How does this tool detect names and places?

It runs a compact BERT-based named-entity-recognition model (about 43 MB, quantized) directly in your browser via WebAssembly using @huggingface/transformers. The model identifies people, organizations, and locations. Emails, phone numbers, URLs, and IP addresses are detected with regular expressions since they have rigid, well-defined formats.

Why is the first run slow?

The first time you use the tool, the model weights download from the Hugging Face Hub (~43 MB) and are cached in your browser. Subsequent runs load the cached model instantly.

Is my text uploaded anywhere?

No. All detection and replacement runs in your browser. Your text never leaves your device.

How accurate is the AI?

The NER model is trained on English news text and performs well on common names, companies, and places. It may miss rare names or produce false positives on uncommon vocabulary. Always review the output before sharing the anonymized text.

Text Anonymizer

Redact names, organizations, locations, emails, and phone numbers from text using a small AI model running entirely in your browser. Ideal for scrubbing PII before sharing chat logs, tickets, or documents.

Your data never leaves your device

Frequently Asked Questions

How does this tool detect names and places?

It runs a compact BERT-based named-entity-recognition model (about 43 MB, quantized) directly in your browser via WebAssembly using @huggingface/transformers. The model identifies people, organizations, and locations. Emails, phone numbers, URLs, and IP addresses are detected with regular expressions since they have rigid, well-defined formats.
Why is the first run slow?

The first time you use the tool, the model weights download from the Hugging Face Hub (~43 MB) and are cached in your browser. Subsequent runs load the cached model instantly.
Is my text uploaded anywhere?

No. All detection and replacement runs in your browser. Your text never leaves your device.
How accurate is the AI?

The NER model is trained on English news text and performs well on common names, companies, and places. It may miss rare names or produce false positives on uncommon vocabulary. Always review the output before sharing the anonymized text.