Local voice safety layer v0.4.0

Record, check and clean a voice transcript before an LLM sees it.

VoiceSafeKit transcribes audio in the browser, finds privacy and safety risks and gives you cleaner text to send on. It now includes SSN, IBAN and CVV detection, confidence scores, conversation history and highlighted transcript spans.

Try the live checker How it was built See progress

VOICE / SAFE / KIT v0.4.0 REVIEW

Review before sending. Found private details.

Simple workflow

Capture. Check. Protect.

Capture

Record your voice, upload a transcript or paste text directly. Conversation mode lets you chain turns and track risk across the session.

Check

Detect PII, credentials, SSNs, IBANs, emotional distress and coercion signals. Every finding shows a confidence score and inline highlight.

Protect

Get a redacted transcript, a detailed redaction map and simple guidance. Export the full result as encrypted JSON when you need to save it.

Live app, runs entirely in your browser

Try it now

01 Capture

Record your voice or load a sample to get started.

Voice to text

Press record, speak then press stop. Your words appear below.

Upload .txt file

Transcript

02 Check

Decision REVIEW

Risk score 80/100

03 Protect

A cleaned transcript with sensitive parts removed, plus guidance for the assistant.

Export key

Safe transcript