When you use the Deep Voice Test on DeepVoiceTest.com, your microphone audio is captured locally in your browser, analysed to detect your fundamental frequency, and then discarded. It is never recorded. It is never stored. It is never transmitted to any server. Nothing you say, sing, or vocalise near your microphone during this session leaves your device.
This page documents exactly how that works — step by step.
How the Microphone Pipeline Works
The Deep Voice Test uses your microphone as a frequency measurement instrument. The entire pipeline — from permission prompt to result display — occurs within your browser. No stage of this process involves a network request of any kind.
Stage 1 — Permission prompt The browser displays a microphone permission prompt before any audio is accessed. This is a browser-enforced security requirement. DeepVoiceTest.com cannot access your microphone without your explicit approval at this prompt. If you decline, no audio of any kind is accessed and the tool does not run.
Stage 2 — Local audio stream Once permission is granted, the browser opens a live MediaStream from your microphone using the getUserMedia() API. This stream exists only in your browser’s active memory. It is not routed to any external server, not logged by any analytics or advertising tool, and not written to any browser cache or persistent storage.
Stage 3 — Web Audio API processing The microphone stream is connected to a MediaStreamAudioSourceNode within the browser’s AudioContext. From there it feeds into an AnalyserNode configured with a 4096-sample FFT window for low-frequency resolution. All processing occurs entirely within the browser’s local audio graph. No audio data exits this graph at any point during analysis.
Stage 4 — F0 detection and result display The FFT data is processed using autocorrelation-based fundamental frequency detection, bounded to the 80–330 Hz deep voice range. The result — a number in Hz — is what appears on screen. The audio frames that produced it are not retained. No voice recording, no waveform capture, and no spectral data snapshot is stored or transmitted.
Stage 5 — Session clear When you close or reload the tab, the MediaStream is closed, the AudioContext is suspended, and all in-memory audio buffers are released. There is no persistent copy of your voice audio anywhere — on your device or on any server — after this point.
What This Site Never Does With Your Audio
- Never records your voice to any file, database, or storage medium
- Never transmits voice audio over the internet to any server, including this site’s own server
- Never shares voice audio with Google Analytics, advertising networks, or any third-party service
- Never stores voice audio in browser localStorage, sessionStorage, IndexedDB, or cookies
- Never retains voice audio after the analysis session ends or the tab is closed
- Never uses your voice audio for any purpose other than displaying the real-time frequency reading on screen
- Never accesses the microphone without displaying the browser’s permission prompt first
Why Voice Data Stays Local — By Design
Voice is one of the most personally identifiable biometric signals. Fundamental frequency, harmonic structure, and vocal tract resonance characteristics are unique to each individual and can be used to identify a speaker. DeepVoiceTest.com is architected so that none of this data ever reaches a server — not as a compliance measure, but as a foundational design decision.
There is no server-side voice processing component on this site. The local-only pipeline is not enforced by policy alone — it is the only pipeline that exists. Even in the event of an unexpected client-side error, there is no server endpoint to which voice data could be transmitted.
Advertising and Analytics Are Separate From Your Audio
DeepVoiceTest.com does not currently run advertising. If basic analytics tools are present on this site, they operate entirely separately from the Web Audio API pipeline. Analytics collect anonymous page-level data — page views, browser type, approximate region — through standard HTTP request metadata. They have no access to the Web Audio API, no access to the MediaStream, and no ability to interact with microphone audio in any form.
The microphone permission you grant to this site is scoped to the Web Audio API analysis pipeline only. It does not extend to any other script or service on the page.
File Uploads
The Deep Voice Test does not involve file uploads of any kind. No audio files are uploaded to any server as part of using this tool. The only audio data processed is the live microphone stream handled locally as described above.
HTTPS Encryption
All pages on DeepVoiceTest.com are served over HTTPS. All data exchanged between your browser and this site’s server — page content, scripts, and voluntary contact form submissions — is encrypted in transit. Your voice audio never travels over the network and therefore requires no network-level encryption — it is protected by the fact that it never leaves your device.
Cookies and Session Data
Your frequency readings are displayed on screen during your session only. They are not written to cookies, not stored in localStorage, and not sent to any server. Closing or reloading the page clears all session data entirely. For a full explanation of what cookies this site uses, see the Cookie Policy.
Children’s Privacy
DeepVoiceTest.com does not knowingly collect personal data from children under the age of 13. The tool does not require account registration and does not request personal information from any user. Voice audio is never transmitted or stored for any user. If you believe a child has submitted personal data through the contact form, contact Lucas Reed at contact@deepvoicetest.com and the data will be deleted promptly.
Reporting a Security Concern
If you identify a potential security vulnerability on DeepVoiceTest.com — including any behaviour suggesting microphone audio or user data is being handled contrary to what is documented on this page — please report it immediately via the Contact page with the subject line “Security Concern.” Lucas Reed reviews all security reports personally within 48 hours.
Responsible disclosure is welcomed. Details of any verified vulnerability will not be published until a fix is in place.
Your Privacy Rights
Your rights regarding any personal data this site may hold are documented in full on the Privacy Policy and GDPR and CCPA pages.
Written by Lucas Reed, founder of DeepVoiceTest.com. Last updated: June 2026.