Repair Corrupt PDF

Under the ISO 32000 specification, a PDF is not read sequentially from start to finish like a text file. Instead, conforming PDF viewers start by reading the very end of the file, locating the startxref token, and reading the cross-reference (XRef) table, which maps every indirect object number to its exact physical byte offset from the start of the file.

Most corruption incidents stem from failures in this indexing system:

1. Truncated Files: If a download terminates at 95% completion, the file trailer (%%EOF) and XRef table at the end of the byte stream are completely missing. Standard viewers fail immediately because they cannot locate the root Catalog.

2. Byte Offset Drift: If a PDF is transmitted via an email server or FTP client that incorrectly treats binary data as ASCII text, line endings can be converted from UNIX (LF) to Windows (CRLF). Inserting an extra byte (0x0D) at every newline shifts all subsequent byte positions, causing every single pointer in the XRef table to point to invalid data.

3. Malformed Trailers: Software crashes during saving often leave the trailer dictionary unclosed or leave circular object references.

When the cross-reference table is corrupted or missing, a professional repair engine must abandon the XRef table entirely and perform linear heuristic byte scanning across the entire file payload.

Our client-side recovery engine scans the raw ArrayBuffer for the regular expression pattern '\d+\s+\d+\s+obj', identifying the physical byte offset of every indirect object header. It reads the object number and generation number, and parses the object dictionary or stream up to the matching 'endobj' token.

Through this linear scan, the engine reconstructs an in-memory inventory of every surviving dictionary, page, image XObject, font, and content stream, completely restoring access to the document's contents regardless of how badly the original XRef table was damaged.

Once all raw objects are inventoried in memory, the repair engine must establish the logical document hierarchy. It searches the object pool for a dictionary containing '/Type /Catalog' (or a dictionary containing '/Pages').

If the original Catalog object is damaged or truncated, our engine synthesizes a brand-new Catalog dictionary. It traverses all objects with '/Type /Page', resolves their respective /MediaBox and /Contents references, and constructs a balanced /Pages tree.

If certain page dictionaries are missing their /Parent references or have orphaned content streams, the engine re-attaches them into a linear page sequence, restoring as many readable pages as physically possible from the surviving data.

Another frequent cause of fatal parsing errors is incorrect stream length declarations. Every stream object in PDF specifies its byte size via the /Length key. If offset drift or transmission truncation alters the stream length, viewers either stop reading prematurely or throw a stream decompression error.

Our repair engine inspects every stream object between 'stream' and 'endstream' markers. It recalculates the exact physical byte length, tests the stream against FlateDecode (Zlib) decompression, fixes malformed stream boundaries, and updates the /Length key accordingly. Corrupted vector graphics and text streams are recovered cleanly.

The final phase of client-side PDF repair is re-serialization. The engine assembles all recovered objects into a clean binary sequence, calculates exact byte offsets for each object from byte zero, and generates a brand-new, perfectly indexed cross-reference table (XRef).

It appends a standard trailer dictionary pointing to the verified Catalog object, writes the 'startxref' offset pointer, and terminates the file with the official '%%EOF' marker. The resulting document is a fully compliant, healthy ISO 32000 PDF file that opens smoothly in Adobe Acrobat, Chrome, Apple Preview, and mobile viewers with zero error prompts.

Corrupted documents frequently contain an organization's most critical un-backed-up records: confidential legal evidence, tax returns, medical scans, or corporate financial models. Uploading corrupted files to commercial third-party repair services exposes privileged communications and trade secrets to unauthorized server retention and potential data breaches.

2run.tools guarantees absolute client-side privacy. The entire byte scanning, object recovery, and XRef reconstruction pipeline executes 100% locally inside your browser's WebAssembly memory sandbox. Your document never leaves your machine, ensuring complete enterprise confidentiality and full compliance with GDPR, HIPAA, and legal privilege rules.

In worst-case scenarios where a storage drive suffers severe hardware failure or a file transfer is cut off midway, critical portions of the PDF structure may be irrevocably missing. In such cases, standard document reconstruction cannot restore the complete page hierarchy.

Our engine incorporates advanced forensic byte carving algorithms. It scans the raw byte stream for known binary signatures, including JPEG markers (\xFF\xD8\xFF to \xFF\xD9), PNG headers (\x89PNG), TrueType font tables (\x00\x01\x00\x00), and compressed FlateDecode streams.

By carving out and reconstructing these isolated asset streams, our tool allows users to recover vital evidentiary photographs, embedded receipts, and raw textual paragraphs even from files that other PDF software declare completely unrecoverable and permanently lost.