M4A to MIDI
Choose your M4A file in Melody Lab, set the melody's pitch range, and select Convert to MIDI. Preview the notes and download the .mid file. If the recording starts with silence, check the exported note positions before aligning it with other tracks.
Single-note melodies · one file · up to 25 MB and 3 minutes · local audio processing
Start with an isolated melody
This converter detects one pitch at a time. It does not separate a full mix into instruments, transcribe spoken words or create sheet music. Use a clear solo melody and review the result before moving it into your music project. The MIDI contains notes rather than the original recorded sound.
A pause before the melody is part of the input's timeline. It is worth checking where the first exported note lands instead of assuming every download begins immediately or that its timing exactly matches the recording. We tested that question with the paired files below.
Our test: the same tone with and without a lead-in
On September 26, 2026, we generated two mono AAC-encoded M4A files locally. Both contain a two-second, 440 Hz synthetic tone at a source sample rate of 22,050 Hz. The second version adds 1,000 milliseconds of silence before the tone. These are controlled diagnostic inputs, not voice memos or a test of singing accuracy.
We selected each file in the live converter using desktop Chrome. Lowest note stayed at C3, Highest note at C6, Detection confidence at 0.85, Quiet-sound filter at −45 dB, Shortest note at 100 ms and Transpose at 0. Both runs displayed one A4 note and offered a MIDI download. We saved both downloads and read their note-event times.
| Input | Browser duration | First MIDI note starts |
|---|---|---|
| No added lead-in | 2.04s | 0 ticks / 0 seconds |
| One-second lead-in | 3.02s | 892 ticks / about 0.929 seconds |
The durations above are what the browser displayed after decoding, not the durations requested from the test generator. We retained that distinction instead of replacing the observed values with the intended two and three seconds.
Both downloaded MIDI headers use 480 ticks per quarter note, and their tempo events specify 500,000 microseconds per quarter note. The conversion is seconds = ticks × 500,000 ÷ 480 ÷ 1,000,000. For the delayed note, 892 × 500,000 ÷ 480 ÷ 1,000,000 = 0.929166… seconds. The onset difference between these two exports is therefore about 0.929 seconds, not exactly the one second added to the source.
This pair shows that the delayed result did not simply move its first note to zero, but it also does not justify a promise of exact alignment. We have not isolated the cause of the onset difference in this experiment. Check your own output rather than applying this measured difference as a universal correction.
Repeat the check before aligning a recording
- Download the two M4A examples above. Load the version without a lead-in, keep the listed controls, convert, and download its MIDI before selecting the other file. A new input replaces the current conversion result.
- Repeat with the lead-in version. Compare the waveform and piano roll, but use the exported MIDI when checking the time precisely. Our immediate MIDI and delayed MIDI preserve the actual downloads.
- Compare the first note's position in your music editor with the source audio. If you need to remove a count-in or pause, work on a copy in an audio editor or move the MIDI notes in your music project. Melody Lab has no dedicated trim or alignment control.
- Keep both the original file and your edited copy. Changing the beginning changes the timeline you are comparing. Reconvert the edited audio and inspect that result rather than assuming a previous offset still applies.
The generation record and input hashes identify our exact samples. The MIDI timing record preserves raw ticks, tempo, resolution and calculated seconds. These measurements apply to this pair of synthetic AAC files and these settings, not every M4A encoding or recording.
Can I use an M4A voice recording?
Try a decodable recording with a single sung melody. Spoken words do not map to the intended musical notes, and overlapping voices or accompaniment can confuse detection. Review the voice-to-MIDI control guide for this site's pitch-range and filter workflow. The timing experiment here used a synthetic tone, not a human voice.
What if loading fails or no notes appear?
Read the displayed status. A decoder error and a completed conversion with no notes are different outcomes. Check the file size and duration, then whether the source plays and contains an audible melody within the selected pitch range. Range changes require another conversion. Our loading versus detection example demonstrates that distinction with a separate format.
Where does processing happen?
Audio decoding and note detection run in your browser without uploading the recording. The site uses Google Analytics for visit statistics. Download results before closing or reloading the page; it does not keep conversion history. Output is one melody track in MIDI, not PDF or MusicXML.