Audio to MIDI Converter

What converting audio to MIDI means

Audio and MIDI store different things: a recording holds a waveform, while a MIDI file holds instructions — which note begins, when it ends and how hard it is struck. Converting between them is transcription, not a change of file format. This tool reads short overlapping slices of the waveform, estimates the dominant pitch in each slice, and groups steady pitches into notes. Because it tracks a single pitch at a time, the result is a monophonic reading of the most prominent line in the recording, not a full arrangement with every instrument recovered.

Which sources you can start from

Any file your browser can decode can be loaded: compressed formats such as MP3, M4A, AAC or OGG, and uncompressed ones such as WAV or FLAC. The container matters less than what is inside it. A clean solo line — one voice, one instrument, one melody — gives a usable transcription; a dense mix does not, because several pitches compete inside the same slice and the estimate jumps between them. Very long recordings are trimmed to the opening minutes so the analysis stays responsive, so keep the passage you care about near the start.

Humming and singing

A hummed or sung phrase is the most natural way to capture a melody before you forget it, and it is close to one line already. It is also the least pitch-stable source. Vibrato and slides keep the estimate moving around the intended note, consonants and breath noise carry no clear fundamental, and a soft or overly breathy tone may not cross the level needed to register as pitched at all. Held, even notes with a little space between phrases convert far more predictably than fast, ornamented singing.

Phone and field recordings

A voice memo is convenient but rarely neutral. Phone microphones apply automatic gain control and noise suppression, which reshape the signal before it is saved; a quiet room is not the same as a treated one, and reverb smears the boundary where one note becomes the next. Distance from the source, handling noise, wind and background speech all add energy that is not part of the melody. Recording close, still and in a quiet space improves the result more than any setting here can.

Audio extracted from video

Soundtracks pulled from video carry their own problems. They are usually already mixed — music under speech, effects on top, several pitched and unpitched layers at once — so a single-pitch tracker follows whichever layer is loudest at that moment rather than the tune you want. They are often loudness-normalized and compressed again during export, which flattens the attacks that help place note starts. Extract the audio first, then trim to a passage where the melody is exposed; if the clip has dialogue or a full backing track, no converter of this kind will separate it for you.

Looking only at uncompressed sources? The WAV to MIDI page explains how that format affects detection, and the original MP3 to MIDI converter is the file-type-first starting point.