How to Use This Tool
This page covers the most common problems encountered when running the Deep Voice Test — including failed detection, results that seem too high or too low, half-period errors, microphone permission issues, and mobile-specific problems. Work through the section that matches your situation. If your problem is not covered here, use the Contact page to submit a technical question.
For background on how the tool works and what a normal result looks like for deep voices, see the How It Works page.
The Tool Will Not Start
Problem: The browser asks for microphone permission but the tool does not respond after I click Allow.
The three most common causes:
- Another application is holding the microphone in exclusive mode. Communication apps — Zoom, Teams, Discord, Google Meet — frequently hold the audio input device exclusively, blocking the browser from opening a new capture stream. Close all other applications that use the microphone, reload the page, and try again.
- The browser has a previously blocked permission for this site. To reset:
- Chrome: Click the lock icon in the address bar → Site settings → Microphone → Allow
- Firefox: Click the lock icon → Connection secure → More information → Permissions → Use the Microphone → Allow
- Safari: Safari → Settings → Websites → Microphone → find deepvoicetest.com → Allow
- A system-level privacy setting is blocking browser microphone access.
- Windows 11/10: Settings → Privacy and Security → Microphone → ensure “Let apps access your microphone” is On and your browser is listed
- macOS: System Settings → Privacy and Security → Microphone → ensure your browser is ticked
Problem: The tool loads but shows no frequency reading when I speak or sing.
Most likely causes:
- Input volume is too low. Check that your system microphone input level is set to at least 60–70%. On Windows: right-click the speaker icon → Sounds → Recording → select your microphone → Properties → Levels. On macOS: System Settings → Sound → Input → Input volume.
- The wrong input device is selected. If multiple audio devices are connected, the browser may be capturing from a device other than the one you are using. Set the intended microphone as the system default input before loading the page.
- Phonation is too quiet or too breathy. The tool requires a clear, voiced sound above the microphone noise floor. Speak or sing at a confident, supported volume. Very quiet or airy phonation — especially at the bottom of a deep voice range where vocal fold closure is least complete — may not register reliably.
- The AudioContext was not unlocked on mobile. On iOS Safari, a user tap is required to unlock the AudioContext before audio capture begins. Tap the start button clearly, then begin phonating immediately.
Result Seems Too High
Problem: My result is consistently showing 160–200 Hz when I expect to be lower.
Harmonic detection. This is the most common accuracy problem in the deep voice range. When a consumer microphone attenuates the low-frequency fundamental, the second harmonic (2×F0) becomes the strongest component in the captured spectrum. The autocorrelation algorithm may peak on the second harmonic rather than the fundamental, returning a result of approximately 2×F0.
How to identify it: If your result is consistently close to double a round number — for example, showing 196 Hz when your expected F0 is approximately 98 Hz — harmonic detection is the likely cause.
Fixes to try:
- Switch to an external USB microphone rated to 80 Hz or below — better low-frequency response makes the fundamental stronger relative to harmonics
- Try humming rather than sustaining an open vowel — nasal resonance can reinforce the fundamental
- Sing on a darker vowel (“oh” or “oo”) rather than a bright one (“ee” or “ah”) — darker vowels tend to have a stronger fundamental
- Ensure OS audio processing enhancements are disabled — high-frequency emphasis applied to the microphone input worsens the harmonic-to-fundamental ratio
Problem: My result is around 300–330 Hz — well above the baritone range.
If you are testing your speaking voice and getting 300+ Hz, check whether your system default input device is set to the microphone you intend to use. Some systems default to a webcam microphone or a communications device that applies voice processing, including pitch enhancement. Also confirm you are not accidentally capturing a higher-pitched voice in the room.
Result Seems Too Low
Problem: My result is below 80 Hz or shows as zero.
