Published by PitchChanger · Updated · How these guides are made
Hear the difference
The C–G–Am–F source used for the reported engine observations.
Pause your workspace preview before comparing these clips.
Original synthetic material. The settings below distinguish authored inputs from processed outputs; no performer or physical-device test is implied.
How to use this workflow
Choose a harmonic section
Find a verse or chorus with bass and chords, not only a transition or percussion. Keep at least ten seconds and enough music to include more than one chord.
Make the excerpt explicit
Save the passage with Audio Trimmer, preserving the original. Import it into Song Key Finder and note the analyzed time range, leading key and alternative.
Compare another tonal section
Repeat with a second representative passage. If the answers differ, check the score, bass movement and phrase endings rather than choosing whichever label you prefer.
Know which part the analyzer actually uses
For sufficiently long input, this implementation removes the outer ten percent at each end and limits analysis to a central window of at most sixty seconds. For a short input where that interior would be under ten seconds, it uses the full input instead. It does not analyze every minute of a long song.
If the musical passage you care about lies outside the reported range, trim it to a separate input and rerun. A whole-file label can otherwise describe a different section from the one you intended to transpose.
Match the excerpt to the musical question
For a backing-track transposition, choose the section you will actually sing over and include its harmonic resolution. For a song with a modulation, inspect sections on both sides. For an exposed melody, include accompaniment when available; the opening note alone does not define the key.
A distorted sustained sound or a dense drum break can have spectral peaks without establishing the harmonic center. Try a clearer section before interpreting a weak estimate. A display of Unknown is a useful limit rather than an instruction to keep reanalyzing identical input.
Use agreement as supporting evidence
The engine compares twelve major and twelve minor pitch-class templates. Its confidence is a score derived from the gap between candidates, not a calibrated probability that a label is correct. The observed scores for this simple fixture are about 0.244 and 0.361; neither should be reported as a percentage accuracy.
Even agreement across two excerpts does not replace a musical check. Use phrase endings, a trusted chord progression or an instrument to test the likely tonic. Uniform pitch shifting preserves mode; selecting another tonic does not turn a major arrangement into minor.
Common questions
How short can the excerpt be?
The tool requires at least ten seconds. Ten seconds is a minimum input requirement, not a guarantee that the passage contains enough harmonic information.
Should I analyze a chorus or verse?
Choose the section relevant to your task and compare another tonal section. A song can change key or contain ambiguous passages, so neither label is automatically the only correct one.