https://api.example.com/invoice/42 | jq -r .pdf | base64 --decode > invoice.pdf

See the full set of Base64 decode examples for Java, C#, PHP, Go and PowerShell.