Our approach
Built from first principles, explained in full.
MindTimbre uses standard acoustics and arithmetic. Here is what each part does, in plain language, with the formulas behind it.
Pitch
MindTimbre finds the pitch of your voice with the YIN method, a well-tested way of finding how often a sound repeats. Pitch is shown as the nearest note and how far from it you are, in cents. One cent is a hundredth of a semitone.
note = 69 + 12·log₂(f ÷ A4)
Harmonics
A voiced sound is made of a fundamental pitch plus harmonics at whole-number multiples of it. If you sing at 110 Hz, you'll see peaks near 220, 330 and 440 Hz. MindTimbre labels these H2, H3, H4 so they aren't mistaken for something else. Evenly spaced peaks are what any voiced sound looks like.
Spectrum settings matter
The shape of a peak depends on how the spectrum is calculated: the frame size, the window and the averaging. MindTimbre shows these settings and lets you change them, so you can see which features come from your voice and which come from the analysis.
Intervals and octaves
Intervals are ratios. A perfect fifth is 3/2, a fourth is 4/3, an octave is 2/1. MindTimbre shows each as an exact whole-number ratio or in equal temperament, as on a piano. Any frequency can be halved or doubled into a common reference octave of 16–32 Hz so values from different octaves can be compared.
The ratio table
The ratio table, sometimes called the Pythagorean table or Lambdoma, lists every ratio of two whole numbers. MindTimbre uses the 16-based version: row r and column c hold (16 + c) ÷ (16 + r). The top row climbs in sixteenth steps to the octave; the left column falls the same way. The diagonal between 16/24 and 24/16 runs from the fourth to the fifth. Try the interactive Lambdoma matrix →
Pattern Lab and chance
Give Pattern Lab a set of frequencies and it looks for four kinds of pattern: values that are the same note, pairs a chosen interval apart, pairs where the fourth of one equals the fifth of the other (which turns out to mean they are a whole tone apart), and differences that repeat.
Then it runs the same search on hundreds or thousands of random sets from the same frequency range. If your set has no more matches than random ones, the pattern is likely coincidence. Results are seeded, so the same input always gives the same answer.
What MindTimbre doesn't do
- It doesn't convert molecular weights into frequencies.
- It doesn't link notes or frequencies to nutrients, organs, toxins, pathogens or emotions.
- It doesn't give diagnoses, health scores or treatment frequencies.
We've looked closely at these ideas, and we haven't found evidence that would let us present them as measurements.
Open specification
The full engine specification, with test values, is available on request. Anyone can check that MindTimbre's numbers are calculated the way this page says.