How to turn your voice into an instrument in 3 steps
- 01
Record or open a hummed melody
Record a melody with your microphone - humming, singing, or whistling one note at a time - or use Open file to load an existing recording.
- 02
Pick an instrument and shape the notes
Choose from 8 instruments, then adjust quantize, snap to scale, and octave shift to clean up the detected melody until the piano roll reads the way you intended.
- 03
Render preview, then export
Click Render preview to hear the instrument version. When it matches your melody, use Download audio for the rendered sound, or export the standard MIDI file to open in a DAW.
What you can do with voice to instrument
- Detect a monophonic melody from humming, singing, or whistling using YIN pitch tracking, a classical pitch-detection method rather than a generative model.
- Play the detected melody back on 8 instruments - piano, electric piano, synth lead, strings, flute, bass, bell, or organ.
- Quantize detected notes to 1/8 or 1/16 at your chosen tempo, or leave quantize off to keep the original timing.
- Snap detected pitches to a scale - off, major, or minor, in any key - to clean up notes that drifted slightly off pitch.
- Shift the whole melody up or down by up to 2 octaves without re-recording.
- View detected notes in a piano roll, so you can see exactly what pitches and timing were picked up from your voice.
- Export both a rendered audio file and a standard MIDI file, ready to drop into any DAW for further composing or production.
What is voice to instrument?
Voice to instrument takes a hummed, sung, or whistled melody and turns it into notes an instrument can play - piano, strings, synth, and more - along with a piano-roll view and a MIDI file you can drop into a DAW. It's a way to capture a melody idea the moment you think of it, using only your voice, without touching an instrument or writing notation by hand.
The pitch detection behind it is YIN, a well-established, classical algorithm for finding the fundamental pitch of a single voice over time. It isn't a generative or AI model composing anything - it's tracking the pitch you actually hummed, moment by moment, and converting that into note data. That distinction matters if you're comparing tools: some voice-to-note products lean on generative models to guess at intent, while this one simply reports the pitch it measured.
Recording tips for cleaner detection
Because detection is built for one voice at a time, how you record has a direct effect on how clean the result looks in the piano roll.
- Hum with a clear, sustained syllable like "doo" or "dah" rather than a closed-mouth hum - a more open, resonant sound gives the pitch tracker more to work with.
- Sing or hum one note at a time. Chords, harmonies, or two notes sung together will not be tracked correctly - this only follows a single melodic line.
- Record somewhere quiet. Background noise, music playing nearby, or a noisy room can confuse pitch detection the same way it would confuse a human ear.
- Keep a steady pace. A rushed or uneven melody is harder for the algorithm and for quantize afterward to make sense of cleanly.
Cleaning up the detected melody
A first pass at detection rarely comes out perfectly quantized or perfectly in tune, and that's expected - humming naturally drifts in pitch and timing more than a played instrument does. The shaping controls exist specifically to fix that.
| Control | What it fixes | Range |
|---|---|---|
| Quantize | Snaps note timing to a rhythmic grid | Off, 1/8, or 1/16 |
| Snap to scale | Pulls note pitches to the nearest scale tone | Off, major, or minor + key |
| Octave shift | Moves the whole melody up or down | -2 to +2 octaves |
| Instrument | Changes the sound the melody is played with | 8 options |
Limitations: monophonic input, one voice at a time
This is a monophonic tool, meaning it tracks a single pitch at a time. Humming a chord, singing harmony with yourself, or recording two people at once will confuse the pitch tracker rather than produce two separate melodic lines - there's no way to separate simultaneous pitches after the fact.
Since detection follows your actual pitch, timing, and phrasing, the rendered instrument version will sound like your voice's melody, not like a performance by someone who plays that instrument for a living - expressive details like breath, vibrato timing, and slight pitch wobble carry through in ways a played instrument line wouldn't have. Quantize and snap to scale clean this up, but they don't make the result indistinguishable from an instrument someone actually played.