The tool’s autocorrelation lag search is bounded to the 80–330 Hz range. A result of zero or a reading below 80 Hz typically indicates one of the following:
- Sub-bass room noise. HVAC systems, traffic rumble, and building structure noise can generate continuous low-frequency energy in the 20–80 Hz range that competes with the voice signal. The tool may lock onto ambient noise rather than your voice. Test in a quieter location or address the noise source.
- Pushing the voice below its natural range. Very strained sub-bass phonation — fry register at extreme depth, for example — is often too aperiodic for reliable autocorrelation detection. The vocal fold closure pattern in extreme fry produces an irregular waveform with a poorly defined period. The tool is calibrated for modal chest voice phonation in the bass and baritone range, not extreme vocal fry or subharmonic registers.
- AudioContext clock desync. Reload the page and retest. If the zero result persists, try a different browser.
Half-Period and Double-Period Errors
Problem: My results alternate between approximately 100 Hz and 200 Hz between runs.
This is a half-period or double-period detection error. The autocorrelation function produces peaks at the true fundamental period and at integer multiples and divisions of it. At very low F0 values — below approximately 100 Hz — the secondary peak at double the period (half the F0) can occasionally be stronger than the primary peak, causing the algorithm to return 2×F0 instead of F0.
Fixes:
- Sustain the vowel more consistently — irregular phonation increases the likelihood of period confusion
- Support the voice with firmer breath pressure — a more closed vocal fold gesture strengthens the fundamental
- Try the test immediately after speaking naturally for 30 seconds — warming the voice slightly tends to produce a more stable phonation pattern
- If the problem persists consistently, your true F0 is most likely the lower of the two alternating values
Microphone-Specific Problems
Problem: My external USB microphone gives a worse result than my built-in microphone.
This can happen when the USB microphone is not set as the system default input device. The browser typically captures from the system default at page load. If you connected the USB microphone after loading the page, reload the page with the microphone set as the system default.
It can also occur if the USB microphone is a communications device with voice processing built in — automatic noise cancellation, de-essing, or high-pass filtering applied internally by the microphone’s DSP can attenuate the low-frequency fundamental just as a consumer microphone’s physical roll-off does. Check the microphone’s companion software and disable any high-pass filter or noise reduction that is applied to the signal before it reaches the browser.
Problem: Results are noticeably different between Chrome and Firefox on the same machine.
Different browsers implement the Web Audio API’s microphone processing differently. Chrome applies automatic echo cancellation and noise suppression to microphone input by default, which can affect low-frequency capture. Firefox handles audio processing differently and may produce a less processed signal. If Chrome gives higher results than expected, try Firefox or Edge. On Chrome, navigating to chrome://flags and disabling hardware echo cancellation may improve low-frequency accuracy.
Mobile-Specific Problems
Problem: The tool does not produce a stable reading on my Android phone.
Android Chrome generally supports the Web Audio API well, but Android’s audio hardware and driver stack introduces additional processing latency and, on some devices, automatic gain control that can compress the low-frequency fundamental. For deep voice measurement, hold the phone close to your mouth — within 15–20 cm — and test in a quiet environment. If results are unstable, desktop use is recommended.
Problem: The tool does not work on my iPhone.
iOS Safari applies strict AudioContext restrictions and additional voice processing that can interfere with low-frequency F0 detection. The AudioContext must be unlocked by a clear tap before capture begins. For deep voice measurement specifically, iOS results should be treated as indicative only. Desktop use with an external microphone gives substantially more reliable results for bass-range F0 detection.
Still Having Problems?
If none of the above resolves your issue, the FAQ covers additional questions about result interpretation and detection accuracy for deep voices. For problems not addressed in either page, contact Lucas Reed directly via the Contact page with the following information:
- Your browser name and version
- Your operating system and version
- Your microphone type and model if known (built-in, USB, headset)
- The result you are seeing and what you expected
- Whether the problem is consistent or intermittent
Response time for technical issues: 24–48 hours.
Written by Lucas Reed, founder of DeepVoiceTest.com. Last updated: June 2026.