How an Online Guitar Tuner Works
A web tuner does the same job as a clip-on or pedal tuner — listen to a note, decide which note it is, and show how far off it is — but it does that work inside a browser tab. This page is the honest version of how that happens, including where the limits sit.
The audio pipeline
When you click Start Tuning, four things happen in quick succession:
- Permission. The browser asks whether the site can use your microphone. Without permission, none of the rest of the pipeline can start.
- Capture. The browser opens an audio input stream — usually 44,100 or 48,000 samples per second — and hands it to the page as a continuous flow of numbers representing the sound pressure at the mic.
- Analysis. A short window of recent samples (a few thousand at a time) is run through a pitch-estimation routine. This estimates the fundamental frequency of whatever is dominating the input.
- Display. The frequency is converted into the nearest note relative to A4 = 440 Hz, and the deviation from that note (in cents) is drawn on the meter.
The whole loop runs many times per second. Closing the tab releases the audio context, and the stream stops cleanly.
Why autocorrelation, not just an FFT
The obvious way to find the pitch of a sound is to run a Fast Fourier Transform and look for the tallest peak. For a synthesised pure tone, that works. For a real plucked guitar string, it doesn't, because the string is rich in harmonics — overtones that sit at integer multiples of the fundamental. Sometimes a harmonic is taller than the fundamental, and a naïve FFT-peak detector will lock onto the wrong frequency by a clean octave.
Autocorrelation looks at the signal differently. It asks: "After what delay does the signal most resemble itself?" The answer is one period of the fundamental, regardless of which harmonic happens to be loudest. The result is far more robust for plucked strings.
Real-world detectors usually combine the two — an FFT-based rough estimate followed by an autocorrelation refinement, with some heuristics to suppress octave errors and ignore non-periodic input like room noise.
From frequency to a note and a cents reading
Once the detector has settled on a frequency, the rest is arithmetic. The nearest equal-tempered note is the one whose frequency is closest to the input on a logarithmic scale. The deviation is expressed in cents, where one semitone equals 100 cents and an octave equals 1,200.
So a reading of "A2, −7 cents" means: the signal is closest to 110 Hz, sitting about 7 hundredths of a semitone below it. That's well within the range a careful turn of the tuning peg can correct.
To a trained ear, anything within roughly ±5 cents is in tune. Most chord work is comfortable inside ±10 cents. Beyond ±20, you can hear it.
Latency: how fast is "real-time"
"Real-time" is shorthand, not a technical claim. There are always small delays:
- The microphone, its driver, and the operating system together usually add a few milliseconds before samples reach the browser.
- The analysis window itself needs to contain enough samples to identify a low E (about 82 Hz, with a period of roughly 12 ms). That sets a hard floor on how quickly the detector can react.
- The browser then renders the result on the next animation frame, which on a 60 Hz display is up to 16 ms.
Added up, the perceived delay between a plucked note and the meter responding is usually well under a quarter of a second. It's not zero, but it's not noticeable while tuning.
What can throw the detector off
- Background noise. A fan, an air-conditioning unit, or a voice in the room can compete with the string for the detector's attention. Quieter rooms produce steadier readings.
- Two strings ringing at once. Damp the strings you don't want to read, or wait for them to die away.
- Very low or very weak signals. If you're tuning a bass or a very low string and the readout flickers, move closer to the microphone or play a touch harder.
- Distortion. A clipped, overdriven signal is rich in non-harmonic energy and confuses pitch detection. Tune clean.
- Sympathetic resonance. Strings can pick up energy from other vibrating strings. Resting a finger on the strings you aren't tuning gives a cleaner reading.
Comparison: web tuner vs. clip-on vs. strobe
Each tool has a place. Roughly:
- Web tuner. Free, always available, works anywhere you have a browser and a microphone. Accurate enough for everyday tuning. Limited by the quality of the input mic and the noise in the room.
- Clip-on tuner. Reads vibration through the headstock, so it's immune to background noise. Convenient on stage. Comparable accuracy in practice.
- Strobe tuner. Designed for setting intonation and for very precise reference work — typically accurate to a fraction of a cent. Overkill for routine tuning, but indispensable when adjusting bridge saddles or matching a piano.
The web tuner on this site is intended for the first category: get the guitar in tune, fast, before you start playing. For a full intonation setup, a strobe tuner — physical or specialist software — is still the right tool.
Privacy of the audio
Every step of this pipeline runs inside your browser. The audio stream is read by the page's JavaScript, analysed locally, and discarded as new samples arrive. Nothing is recorded, nothing is sent to a server. The privacy policy covers this in more detail, but the short version is that the tuner can't share something it never stores.
Putting it into practice
Knowing how the detector works changes how you use it. Play cleanly. Let the note ring. Damp the strings you're not reading. Quiet room, fresh attack, second pass. With those habits, a browser tuner gets you into tune as reliably as the hardware sitting next to it on the desk.
To try it now, open the tuner. To understand the targets it's aiming at, see standard tuning explained or the alternate-tunings reference.