Strip personal data from text. Offline.
NixText finds names, EGN, phone numbers, addresses, IBANs and more in Bulgarian and English text. You review every finding, then it replaces them on your own computer. Nothing is uploaded.
- Works without internet
- Bulgarian, English and mixed text
- No GPU or Python needed
- Word, Excel, PDF and text files
Your text
Anonymized
The app
Paste text or drop in whole documents. Every finding is listed with its type, so you can skip anything that should stay before you save.
Supported formats
Word
Text in paragraphs, tables, headers and footers is replaced in place. Formatting stays; author metadata is cleared.
Excel
Every cell on every sheet. Formulas are left as formulas.
Your choice: a PDF with the personal data blacked out (layout kept, nothing left underneath) or an editable Word copy.
Text files
Also .tsv, .json, .xml, .html, .yaml, .ini, .log and .srt subtitles. The output keeps the same format.
Pasted text
Paste from email, chat or anywhere else, review the findings, then copy or save the result.
Not yet
Scanned PDFs (images without text) and old .doc/.xls files. Save old files as .docx/.xlsx first.
Built for Bulgarian documents
Generic tools treat Bulgarian as an afterthought. NixText checks Bulgarian identifiers the way the institutions issue them, and knows every settlement in the country.
People
Full names with patronymics, female and male surname forms, titles like г-н, д-р, инж.
Places
All 5,351 towns and villages and 28 provinces, plus foreign cities.
Organisations
Companies by legal form, ministries, courts, universities, hospitals, banks.
Phone numbers
Validated against the Bulgarian numbering plan: mobile, landline, +359, 00359, 0800, 0700.
Addresses
Streets and boulevards with numbers, residential complexes, block, entrance, floor, flat, postcode.
IDs and money
EGN, LNCh, EIK/BULSTAT and VAT numbers with checksum checks, IBAN, card numbers, licence plates.
People
Full names, first names on their own, names after titles such as Mr, Mrs and Dr.
Places
Cities, countries and regions anywhere in the world.
Organisations
Companies, institutions and brands, including legal forms like Ltd and Inc.
Phone numbers
International and national formats, validated against official numbering plans.
Email and web
Email addresses, web addresses, IP and MAC addresses.
Money and IDs
Payment cards with checksum checks, IBANs, crypto wallets, and US and UK identifiers such as SSN and NHS numbers.
- Date of birth: 2 January 1975
- Region and birth order; the 9th digit encodes sex
- Check digit: weighted sum 63, 63 mod 11 = 8 ✓
Why checksums matter
A ten-digit number is only flagged as an EGN if its date is real and its check digit matches. Order numbers, invoice numbers and random digits stay untouched, so your documents keep their meaning.
Your documents never leave your computer
Anonymization runs locally. There is no account, no cloud service and no telemetry.
Analyse
Paste text or add files. NixText detects the language and finds personal data in seconds.
Review
Every finding is pre-selected. Skip what should stay, or add words to a permanent ignore list.
Replace
Choose how: type labels like <PERSON>, masking with stars, removal, or one-way hashing.
Optional enhanced models are downloaded only when you click Download, verified against published checksums and stored on your computer.
Measured on Bulgarian text
F1 score for people, places and organisations on a held-out test set of 103 Bulgarian documents with 293 labelled entities. Higher is better.
Exact-match F1. The test set is small and was written for this evaluation, not drawn from real case files, so treat the numbers as a comparison rather than a guarantee. Identifiers such as EGN, IBAN and phone numbers are rule-validated and not part of this score.
Download
Everything is bundled. Install and run, no Python or other software needed.
Windows
Builds coming soonmacOS
Builds coming soonLinux
Builds coming soon| Minimum | Recommended | |
|---|---|---|
| Processor | 2 cores, 64-bit. No graphics card needed | 4 cores or more |
| Memory | 4 GB (built-in models) | 8 GB (enhanced models) |
| Disk | About 1.3 GB | About 2 GB with enhanced models |
| Internet | Not needed | Only to download enhanced models once |
# command-line version, included nixtext-cli contract.txt # → contract.anonymized.txt nixtext-cli letters/*.txt --operator mask nixtext-cli --download bg-enhanced
Questions
Is my text sent anywhere?
No. Analysis and replacement run on your computer. The app only connects to the internet when you choose to download an enhanced model.
Which files can I process?
Plain-text formats such as .txt, .csv, .md, .log, .json and .xml, one at a time or in batches. Results are written next to the originals or into a folder you choose.
Does it catch everything?
No automatic tool does. That is why NixText shows every finding before replacing it, so you can check the result. Rare names and unusual spellings can be missed.
Can I use it in Bulgarian?
Yes. The whole interface is available in Bulgarian and English, and it follows your system language on first start.