WAV to MIDI
Short answer
No, and the reason takes two minutes to see. A WAV file holds waveform samples, tens of thousands of amplitude readings for every second of sound. A MIDI file holds performance instructions: which note, when it was pressed, how long it was held.
Nothing in the first file can be re-wrapped into the second, because the two share no container. What this page does with a WAV is transcribe it.
The tracker estimates notes from the waveform, and the file you download is that estimate written out as SMF format 1 at 480 tpqn.
That distinction is not pedantry. A conversion is deterministic, and an estimate is not.
This page carries Sensitivity, Note range, Min note length and Quantize controls. Moving Sensitivity alone changes how many notes come out. A format converter has no sensitivity slider, because a conversion has exactly one right answer.
So the question that decides your result is not whether WAV converts better. It is how many things are sounding at once inside the file. One part at a time is what changes the outcome.
Try it on a WAV Runs in your browser, no upload, no sign-up.
Why a WAV cannot be converted into MIDI
Start with what each file is for. A format conversion stays inside one domain: PNG to JPG, MP3 to FLAC, MIDI format 0 to format 1. The input and the output describe the same kind of thing.
That is why the transformation is mechanical and repeatable. Run it twice on the same file and you get the same bytes.
WAV to MIDI crosses domains. The WAV side is a recording of pressure over time. The MIDI side is a list of instructions for a performer.
There is no shared container, so there is nothing to move. That leaves exactly one route from one to the other, and it is not conversion.
Something has to look at the waveform and decide which notes were played. That decision is an estimate, and the rest of this page is about living with it.
The file size gives the game away. A real conversion scales with duration, because five minutes of audio becomes five minutes of audio in another wrapper.
A MIDI file has no notion of seconds inside it. A long silent stretch costs a few bytes, one large time value. Every single note costs a note-on event and a note-off event.
In this page's own writer each note is two three-byte channel events plus their timing. So the size of the MIDI follows how many notes were estimated, not how long you recorded.
The controls are the proof it is an estimate
If WAV to MIDI were a conversion, the page would need one button and nothing else.
Instead it has Sensitivity, Note range, Min note length and Quantize. Each one exists because the result is not fixed by the input. Why would a converter need a slider called Sensitivity?
Sensitivity sets how strong a pitch has to be before the tracker believes it. Note range sets the window it searches. Min note length discards anything shorter, and Quantize moves what is left onto a grid.
Change Sensitivity and the note count changes. Lower it and quiet false notes appear. Raise it and real notes in soft passages stop being found.
That is the behaviour of an estimator, not of a converter. A conversion has one right answer per input.
This page has a range of plausible answers, and your job is to pick the setting where the estimate matches what you played.
The two numbers on a WAV file do not add notes
A WAV file carries two quality numbers, its sample rate and its bit depth. Both describe how faithfully the waveform was stored. Neither describes how many notes are in it.
A 96 kHz, 24-bit WAV of a full mix is not more transcribable than a 44.1 kHz, 16-bit WAV of the same mix. This build has not run that comparison side by side, so what follows is the mechanism rather than a measured difference.
The build itself settles it. It downsamples the audio before any pitch decision is made, to about 11 kHz in Melody mode and about 22 kHz in Chords mode. A 96 kHz WAV is cut down by roughly nine and a 44.1 kHz WAV by roughly four, so both arrive at about the same working rate.
The tracker looks for a repeating period in the waveform, and it needs only enough resolution to see that period. Extra bits and extra samples describe the same performance more precisely. If the file was recorded at 96 kHz, does that mean more notes were played?
The estimate is limited by how many notes sound at once. So exporting at a higher rate is not the lever people hope it is. If you are choosing an export setting to improve a transcription, you are turning the wrong dial.
Sample rate and bit depth are about fidelity. Transcription is about polyphony. They are different axes, and only one of them is a control you can turn.
What to export from your DAW
The export setting that matters is not the sample rate.
Export one instrument at a time. A stem of the vocal, or the bass, or the lead line is closer to a single voice than a mixdown is, and a single voice is the case this tracker handles well.
On this build's own test audio, a clean single-line take returned 7 to 8 of every 8 notes at the correct pitch. The same tracker on a finished mix returns the loudest part only.
If the part you want is buried, isolate it first. Mute everything else, or bounce that one track on its own. The goal is a file where one pitch sounds at a time as often as possible.
Trim the head and the tail. Silence at the start costs you time and nothing else.
A long reverb tail at the end is worse than useless, because in this build's test reverb split one sustained note into two.
Stereo does not buy you anything either. Before tracking starts, this page averages every channel into one, so a stereo export gives the pitch tracker a single signal. Two channels go in, one comes out, so what did the second one buy you?
