FL Studio audio to MIDI: what every DAW's built-in capture shares, and where a separate converter wins

The short answer: FL Studio, Reaper, Cubase, Audacity and GarageBand all ship a way to turn audio into MIDI, but every one of those built-in paths is, at heart, a single-line pitch follower — and a separate converter earns its place the moment you need polyphony, a checkable note list, or a result you are not already inside that DAW to get. We measured the limits that apply to all of them on our own detector (dsp.js): a readable pitch window of 55–1800 Hz in monophonic mode and up to 4000 Hz in polyphonic mode, a hard ceiling of six simultaneous notes per 11.6 ms frame, and a timing floor at the 60 ms shortest-note default. A clean eight-note phrase came back as exactly eight distinct pitches in polyphonic mode but only two in monophonic mode on the same take; a seven-note chord returned six notes because the analyser holds at most six per frame. So the question is never “can the DAW do it” — it is “what will the capture actually contain.”

The five DAWs, and the one thing they have in common

Each of the host-software searches — FL Studio audio to MIDI, Reaper audio to MIDI, Cubase audio to MIDI, Audacity audio to MIDI and GarageBand audio to MIDI — is really the same question wearing five different menus. Here is the rough shape of what each built-in path is:

Read that column on the right and the pattern is unmistakable. Four of the five follow one pitch at a time; the other two do not capture audio to MIDI inside the app at all and ask you to export the file. None of them is built to hand you, say, a verifiable list of every note with its timestamp. That is not a flaw in any DAW — it is what a built-in capture is for. The rest of this page is about the limits that follow from that design, measured on our own converter because it is the one we can measure.

The limits that decide every capture, measured here

Because each DAW's capture is a pitch follower, the same three boundaries show up everywhere. We ran them on this site's detector (dsp.js) at the default sensitivity of 60, on synthesised tones whose correct answer we already knew:

The first row is the one worth sitting with. The same eight-note phrase, played once, returned all eight distinct pitches in polyphonic mode and only two in monophonic mode. Monophonic mode is a strict single-pitch tracker: on a quick melody it latches onto a couple of the strongest pitches and lets the rest go. So even inside the DAW, the “best with monophonic material” warning you see in one manual is the same physics the other four are fighting. The pitch window itself is fixed: 55–1800 Hz in monophonic mode and up to 4000 Hz in polyphonic mode, so a sub-bass line or a very high piccolo is outside the search on every one of these captures.

A chord collapses, and there is nothing the DAW can do about it

The six-per-frame ceiling is the second universal limit, and it is why a chord is the thing every built-in capture handles worst. We fed a six-note chord and a seven-note chord through polyphonic mode. The six-note chord filled the ceiling exactly — six in, six out. Adding one more note, a seventh, did not give seven: it gave six, because the analyser will not hold more than six pitches in any single frame, and the extra voice gets merged away. A tight chord is exactly where overlapping octaves and close voicings get collapsed first.

This is not a shortcoming of one DAW's algorithm versus another's. It is the arithmetic of following pitches in short windows. Whatever the menu says, a capture that reads one or a few pitches per moment will return the skeleton of a chord, not the full voicing — and on a wide cluster it will over-read, accumulating extra pitches as the frames drift. If your goal is the actual harmony, no built-in capture gives it to you; you export the audio and run a converter whose polyphonic mode is built to hold several peaks at once, then clean up.

The timing floor that no sensitivity slider lifts

The third limit is about speed, not pitch. On this converter the shortest note defaults to 60 ms, and notes shorter than that do not register at all. We played the eight-note phrase again with the notes only 50 ms apart and monophonic mode returned zero notes — each note was simply too brief to clear the gate. The same is true of any built-in capture: there is a shortest event it will write, and a run of fast notes under that length collapses into nothing or into one smeared note. This is the part people least expect, because the pitch was right and the playing was clean — the capture just could not keep up with the rate.

Two numbers frame the whole timing picture. The analyser reads audio in frames 11.6 ms long (about 86 frames per second), and the default shortest note is 60 ms — so the practical ceiling on how fast distinct notes can arrive is set by that 60 ms, not by how fast you played. A DAW's own audio-to-MIDI has its own equivalent floor, and none of them is removed by turning a knob. When the timing matters, slow the part down, or record it cleanly enough that the capture can resolve each attack.

Velocity, and the one number that tells you when a capture is trustworthy

One more shared behaviour: the loudness-to-velocity mapping. On this converter velocity is round(20 + (dB + 60) × 100 / 54), so a note hits the maximum of 127 at −2.22 dBFS and a quiet note drops toward 20. A built-in capture has its own curve, but the shape is the same — velocity is a relative reading of loudness, not a measured performance dynamic. Where this matters is verification: a capture gives you a region or a clip you play, but you cannot open it and ask “did it get this note?” line by line.

That is the gap a separate converter fills. Ours writes the same Standard MIDI file (format 1, two tracks, one channel, no instrument-label fields) and a CSV with start, end, MIDI note, note name and velocity to four decimal places. On 11 known-answer clips it measured 86.7% recall and 91.5% precision. The honest move on any of these five DAWs is the same: use the built-in capture for a quick idea inside the project, and export the file to a converter when you need to know what the audio actually contained.

When to use the DAW, and when to export the file

Audacity and GarageBand are the easy cases: they do not capture audio to MIDI inside the app at all, so the answer is simply to export the recording and convert it. FL Studio, Reaper and Cubase each have a capture worth using on a single line — but the moment the source is a chord, a dense mix, or something you must verify, exporting the file is the move. What that exported file should look like is its own question: what exporting actually changes measures the region length, level, channels, silence and fades that decide the notes once the audio reaches a separate converter, including the four hosts covered there.

How to do each

In the DAW

Run the built-in capture (Edison Convert to score in FL Studio, Create MIDI from selection in Reaper, VariAudio in Cubase). Keep the original audio — the MIDI is a copy. For anything polyphonic, expect a single moving line and plan to rebuild the harmony by hand.

Export and convert separately

Drag the region to the desktop to get the audio file (or use the DAW's Export Audio), then drop it into a converter. Pick Mono for a single line or Polyphonic for chords — on this converter, polyphonic mode holds several peaks per frame while mono follows one pitch at a time. Remember the file is summed to mono before analysis, so a hard-panned part loses 6.02 dB of loudness, which drops its velocity by 11 steps — something no mode or sensitivity setting recovers.

Once a separate converter hands you that .mid, the work is not finished — Getting an audio-to-MIDI result into FL Studio covers the import itself and the one step that surprises everyone: the file carries no instrument, so you have to load a generator before you hear anything.

This page compares the built-in audio-to-MIDI captures of five DAWs with a separate converter. It is not affiliated with any DAW maker. The measured limits are from this site's own detector, run in Node — try the converter itself on the home page: drop a file, get a MIDI and a CSV, nothing uploaded.

Related guides

• Ableton audio to MIDIWhen Live's converter is enough, and when to export
• Logic Pro audio to MIDIWhat the built-in capture can and cannot hold
• Other DAWs audio to MIDIWhich export settings change the notes, measured on this converter
• Mono vs polyWhat each detection mode misses
• Microphone to MIDIWhat this converter does with a recording
• Piano to MIDIWhere the keyboard falls outside the detector
• Audio to MIDI pluginWhen to install one, and when the browser is enough
• MIDI converterWhat is actually inside a MIDI file
• Clean up a converted MIDIThe 5 passes that turn a raw export into a usable part