July 22, 2026 · CensorMate Blog

How to Censor Names, Addresses & Personal Info in Audio and Video

Profanity gets all the attention, but the censoring job that actually keeps people out of trouble is quieter: a vlog that catches your kid’s school name, a podcast guest who blurts out their phone number, a customer-support recording with a card number in it, an HR interview that names a third party. This guide covers how to find and censor spoken personal information in audio and video — and why this particular job should never involve uploading the recording to someone’s server.

What counts as personal information in a recording?

The usual suspects, roughly in order of how often they slip into recordings:

• Names — full names of people who didn’t agree to be in your content, kids’ names, employee names in internal recordings
• Phone numbers and email addresses — spoken aloud far more often than people realize (“just call me at…”)
• Street addresses and locations — home addresses, school names, workplace details
• Identifying numbers — Social Security numbers, account numbers, card numbers read aloud on support calls
• Third-party details — in legal, HR, and FOIA contexts, information about people who aren’t party to the disclosure

One useful mindset: the microphone hears everything twice. Once in the audio, and once more in the auto-captions every platform generates — so a name you missed in the sound gets typed out in searchable text. Censoring the audio and using matching censored captions closes both doors.

Why shouldn’t I upload sensitive recordings to a cloud censoring tool?

Because the fix would create the exact exposure you’re trying to remove. Uploading a recording that contains someone’s personal information to a third-party server is itself a disclosure — one you can’t take back, governed by whatever that service’s retention and training policies say. For regulated contexts it can be worse than pointless: FERPA-covered classroom media, privileged legal recordings, and internal HR material generally can’t be handed to an unvetted processor at all.

The alternative is doing the processing on your own device. CensorMate runs its speech-recognition AI entirely in your browser — the file never leaves your computer or phone, there’s no account, and nothing is retained because nothing was ever sent. For personal-info censoring specifically, that architecture isn’t a nice-to-have; it’s the whole point.

How do I censor a name or number in a recording?

The workflow in CensorMate takes a few minutes:

1. Open censormate.com/app in your browser (computer or phone — no install).
2. Type the terms to censor: the names, numbers, or phrases you already know about. Multi-word phrases work (“123 Maple Street”).
3. Drop in your file — audio or video.
4. Let the AI transcribe it. Every word gets a timestamp, and your listed terms are highlighted automatically.
5. Review the transcript. This is the step that catches what you didn’t list: read it once, click any word to censor it. CensorMate also flags patterns that look like phone numbers, email addresses, or SSN-shaped and card-shaped numbers, and offers to censor them in one click.
6. Pick silence, not a bleep, for sensitive material. A bleep is comedic; silence is neutral and irreversible — the exported file simply contains no audio in the censored spans.
7. Export. Your video quality is untouched (only the audio track is rebuilt), and you can download matching censored captions so platform auto-captions don’t reintroduce the information.

What about a paper trail — proving what was redacted?

For casual use, the clean file is the goal and you’re done. But legal, HR, records-request, and compliance workflows usually need to document the redaction: what was removed, where, and by what rule. CensorMate’s redaction mode produces exactly that — a machine-readable redaction log (term, rule, timestamps) and a printable report with SHA-256 hashes of the source and output files, generated entirely on your device. There’s a fuller write-up on the audio & video redaction page.

Frequently asked questions

Can AI find every name automatically? No, and you should distrust any tool that claims it can. Speech recognition transcribes what was said; deciding what’s sensitive still needs a human pass. The honest workflow is: list what you know, let pattern-detection flag numbers and addresses, then read the transcript once yourself. On a one-hour recording that review takes a few minutes and is the difference between “probably clean” and “clean.”

Does censoring the audio remove the information from the video picture too? No — audio censoring only affects sound. If a document, screen, or street sign is visible, that needs visual blurring in a video editor. Most spoken-PII jobs don’t have this problem, but check once before exporting.

Is a bleep or silence better for personal information? Silence. Bleeps draw attention and carry a comedy connotation; silence reads as neutral redaction. In CensorMate, redaction mode uses irreversible silence — the audio in those spans is genuinely gone from the export, not muted or recoverable.

Can I do this on my phone? Yes — the same tool runs in iPhone and Android browsers with a lighter on-device AI model. Short recordings work best on phones; long ones are happier on a computer.

CensorMate is a free browser-based tool that finds and censors any word in audio and video using on-device AI — files never upload anywhere. Try it free, or read how it compares to other censoring tools.