Audio to MIDI free: what the top tools actually return

We fed one fixed 8-note piano clip — C4, D4, E4, G4, A4, G4, E4, C4, spaced 0.40 s apart — to the tools that actually rank for “audio to midi free” and timed every one that let us reach a download. Two findings, both measured rather than quoted. First, not one tool returned the 8 notes we put in: the three that completed returned 15, 16 and 16 notes, every one of them over-counted. Second, four of the seven never reached the download at all — one needs a paid account, one a Google login, one sits behind an anti-bot wall, and one needs GPU hardware this test machine does not have. “Free to run” and “free to actually get a file out” turn out to be two different things. All three completion times below were measured on the same machine in the same browser, so they are directly comparable to each other.

1. How this was measured, and what “correct” means

The test clip is folk-phrase.wav, 4.0 s at 44.1 kHz, a single piano line of eight notes. Because we wrote the clip, “correct” is not a matter of taste: the answer is exactly 8 notes, and any tool that returns 8 is right while one that returns 15 is wrong in a specific, countable way.

Each tool was driven in a real logged-in Chrome on this machine, uploading the same WAV, clicking its convert button, and waiting for the .mid to become downloadable. We recorded the wall-clock time from upload to download and parsed the .mid for note-on events (velocity > 0) to count what came back. The three completion times were all measured in the same software-rendered browser, so they are directly comparable to each other; they reflect this machine, not a GPU workstation. The note counts are the honest part, and they do not depend on hardware at all.

2. Every tool that finished over-counted

Eight notes went in. Nothing came back at eight. The three tools that completed returned 15, 16 and 16 notes. This is not a one-off — it is what happens when a single piano note is read as a fundamental plus one or more octave-and-harmonic ghosts, and all of them are written into the file as separate notes.

This is not a reason to write the tools off — it is a reason to know what you are getting. A doubled count mostly means an octave above the real note, which is easy to spot and delete in a piano roll. What matters more for a single instrument is whether the right note is present at the right time, and on this clip all three that finished kept the real line intact; the clutter was addition, not substitution.

3. Four never reached the download

Four of the seven tools could not hand over a file on a fresh machine, each for a different, verifiable reason — not a crash on our side but a gate on theirs.

On a browser where you are logged in, or hold a paid account, three of these four would very likely complete. We are not claiming they are bad — we are recording that, on a fresh machine, “free” did not get us a file. That is exactly the information the phrase “audio to midi free” is asking for.

4. What “free” does and does not buy you

The one thing every tool we tested had in common: uploading and converting costs nothing. The thing that separated them was the last step. Free to run is not the same as free to finish — a tool that shows you a preview and then asks for a card, or a login, has taken your time and kept your file out. When a page says “free”, the question that matters is not “can I upload” but “can I download”.

The second thing “free” does not buy is accuracy. The cleanest single line we measured still added one ghost note for eight real ones — and that was the best case. The worst roughly doubled the count. A free tool saves you money; it does not save you the cleanup pass.

5. How to check any tool on your own file

If you only remember one number from this page, make it this: the truth was 8 notes, and nothing returned 8. That means you cannot judge a free tool by its marketing page — you have to run a clip with a known answer and count. The two-minute check:

1. Use a clip with a known answer. Five to ten notes you can hum, with a beat between them. If you do not know the right answer, you cannot tell whether a tool is right.
2. Count what comes back. Open the .mid and count the notes. More notes than you played usually means octave duplicates and harmonic ghosts — deletable, but you should know they are there.
3. Check the download step before you invest time. On this test, four of seven tools would not hand over a file on a fresh machine. Try the download before you build a workflow around a tool.
4. Time it once. A tool that takes 40+ seconds per 4-second clip will hurt on a full song. On our machine the three that finished split clearly into a quick one and two slower ones, which is what matters for a long file.

For the single-instrument case, none of this changes the conclusion that the tools work — every one that finished kept the real line and only added clutter. The difference between them is how much clutter, how fast, and whether you get a file at all.

Questions people actually ask

Why did every tool return more notes than I played?

Because a single piano note is not one frequency — it is a fundamental plus a series of harmonics, and most tools write the strong ones (often an octave up) as separate notes. Eight notes went into our clip and the three tools that finished returned 15, 16 and 16. The real melody was still there; the extra notes were duplication, not a wrong pitch.

Is a free audio to MIDI converter actually free?

It depends on what you mean by the download. Every tool we tested let you upload and convert for free, but four of the seven could not hand over the MIDI on a fresh machine: one wanted a paid account, one a Google login, one sat behind an anti-bot wall, and one needed GPU hardware this machine lacks. Free to run is not always free to finish.

Do I need to sign up to use a free converter?

For the tools that completed, no account was required to reach the download. The blockers were the other four: a paid account, a login, an anti-bot challenge, or a hardware requirement — not an optional account.

How do I know which tool is accurate?

You do not, until you run a clip with a known answer and count. On our 8-note clip the tools returned 15, 16 and 16 notes — none returned 8. Run your own melody through the candidates, count the output, and pick the one nearest to the number of notes you actually played. Then plan to delete the ghosts.

Is a free tool good enough for a single instrument?

Usually yes, if you accept the cleanup. Every tool that finished our test kept the real line intact and only added clutter, and clutter is cheap to delete in a piano roll. The real risk with “free” is not wrong notes — it is a tool that never hands you the file at all.

Related guides

• Monophonic vs polyphonic: what each mode actually misses — the same clip run both ways, chord-by-chord.
• How many notes a converter can hold at once — the polyphony ceiling and the harmonic rule.
• Why MP3 loses notes that WAV keeps — the FFT reason, measured.
• What is actually inside a MIDI file — the bytes, offset by offset.
• Convert a file now and read the CSV — every number on this page can be checked against your own export.
• Audio to MIDI converter — How accurate a conversion really is
• Audio into MIDI — Where the note boundaries land
• MP3 to MIDI converter — How loud a note has to be before it hits velocity 127
• WAV to MIDI converter — What bit depth actually changes