Does perfectly in-tune pitch mean a track is AI-generated?

No. Software instruments, sample libraries and pitch correction all place notes exactly on the semitone grid, and pitch correction has been standard on commercial vocals since the late 1990s. Grid-exact pitch indicates software or editing, not authorship.

What is the reference tuning figure?

The tool searches for the semitone grid that best fits the pitch content and reports where its A4 sits. A result near 440 Hz means concert pitch. An offset points to deliberate alternative tuning, a tape or vinyl transfer running fast or slow, a pitch-shifted upload, or an instrument tuned by ear.

What counts as human pitch accuracy?

Roughly 6 to 20 cents median distance from equal temperament is what competent singing and playing produce — audibly in tune, measurably imperfect. Under about 6 cents with most frames on the grid points to sequencing or hard correction.

What does the jitter figure measure?

How much pitch moves from frame to frame inside notes. Natural voices and acoustic instruments drift by a few cents even on a note meant to be held flat. Values near zero mean the pitch line is mathematically still, which synthesis and heavy correction both produce.

Why does my track report no pitch at all?

Because there is no single periodic component to lock onto. Percussion-led tracks, dense mixes, heavy distortion and noise-based textures all defeat pitch tracking, so the tool says so rather than reporting a number it did not find.

Is my audio uploaded anywhere?

No. The file is decoded and measured by your own browser and released when you close the tab. Nothing is transmitted, stored or logged, so unreleased material is safe to check.

Free tool

Pitch & tuning checker

Track a file's fundamental frequency, find the reference tuning it implies, and measure how closely notes land on the semitone grid and how steady held pitches are. Decoding and measurement happen in your browser — the file never leaves your machine.

Short answer

Can tuning tell you whether a song is AI-generated?

No. Notes landing exactly on equal temperament means the pitch came from software — a synth, a sampler, or a correction plug-in — and that describes most records made in the last twenty-five years. What pitch measurement genuinely settles is narrower: whether a performance was corrected, and whether a file has been pitch-shifted or transferred off-speed.

Where detection misleads →

Drop an audio file here

MP3, WAV, FLAC, M4A, Ogg or Opus. Long files are measured on three minutes from the middle.

What each number means

Reference tuning is found by testing every semitone grid between 50 cents flat and 50 cents sharp of A440 and keeping the one your file's pitch content fits best. A result within a few cents of 440 Hz is the modern default and tells you almost nothing. A clear offset is more interesting: it appears in alternative tunings, in transfers from tape or vinyl running slightly off speed, and in uploads pitch-shifted to slip past content matching.

Off grid (median) is the typical distance from each tracked pitch to the nearest semitone on that grid, in cents — hundredths of a semitone. Under about 6 cents means placement is effectively exact. Between 6 and 20 cents is what singing and playing produce. Larger values mean expressive movement, non-equal tuning systems, or a mix dense enough that the tracker is following a blend rather than one note.

In tune is the share of tracked frames within 15 cents of a semitone, and note jitter is the median frame-to-frame pitch change inside voiced runs. Read them together: high in-tune share with near-zero jitter is a mathematically still pitch line, which synthesis and hard correction both produce, while a high share with a few cents of jitter is a corrected but still breathing performance.

Why in-tune pitch is not an origin signal

A software instrument plays exactly the note it is asked for. A sampler transposes to exactly the requested interval. A pitch-correction plug-in moves a sung note to the nearest scale degree and, on its more aggressive settings, removes the drift within the note as well. All three predate generative audio, and the third has been standard on commercial vocals since the late 1990s. A file whose notes sit two cents from the grid has told you software was involved in the pitch, and nothing about who or what wrote it.

The reverse inference fails too. Generated vocals frequently carry believable drift and vibrato, because the models learned from performances that had them. Imperfect pitch is not evidence of a human, and perfect pitch is not evidence of a machine author.

What pitch measurement does settle

The useful questions are specific ones. If a file is presented as an unprocessed live vocal, a pitch line that snaps to the grid with no drift inside notes is inconsistent with that description. If a recording is presented as an original transfer, a reference tuning 30 cents off concert pitch says the speed was wrong or the file was shifted. If a track was uploaded to avoid a content match, that shift often shows up here first.

For the full argument — why correction is not an origin signal, how to read cents and jitter, and where pitch readings mislead — see pitch, tuning and AI music.

Use it alongside the other measurements here: tempo and timing for quantisation, structure and repetition for copied passages, file metadata for encoder history, a spectrogram for band structure, and stereo field for mix decisions.

Limits of this tool

Pitch tracking follows one fundamental at a time, so it works best on solo voice, lead lines and sparse arrangements. In a dense mix it will follow whatever periodic component dominates each frame, which may switch between bass, vocal and pad — the numbers stay arithmetically correct while describing something you did not intend to measure. Percussion, distortion and noise textures yield no confident pitch at all. Measurement is bounded to three minutes from the middle of the file, the analysis range is 65 Hz to 1 kHz, and a low-bitrate encode can move tracked pitch by a cent or two from the master.

Questions

Does perfectly in-tune pitch mean a track is AI-generated?: No. Software instruments, sample libraries and pitch correction all place notes exactly on the semitone grid, and pitch correction has been standard on commercial vocals since the late 1990s. Grid-exact pitch indicates software or editing, not authorship.
What is the reference tuning figure?: The tool searches for the semitone grid that best fits the pitch content and reports where its A4 sits. A result near 440 Hz means concert pitch. An offset points to deliberate alternative tuning, a tape or vinyl transfer running fast or slow, a pitch-shifted upload, or an instrument tuned by ear.
What counts as human pitch accuracy?: Roughly 6 to 20 cents median distance from equal temperament is what competent singing and playing produce — audibly in tune, measurably imperfect. Under about 6 cents with most frames on the grid points to sequencing or hard correction.
What does the jitter figure measure?: How much pitch moves from frame to frame inside notes. Natural voices and acoustic instruments drift by a few cents even on a note meant to be held flat. Values near zero mean the pitch line is mathematically still, which synthesis and heavy correction both produce.
Why does my track report no pitch at all?: Because there is no single periodic component to lock onto. Percussion-led tracks, dense mixes, heavy distortion and noise-based textures all defeat pitch tracking, so the tool says so rather than reporting a number it did not find.
Is my audio uploaded anywhere?: No. The file is decoded and measured by your own browser and released when you close the tab. Nothing is transmitted, stored or logged, so unreleased material is safe to check.