Pitch Accuracy Test
Choose a target manually, or prepare a target session.
No session score yetPlay the target, then start the microphone. Each attempt ends after one voiced second with a stable note, or after eight seconds. No grade is generated from silence alone.
Use one clear voice or instrument note in a quiet room. Audio stays in your browser.
In-tune share is the voiced duration within your selected tolerance. Pitch spread is the standard deviation in cents. These describe this attempt, not overall musical ability.
| Target | Tolerance | In tune | Offset | Voiced |
|---|---|---|---|---|
| Complete an attempt to see its measurements. | ||||
Settings & measurement notes
Microphone features need HTTPS and browser permission. Stop or hide/leave the page to release input. Reference tones pause measurement. Noise, harmonics and polyphonic sound can affect estimates.
A4 defaults to 440 Hz. An octave doubles frequency; twelve equal-tempered semitones make an octave.
Related tools
- Vocal Range Test — Find comfortable singing endpoints and choose the next appropriate task.
- Voice Hz Test — Understand retained fundamental-frequency statistics in a short sample.
- Vocal Warm-Up Generator — Organize a fixed-outline timed routine with optional bounded demos.
- AI Voice Analysis — Interpret local acoustic features and the optional custom consistency formula.
The Pitch Accuracy Test compares a sung note with a reference target. Choose a comfortable pitch and tolerance, listen, then attempt the note with the microphone. Review voiced-time accuracy, cents offset and pitch spread rather than treating one percentage as a complete singing grade.
Check one target first
- Choose a comfortable target within C2–C6.
- Select a tolerance of ±10, ±25 or ±50 cents.
- Play the target and allow the reference sound to stop.
- Start the microphone and sing a clear comfortable vowel.
- Review the completed attempt before repeating.
Use headphones when speaker playback might enter the microphone. A reference tone captured from speakers can be mistaken for your own matched note.
The attempt ends after a usable stable voiced second or approximately eight seconds. A target outside your comfortable voice is a reason to choose another pitch, not to push through the time window.
Understand cents and target offset
A cent is one hundredth of an equal-tempered semitone. The offset compares detected frequency with the selected target using their ratio.
At A4 = 440 Hz, a detected 448 Hz is approximately 31.2 cents above the target. It remains near the note A4 but falls outside a ±25-cent tolerance.
A positive offset means sharp and a negative offset means flat relative to the chosen target. Check the target octave and A4 reference before interpreting the number.
Read the attempt measures together
| Measure | Meaning |
|---|---|
| In-tune share | Duration-weighted detected voiced time inside tolerance |
| Median offset | Middle signed distance from the target |
| Pitch spread | Variation in cents during the attempt |
For a hypothetical four voiced seconds with three seconds inside the band, in-tune share is 75%. Silence and rejected frames are not counted as correctly sung time.
A median near zero can coexist with wide movement above and below the target. A narrow spread can also coexist with a consistently sharp or flat note.
Use the prepared target session
Choose 5, 10, 15 or 20 targets and a lower C3–C4, middle C4–C5 or higher C5–C6 group. Prepare the session, then hear and attempt each target.
The targets cycle through a major-scale pattern in the chosen group. This is not a complete-song test, a random melody examination or an automatic personalized range assessment.
Next / skip target moves the session forward. An unattempted or untracked position remains distinct from a measured target hit.
Distinguish session hits from voiced-time accuracy
The session score counts targets whose median offset falls within tolerance. Its denominator includes attempted or skipped positions once the session has voiced results.
For example, three target hits, one voiced out-of-band attempt and one skipped position give three out of five, or 60%. That is not the same calculation as the within-tolerance share of every voiced frame.
A sample can therefore have different session and attempt percentages. Read the labels and detected/skipped counts rather than assuming they are contradictory measurements.
Keep repeated comparisons fair
Use the same target group, tolerance, A4 reference, microphone and distance when comparing sessions. A wider tolerance makes the same input easier to classify as in tune.
Natural vibrato can cross a narrow band. The tool measures that movement but does not decide whether it is artistically appropriate.
Review the last ten attempts and use the available attempt CSV if you want a record. These are working-session measurements rather than a permanent practice account.
Know what this test does not establish
This is vocal production against a reference. It is not a fine-pitch hearing-discrimination test, a diagnosis of tone deafness or a perfect-pitch assessment.
It does not score rhythm, diction, expression or a full song performance. For endpoints, use the Vocal Range Test. For a short unguided frequency summary, use the Voice Hz Test.
The AI Voice Analysis page offers a different held-note consistency formula without an external target. A high consistency result there is not proof of target accuracy here.
Resolve input problems comfortably
Check microphone access and input selection, reduce nearby sound and avoid distorted levels. A harmonic can create an octave error, while an irregular sound may not be tracked.
No detected pitch is not automatically a failed musical attempt. Resolve the recording condition and repeat an easy note before drawing a conclusion.
The microphone is analyzed locally without recording or upload. Stop input when finished. The Vocal Warm-Up Generator can organize a separate gentle routine but does not evaluate performance.
Frequently asked questions
Is this a fine pitch hearing test?
No. It measures what you produce against a target, rather than the smallest pitch difference you can hear.
Does a high percentage mean I am a good singer?
It describes a limited target-matching task. Many aspects of a complete performance are untested.
Why do session and attempt percentages differ?
Session hits use median target offset by position; attempt accuracy uses detected voiced duration inside the tolerance.
Can I choose a tighter tolerance?
Yes. Keep the setting with the result and avoid comparing percentages as if the tolerance had not changed.
Can I grade an uploaded song?
This workflow uses live target attempts. It does not judge an uploaded complete performance.
