⚠️Beta — you are the authority. This silences the ranges you mark and (optionally) suggests where spoken PII is, but a transcript is never perfect. Listen back to the exported file before releasing it. Everything runs in your browser — the recording is never uploaded. DubbleBooked does not certify compliance; counsel remains responsible under CrR/CrRLJ 4.7.
1 · Load the recording
Witness interview audio (mp3, wav, m4a). Stays in your browser — nothing is uploaded.
2 · Mark what to silence
Drag across the waveform to mark a range. Play to review. Or auto-suggest spoken PII below.
0:00 / 0:00
Auto-redact spoken PII · runs locally · no listening
Transcribes the recording in your browser (nothing uploaded), then auto-silences where it hears SSNs, dates, phones, emails, account numbers — and names. You confirm by reading the timestamped transcript, not by listening to the whole thing. First run downloads the models.
Names are detected automatically now, but a transcript is never perfect — read the highlighted lines and add or trim ranges in step 2 before you export. Chunk-level timing means flagged ranges are silenced a little generously (safer to over-cover).
3 · Silence & export
Zeroes the audio inside every marked range and exports a new WAV. The redacted audio is gone, not muted-on-playback.
Keep the exact redacted file in the client's case record and serve a copy on the prosecutor per the rule. Output is WAV; convert to mp3 afterward if you need a smaller file.
DubbleBooked · Audio Redactor · runs entirely client-side · beta, listen back before releasing.