About Pitch Detector

The Problem We Are Solving

The pitch detection problem sounds simple: listen to audio, report the note. But singers have known for decades that existing tools fail them in specific, frustrating ways. The three most common complaints we encountered were:

• The octave error.

  A tenor sings A4 and the tuner shows A3. This happens because pitch detection algorithms that are not tuned for human voice often lock onto sub-harmonics instead of the fundamental frequency. For singers, this makes standard guitar tuners nearly useless.

• The unstable high-note reading.

  In the upper register, confidence drops and the displayed note jumps between adjacent semitones. Without a confidence indicator, singers cannot tell whether the reading is accurate or just noise.

• No way to compare live to recorded.

  Singers know they sound different on recordings than in their head. But most real-time tools show only the current moment, never a comparison against a previous take. The intonation drift between takes goes unmeasured.

Our Approach

We built this pitch detector around three principles that directly address those problems:

1. Voice-type frequency priors.

   Before pitch estimation runs, the algorithm constrains its search space to the frequency range of the selected voice type. A bass singer is bounded to E2-E4. A soprano is bounded to F3-G5. Sub-harmonic candidates outside the range are discarded. This eliminates the octave error for all six voice types.

2. Explicit confidence reporting.

   The YIN pitch estimator already produces a confidence value alongside its frequency estimate. We expose that value directly to the user as a color-coded confidence bar. Singers always know whether to trust the current reading.

3. Dual-track pitch curve display.

   We store the recorded pitch curve separately from the live curve and display both simultaneously on the same canvas. Singers can now see take-to-take intonation consistency at a glance — something previously only available in professional DAW plugins.

Technology Stack

The pitch detection runs on the YIN algorithm (de Cheveigné & Kawahara, 2002), implemented in TypeScript and executed entirely in the browser via the Web Audio API. No audio data is ever transmitted to any server. The application is built with Next.js and deployed on Cloudflare Workers for fast global response times with no cold start latency.

All five tool pages (Pitch Detector, Voice Pitch Analyzer, Singing Note Detector, Vocal Range Test, and this About page) are statically generated at build time, meaning they load instantly from Cloudflare's edge cache with no server-side processing on each visit.

Privacy Commitment

We collect no personal data. No accounts, no cookies for tracking, no audio uploads. The microphone stream is processed in memory inside your browser tab and is discarded when you close or navigate away. The only external request the page makes is to load Google AdSense advertisements that help keep the service free. Ad network behavior is governed by their respective privacy policies.