A small page that turns a single-line recording into a MIDI file, in your browser.
Easy Audio to MIDI is one page with one job: you drop in an audio file, it listens to
the pitch, and it hands you back a .mid file you can open in a DAW or notation program.
There is no account, no queue, no upload and no server doing the work - the analysis runs as JavaScript
on the device you are holding.
Most "audio to MIDI" pages either send your file somewhere or ask you to register first. This one does neither. The trade-off is that everything has to fit in a browser, which sets the limits below.
For why that matters and what the browser does instead, see MP3 to MIDI without uploading your file.
The page decodes the file with your browser's own decoder, mixes it to mono and resamples to 16 kHz, then walks through it in short overlapping frames. Each frame is scored with YIN pitch detection - a difference function followed by a cumulative mean normalised difference - restricted to lags matching 60-2000 Hz. The first lag that drops below the threshold is treated as the fundamental. Runs of frames that agree on a semitone become one note, and the notes are written out as a standard MIDI file (format 0, one track, 480 ticks per quarter note).
The lag search that produces that 60-2000 Hz window, and what the detector reports at each end of it, is walked through in how YIN pitch detection works.
What that MIDI file actually contains, byte for byte, is read back in what is inside the downloaded .mid file.
Those are limits of the method, not settings you can tune past. Where a paid service offers chord recognition or one MIDI track per instrument, it is running a trained model. This page does not, and does not claim to.
The 80 MB figure is a file-size limit rather than a length limit, so what it means for a file you already have depends on its format - the measured durations are in what the limits actually are.
A clean monophonic line usually comes out usable. Anything denser needs cleaning up in a DAW. The page shows you how many notes it found, how long the analysis took and the pitch range it landed on, so you can judge the result instead of taking a promise on faith.
What does and does not change that accuracy was measured in 128 kbps vs 320 kbps MP3 to MIDI.
How to check a result yourself, and what each failure looks like, is set out in how to check a converted MIDI before you trust it.