Take the names out before you share the recording

It finds the names, phone numbers, addresses and ID numbers in a recording or transcript, and you choose what comes out. You get bleeped audio, a cleaned-up transcript and a log of every change. It's free, and the file never leaves your computer.

How it works

Doing this by hand means listening to every minute, often twice. Redact finds the details and you decide what goes.

  1. Load or transcribe the recording

    The Whisper speech model runs inside your browser. If you already have a transcript from Zoom, Teams or YouTube, load that instead.

  2. Let it find the personal details

    It looks for names, phone numbers, emails, street addresses and full dates. Tax file, Medicare and ABN numbers are checked against their real checksums, so random digits don't get flagged.

  3. Go through every one

    Click a highlight to keep that word, or click any other word to remove it. You can also list terms that must always go, or never.

  4. Download what you need

    Take the bleeped or silenced audio, the redacted transcript and subtitles, and a log of every change. If you swapped names for labels, there's a private key too.

Questions

Is my recording uploaded?
No. Your browser opens the file and does the work on your computer. The first time you transcribe, it downloads the speech model from Hugging Face and the engine from jsDelivr, and keeps them for next time. Your recording and transcript aren't sent anywhere. Once the model has loaded you can disconnect from the internet and it keeps working.
Will it catch everything?
No tool does. It catches most of it quickly and shows you every decision, so you make the final call. It can miss nicknames, unusual names at the start of a sentence, and things described rather than stated, like "the house next to the servo". Read the whole transcript before you share anything.
What does it cost?
Nothing. It's free here in your browser. I'm thinking about a desktop version for long recordings, video and batches. Tell me if you'd use one.
Which languages work?
Transcription works in most languages. Name detection is tuned for English and Australian formats for now. In other languages it still finds numbers, emails and anything on your always-redact list.
Can it do video?
It reads the audio track of most MP4 files and gives you back redacted audio and text. It doesn't blur faces or re-export the video yet.
Is this legal advice?
No. Whether you can share a recording, and what has to come out of it first, depends on your obligations. Check them before you rely on it.

Need this for long recordings, video, a whole team, or on machines that never touch the internet? Tell me about it.