When a file is too damaged for its own program, its words are often still inside. This pulls every readable string out of the raw bytes.
This is the strings technique from digital forensics, simplified for rescue work. Many file formats store their text as plain readable runs of characters inside a binary container, so when the container breaks (corrupt document headers, truncated databases, mangled backups), the words survive even though the file will not open. The tool scans the raw bytes for runs of printable characters, in both plain (ASCII/UTF-8-style) and UTF-16 (every-other-byte) encodings, the latter being how Office documents and Windows programs store text, and collects everything found into one downloadable text file.
Set expectations correctly: recovered text arrives without formatting, order can jump around inside complex files, program code and metadata mix in with document content, and compressed or encrypted regions (modern .docx/.xlsx innards are ZIP-compressed, for example) yield nothing here, for those, repair the container first or use the format-specific tools. Where this tool shines is the classic disasters: the corrupt .doc/.txt/database/log file whose contents exist nowhere else. Ten minutes of reading raw strings beats losing the content outright.
The output is full of gibberish fragments, is that normal? Yes: file headers, internal labels and code all qualify as 'printable strings'. Raise the minimum length to 6 or 10 to filter toward human sentences, and search the output for words you know were in the document.
Why is there almost nothing from my .docx? Modern Office files compress their text (ZIP/DEFLATE inside), so the words are not stored readably. Try the ZIP Recovery Tool on it, a .docx is a ZIP archive, and once extracted, its document.xml contains the text.
Is any of my file uploaded? No, the scanning happens in the page on your machine, which is exactly what you want for a damaged file containing private material.