Voice to instrument vs. the vocal tuner
Both tools use pitch tracking on a monophonic voice, but they do different jobs with it. The vocal tuner corrects the pitch of your vocal recording itself - you still hear your voice, just tuned to a key. Voice to instrument instead uses the detected pitch to drive a completely different sound - piano, strings, synth - so what comes out is an instrument, not your voice.
Use the vocal tuner when you want your own voice, pitch-corrected. Use voice to instrument when you want a melody idea captured as notes and turned into an instrument part, with a MIDI file you can take into other software.
Tips for better voice to instrument results
- Hum on an open syllable like "doo" or "dah" instead of humming with a closed mouth - it gives the pitch tracker a clearer signal to follow.
- Sing one note at a time. Chords and harmonies aren't supported - this tracks a single melodic line only.
- Record in a quiet space without background music or noise competing with your voice.
- Try snap to scale first if notes look slightly off in the piano roll - it's often a faster fix than re-recording.
- Use quantize if your original melody's timing was uneven and you want a cleaner, more rhythmic MIDI file.
- Check the piano roll before exporting MIDI - it's the most direct way to confirm the detected notes match what you actually hummed.
- Shift octaves after recording rather than trying to hum unusually low or high - it's easier to get clean detection in your natural range.
Frequently asked questions
Is voice to instrument free to use?
Yes. There's no sign-up and no limit on how many melodies you can record and convert. Everything runs in your browser, with no account required.
Does my recording get uploaded anywhere?
No. Pitch detection, instrument rendering, and MIDI export all happen on your device. Your recording is never sent to a server.
Is this AI?
No. Pitch detection uses YIN, a classical signal-processing algorithm for finding the pitch of a single voice - the same kind of method used in tuners and pitch-tracking software for decades. It isn't a generative model composing or guessing at a melody; it's tracking the pitch you actually hummed.
What should I hum, sing, or whistle for a clean result?
An open syllable like "doo" or "dah," sung one note at a time, in a quiet space. Avoid chords, harmonies, or humming with a closed mouth - all of these make detection less reliable.
How many instruments are available?
8: piano, electric piano, synth lead, strings, flute, bass, bell, and organ. Switch between them freely to hear the same detected melody played with a different sound.
Can I record chords or harmony?
No. Detection is monophonic, meaning it follows a single pitch at a time. Humming a chord or two notes at once will not produce two separate melodic lines - only a single-voice melody works reliably.
Can I export a MIDI file?
Yes. Alongside a rendered audio version, you can export a standard MIDI file of the detected melody, ready to import into a DAW for further arranging, layering, or production.
What's the difference between the audio and MIDI downloads?
The audio export is a rendered recording of the melody played on your chosen instrument. The MIDI export is note data - pitch, timing, and velocity - that you can open in a DAW and reassign to any instrument or virtual sound you have there.
Why doesn't the result sound exactly like a real player?
Because it's tracking your actual voice's pitch, timing, and phrasing rather than generating an idealized performance. Small pitch wobble and timing quirks from your humming carry through - quantize and snap to scale help clean this up, but the result reflects what you actually sang.
Does this work on mobile, including iPhone?
It runs in a mobile browser, though microphone quality and background noise can be harder to control on a phone. For a cleaner detected melody, a quiet room and a decent microphone matter more than which device you use.
What's the difference between this and the vocal tuner?
The vocal tuner pitch-corrects your actual voice recording - you still hear yourself, tuned to a key. Voice to instrument uses your pitch to drive a different instrument sound entirely, so the output is piano, strings, or another instrument, not your voice.
Can I fix notes that came out off-pitch or off-beat?
Yes. Snap to scale pulls pitches toward the nearest note in a chosen key, and quantize aligns timing to a rhythmic grid at 1/8 or 1/16. Between the two, most minor detection drift can be cleaned up without re-recording.
Can I use the exported MIDI file in any DAW?
Yes. The export is a standard MIDI file, which any modern DAW can import. From there you can reassign the melody to any virtual instrument the DAW offers, layer it with other parts, or edit individual notes with the DAW's own piano roll.