How to turn a WAV into MIDI, in order
- Choose the file. WAV is listed as working in the format table on the home page, and the file picker takes it directly.
- Set Note range to the real span of the part. This is the highest-value control on the page, because it stops the tracker hunting for pitches that are not there.
- Leave Mode on Melody. Chords mode reads peaks in the frequency spectrum instead of tracking one pitch, and the page labels it experimental.
- Set Analyse to the first 30 seconds and convert once. Judge that clip before you commit a long file. Anything past that window is trimmed before tracking starts, so it will not be in the result at all.
- Read the status line. It reports how many notes were found, over how many seconds of audio, the pitch range and the tempo. If the pitch range reads unknown, nothing passed the thresholds.
- Adjust and repeat. Raise Min note length to clear clusters of tiny notes. Lower Sensitivity when quiet parts sprout notes nobody played.
Which of those steps do you think costs the most notes when it is skipped?
FAQ
Can a WAV file be converted to MIDI?
Not by conversion. WAV stores samples and MIDI stores note instructions, so the two share no container and there is nothing to re-wrap.
What happens instead is transcription. Turning a WAV into MIDI means analysing the waveform, estimating the notes and writing those estimates to a MIDI file. You get a result either way, and it is an estimate rather than a transformation.
What is the difference between transcription and a format conversion?
A conversion is mechanical and repeatable. The same input always produces the same output, and no interpretation is involved.
Transcription estimates information that is not written down anywhere in the file. It reads a waveform and decides which notes were played, and a different threshold gives a different answer. This page does the second thing.
Does a higher sample rate or bit depth give me more notes?
No. Sample rate and bit depth describe how precisely the waveform was stored, not how many notes are in it.
The tracker needs enough resolution to see a repeating period, and ordinary CD rates provide that. A 24-bit, 96 kHz export of a full mix is still a full mix.
This build reduces the rate anyway. Melody mode tracks at about 11 kHz and Chords mode at about 22 kHz, whatever the file carries. A 96 kHz WAV and a 44.1 kHz WAV both land at about the same working rate, so the extra samples never reach the tracker.
Why does the MIDI file not scale with the length of my WAV?
Because a MIDI file has no samples in it. It is a list of events with time gaps between them, so a long silence is a few bytes while each note is an event with a pitch, a start and an end.
The size of the file follows the note count. That is also the clearest sign that nothing was converted.
Why does my WAV come back with fewer notes than I played?
Almost always because more than one note was sounding at the same moment. This tracker follows one pitch at a time, so where several are present it keeps the loudest and drops the rest.
The count is decided by the recording, not by the file type. Convert one instrument at a time and the count comes back up.
What should I export from my DAW?
One instrument, not the mixdown. A stem of the part you actually want is closer to a single voice, which is the case this tracker handles well.
Trim the silence at both ends and cut any long reverb tail, because reverb makes one note look like two. Sample rate and bit depth can stay at whatever your project uses.
Stereo is not worth fussing over. Every channel is averaged into one before tracking starts, so a stereo export gives the tracker one signal. If the left and right channels carry different parts, that average stacks them together and you are back to the mixdown problem.
Does a WAV carry anything the tracker can use?
No. The tracker reads samples and nothing else. A WAV usually carries no useful metadata at all, and a file that did would not change the result.
Whatever you know about the recording, its key, its tempo, its intended instrument, is not written in the file. It cannot be part of the estimate.
Is my WAV uploaded anywhere?
No. Your browser decodes the file and runs the tracker on your own device, and the MIDI is assembled in memory.
There is no account, no upload, no watermark and no e-mail. You can confirm it yourself: open the network panel, convert something, and watch. Nothing leaves the machine.
How do I convert WAV to MIDI?
You do not convert it, you transcribe it, but the practical steps are short. Pick the file, set Note range to the real span of the part, leave Mode on Melody, and convert a short clip before you commit a long one.
The page returns a .mid file either way. The difference is that the notes are an estimate you can tune with the controls, not a byte-for-byte re-wrap of the original audio.
Analyse is a cut, not a preview. The decoded audio is trimmed at 30 seconds, 60 seconds or 2 minutes before tracking begins, unless you pick the whole file. Notes past that point are absent from the .mid rather than wrong, which is why the status line tells you how many seconds were read.
Is this a WAV to MIDI converter?
In output, yes: you give it a WAV and get a MIDI file back. In mechanism, no, and the controls are the tell.
A converter maps one container onto another and has a single fixed result. This page carries Sensitivity, Note range, Min note length and Quantize, because what it produces is an estimate you steer rather than a fixed transformation.
Judge it on a single part, not on a whole mixdown. Thirty seconds of one instrument tells you more about this tool than any page can, including this one.