How to record a vocal that converts well
| Do | Why it helps |
|---|---|
| Hum or sing "la" or "ah" on clear notes | Consonants and breaths are not pitched and create junk notes |
| Sing one line at a time | Two voices at once are read as chords |
| Leave a short gap between notes | Repeated notes at the same pitch merge into one long note otherwise |
| Keep a steady volume and stay close to the mic | Quiet notes fall below the detection threshold |
| Record in a quiet room, no backing track | Music behind you is transcribed as notes too |
Do not worry about being exactly in tune. The tool records the pitch it hears, so a flat note becomes the nearest note or sometimes the wrong neighbour. If you want a grid, fix the notes in your DAW.
Vocals from a song: remove the music first
If the voice you want is in a finished song, the backing music will confuse the transcription. Run the vocal remover (or the acapella extractor) to get the vocal alone, then drop that stem here. Even then, backing vocals, reverb and doubled lines can add extra notes: raise the shortest note and keep One note at a time on.
What to do with the MIDI
- Play the melody on any instrument: piano, strings, synth lead, guitar sampler.
- Write chords under it: pair it with the chord finder to see which chords fit your key.
- Sketch a hook before you forget it, then edit the notes in a piano roll.
- Import to notation software to get a rough lead sheet (expect to fix rhythm).
Accuracy: what to expect
Honest answer: approximate. Clean, steady humming of a simple tune converts well. Fast runs, heavy vibrato, falsetto breaks, whispering and rap do not: speech and rap are not sustained pitches, so the output is mostly noise. The tool does not know words or syllables, and it cannot split harmonies into separate parts.
| Input | Result |
|---|---|
| Steady humming or "la" singing | Good for a melody sketch |
| Sung with words and vibrato | Fair: some notes split or merged |
| Vocal stem from a song | Fair: reverb and doubles add extra notes |
| Choir or harmonies | Chords, not separate parts |
| Rap or speech | Not suitable |
Frequently asked questions
Can I record directly in the tool?
Not on this page. Record with your phone's voice memo app or any recorder, then drop the file in. If you need a recording tool, the voice recorder page works in the browser too.
Is my voice uploaded?
No. The recording is decoded and transcribed in your browser, and only the small model files are downloaded.
Can it convert the vocals of a song?
Yes, after you remove the music. Run the vocal remover, then drop the vocal stem here. Expect some extra notes from reverb and backing vocals.
Why is my melody chopped into many short notes?
Vibrato and breaks in the voice make the pitch wobble. Raise Shortest note to 100 to 150 ms and move the notes slider toward fewer.
Does it work for rap or speech?
No. It follows sustained pitches, and speech does not have them. You will get scattered notes rather than a melody.
Does it write the lyrics or timing of words?
No. It only outputs note pitches, lengths and velocities. For lyrics timing see the lyric video maker.
Is it free?
Yes: no sign-up, no card and no per-file cap. Recordings can be up to 6 minutes.