The short answer: a studio take almost always transcribes cleaner than a live recording, because a live room adds two things a studio isolates away — reverb (your sound bouncing off walls and coming back a fraction of a second later) and stage bleed (other instruments and the room tone underneath you). In our own measurements with this site's detector, a fast 8-note line (C–C) came back 8/8 from a dry studio take in both modes, but from a live room it came back 5/8 in monophonic mode and 14 notes in polyphonic mode — 8 real ones plus 6 reverb artefacts smeared into the gaps. So: when you can, record close and dry. When you cannot avoid a live take, use polyphonic mode (it keeps the real notes) and read the piano roll, because the room mainly adds mess rather than deleting your playing.
What a room actually does to the file
A studio take and a live take are not the same recording with different labels. The moment you play in a room, two physical things change the waveform before any converter sees it:
- Reverb (reflections). Your note leaves the instrument, hits a wall, and a quieter copy returns. The converter does not hear one clean note — it hears the note followed by a decaying tail of its own echoes. That tail bleeds energy across the silent gaps between notes.
- Stage bleed. In a live setting there is almost always something else sounding: a bass amp, the drummer, the PA, the crowd. That low, continuous sound sits under your line the whole time.
The detector scores pitch from the harmonic structure of each short frame. Reverb does not change your pitch, but it fills the gaps the detector uses to tell one note from the next; bleed adds a second pitch the detector may score as if it were part of your music. Neither is a codec error — it is the room, baked into the audio.
What we actually measured: a fast run, dry vs live
Numbers you cannot reproduce are worthless, so we ran it on this site's own pitch detector (dsp.js). We synthesised a fast 8-note line (C4 D4 E4 F4 G4 A4 B4 C5, about 0.14 s per note with tight 12 ms gaps) and made two versions: a dry studio take (the line alone) and a live take (same line convolved with a ~1.1 s reverb tail plus a low ambient noise floor). Both ran at the default settings:
| Take | Mode | Notes detected | Real notes recovered | Extra (artefact) notes |
|---|---|---|---|---|
| Studio (dry) | Polyphonic | 8 | 8 / 8 | 0 |
| Studio (dry) | Monophonic | 8 | 8 / 8 | 0 |
| Live (room) | Polyphonic | 14 | 8 / 8 | 6 |
| Live (room) | Monophonic | 5 | 5 / 8 | 0 |
The pattern is the whole point of this page. The reverb tail did two opposite things at once: in polyphonic mode it left all eight real notes in place but invented six extra ones in the gaps (the echoes looked like short notes), while in monophonic mode it swallowed three of the fast notes entirely — the line and its echo blurred into one continuous blob the single-pitch tracker could not split. A studio take showed neither problem.
Two distinct ways the room makes a mess
1. Boundary smear (polyphonic mode gains artefacts)
Polyphonic detection looks for several spectral peaks per frame and groups frames that share a pitch into notes. A reverb tail keeps your note's spectrum alive through the gap before the next note, so the detector often scores those echo frames as a short extra note. Your eight real notes survive — but the piano roll now shows six ghosts you have to delete. The more reverberant the room, the more ghosts.
2. Swallowed fast notes (monophonic mode loses notes)
Monophonic detection follows one pitch at a time. When a note's echo overlaps the next note, the tracker cannot tell where one ends and the next begins, so it holds one pitch straight through. On a slow ballad you may not notice; on a fast run the missing notes add up fast — five of eight in our live test. This is why a live single line is safer in polyphonic mode than in monophonic mode.
Mono or poly for a live take
For a live recording of one line — one voice, one horn, one guitar — reach for polyphonic mode. Our table is the reason: polyphonic kept all eight real notes in the live room while monophonic dropped three. Polyphonic trades cleanliness for completeness: it will return a few reverb ghosts, but those are easy to spot and delete on the piano roll. Monophonic, by contrast, silently loses the notes. The same rule shows up on plain room noise too — see microphone to MIDI for the SNR measurements behind it.
How to record a live take that converts cleanly
- Get close to the source. A close microphone gives your note a stronger level over the room's reflections, which is exactly the signal-to-echo ratio the detector needs.
- Cut the reverb. A smaller, drier space keeps note boundaries sharp. A tiled hallway or a stone church will smear more than a carpeted bedroom.
- Kill the bleed. Isolate your instrument. A vocal in headphones, an amp in a separate room, or a tight cardioid pattern all reduce the other sound the detector would score as music.
- Do not clip. A distorted peak flattens the harmonic structure the detector scores, which hurts in a room as much as in a studio.
- Mono is fine. The converter averages every input to mono before analysis anyway, so a stereo live recording buys you nothing here.
- Slow the line down if you can. The swallowed-note problem is a fast-note problem. A live take with roomier spacing converts far more completely than a machine-gun run.
How to recover a live recording you already have
You cannot un-ring the room, but you can stop the converter from trusting it. Drop the file in and let the piano roll draw, then:
- Use polyphonic mode so the real notes are not silently dropped.
- Lower the sensitivity a step or two. The reverb ghosts sit lower in the mix than your note; a lower threshold drops many of them without touching the line.
- Download the .csv and open it in a spreadsheet. The columns are
index,midi_note,note_name,start_sec,end_sec,duration_sec,velocity. Sort bystart_secand delete the short, low-velocity artefacts the reverb left in the gaps. - Check the piano roll, not the note count. A live take can return the right number of notes and still be full of ghosts — the picture is the truth.
Questions
Is a studio take always better than a live one?
For conversion input, yes. A dry, close-mic'd studio take has no reverb tail and no stage bleed, so the detector sees clean note boundaries. In our measurement the studio take came back 8/8 in both modes while the live room lost three notes in mono and added six artefacts in poly. The gap is real, not cosmetic.
Does reverb delete my notes?
Not usually — it mis-files them. In polyphonic mode the real notes survive but the echo tail becomes extra short notes you delete; in monophonic mode a fast passage can blur into one held pitch and drop notes. Slow, separated playing survives a room far better than a fast run.
Can I just clean up the MIDI afterwards?
Yes, and for a live take you usually should. Convert in polyphonic mode, lower sensitivity to shed the reverb ghosts, then trim the short low-velocity artefacts in the .csv. The notes you played are almost all still there; the room mostly adds mess on top.
Does this mean I should re-record in a studio?
Only if the live take is unusable. A close, dry live recording converts well enough for most purposes — the measurement above used a heavy ~1.1 s reverb, heavier than a typical small-venue take. The practical rule is simply: the drier and closer, the cleaner, and polyphonic mode is the safe choice when you cannot control the room.
How is this different from plain room noise?
Room noise is a steady hiss that the detector's voice gate can ignore between notes. Reverb is your own note delayed, so it lands inside the gaps and fools the gate — that is why reverb smears boundaries while steady noise mostly just lowers the floor. The two often travel together in a live take, which is exactly the case measured above.
Convert a file now — runs in your browser, nothing is uploaded.
Related guides
- Microphone to MIDIWhat this converter does with a recording
- Mono vs polyWhat each detection mode misses
- MP3 to MIDIWhy MP3 loses notes that WAV keeps
- Voice to MIDIWhat a single-line converter can follow
- Guitar to MIDIWhat a bend does to your notes
- Piano to MIDIWhere the keyboard falls outside the detector
- Vocal to MIDIWhy a sung note comes back an octave up
- Humming to MIDIWhat wordless vocals come back as
- Drums to MIDIWhat a pitch detector does with no pitch to find