Voice Anonymizer for Recordings

Protect a source's identity before publishing. Upload an interview or spoken audio file — pitch shift, formant shift, and timing jitter are applied entirely in your browser. No audio ever leaves your device.

🎙️
Click to choose a file, or drag & drop here
MP3, WAV, OGG · Max ~50 MB
File: —  ·  —
0 st

Negative = deeper voice, positive = higher. Set to 0 to auto-randomize ±4 st.

1.00×

Shifts vocal tract characteristics independently of pitch. Set to 1.00 to auto-randomize 1.15–1.35×.

auto

Adds random micro-timing variations (ms per segment) to break voice rhythm patterns. 0 = auto (±30 ms).

Pitch: auto Formant: auto Jitter: auto 🔒 100% local
🔒 All processing runs in your browser using the Web Audio API. Your audio file is never uploaded to any server. This tool works entirely offline once the page loads.
Decoding audio…
✅ Voice anonymized

How it works

Three independent transformations alter the speaker's vocal identity while keeping speech intelligible:

1. Pitch shift Shifts the fundamental frequency (F0) of the voice by ±4 semitones. A semitone is 1/12 of an octave; ±4 st changes apparent gender/age without making speech sound robotic.
2. Formant shift Alters the vocal-tract resonances (F1–F3) independently of pitch by a factor of 1.15–1.35×. Formants encode the speaker's unique mouth and throat shape — shifting them is the key to defeating voice recognition.
3. Timing jitter Adds ±30 ms of random micro-delay per audio segment (~300 ms chunks). Speech rhythm and pause patterns are speaker-specific biometrics; jitter breaks the temporal fingerprint.
Client-side only Everything runs via the Web Audio API and OfflineAudioContext in your browser. The WAV export uses raw PCM encoding — no server round-trip, no cloud storage, no account needed.

Pitch shift math: A semitone ratio is 2^(1/12) ≈ 1.0595. Shifting by N semitones multiplies frequency by 2^(N/12). For example, −4 st → rate = 2^(−4/12) ≈ 0.794 (deeper). +4 st → rate = 2^(4/12) ≈ 1.260 (higher). Formant shift is applied as an additional spectral envelope warp on top of pitch adjustment.

Frequently asked questions

Will the anonymized voice still be intelligible?
Yes, when pitch shift stays within ±4 semitones and formant shift stays below 1.4×, speech remains clearly understandable. The tool defaults are tuned for this balance. Going to extremes (±6 st, 1.5× formant) will produce more distortion at the cost of some intelligibility — use the sliders if you need stronger protection.
Can automatic speaker identification software still identify the speaker?
Combined pitch + formant + jitter transformation significantly degrades the accuracy of modern speaker-identification systems (e.g. those using i-vectors or x-vectors). Formant shift is the most important layer because it alters the vocal-tract spectral envelope that speaker-ID models rely on most. No client-side tool can offer a 100% guarantee against all adversarial attacks, but this combination substantially raises the bar for casual or semi-professional de-anonymization attempts.
What output format does the tool produce?
The download is a standard WAV file (PCM 16-bit, same sample rate as the input). WAV is uncompressed and universally accepted by editing software, broadcast systems, and publication platforms. If you need MP3, import the WAV into any audio editor (Audacity, GarageBand, etc.) after downloading.
Is my audio uploaded to any server?
No. The Web Audio API runs entirely in your browser. Your file is read locally by the browser's file picker — it is never transmitted over the network. You can verify this by opening your browser's developer tools → Network tab and observing zero outbound requests while processing. This makes the tool safe for sensitive journalistic, legal, or investigative recordings.
What file formats are supported as input?
Any audio format your browser can decode natively: MP3, WAV, OGG Vorbis, AAC (in Safari/Chrome), FLAC (in Firefox/Chrome). Most browsers support MP3, WAV, and OGG. If your file is in a less common format (e.g. M4A, AIFF), convert it first with a tool like VLC or FFmpeg, or rename an AAC file to .m4a and try it in Chrome.