Hi, I'm Brian, a Fractional Head of Research & AI Builder who's deep diving into AI by building practical tools.

I built Redact to solve a problem I kept running into: manually scrubbing PII from user interview transcripts before sharing them with ChatGPT, Claude, or NotebookLM to help with analysis. It was tedious, error-prone, and slowing down my client work.

Plus, I wanted an excuse to learn Claude Code (officially addicted now).

How it works

  • 100% local processing using a BERT ML model trained for NER (Named Entity Recognition) + smart regex patterns, powered by Transformers.js
  • No data leaves your browser - processing is all done locally using JavaScript, ML model is downloaded to your local memory, no data is sent to any server
  • Your files are never uploaded, stored, or transmitted anywhere - they stay on your device
  • Supports .txt, .vtt, and .docx files up to 1MB
  • Redacts names, emails, phone numbers, locations, organizations, Social Security Numbers, credit card numbers (with validation), and dates of birth automatically
  • Export redacted transcripts as Text, Markdown, or Word formats, with optional line numbers

How you can help

I made this for myself, but I think it's helpful for the broader community. It'd be hugely appreciated if you tried it out and let me know what's missing.

Found a bug? Have an idea?and help make this better for everyone.

Want updates on new Redact functionality?.