Hum to song: the tune in your head becomes the record.

Two very different things get called humming apps. Hum to search — the feature in Google, Shazam or SoundHound — listens to your hum and tries to name a recording that already exists. Hum to song, what this page is about, goes the other way: it takes the melody you hummed as the seed of a new song and writes everything else around it.

Octa is in the second category. The hum sets the melody; a line of text (or nothing at all) sets the genre, the mood and the voice; the engine composes the arrangement, writes or sings the lyrics, and masters the result. Nothing to assemble, nothing to transcribe by hand. The same engine that powers text to song drives it — the melody simply arrives from your voice instead of your keyboard.

You bring the tune. The song comes back finished.

• You humTap the mic in the app and hum or sing one phrase — da-da, la-la or words, whatever feels natural.
• Pitch becomes notesThe recording is traced note by note, octave slips and slow drift are corrected, and the key and tempo are read from what you sang.
• The melody becomes the guideYour tune is laid out across the length of a song so the engine follows it, rather than quoting it once and wandering off.
• Arrangement, lyrics, vocalInstruments, sections and words are written to fit the melody and your style line; a real voice sings it.
• Mix, master, artworkThe result lands in your library as a finished track with cover art, ready to play, share or release.

You get more than one take to choose from, and each take is a different interpretation of the same melody. Pick the one that holds your tune best; delete the rest.

The engine can only follow what it can hear. Five habits make the difference between “that is my tune” and “that is a song in the same mood”:

• Record in a quiet room, phone close to your mouth — no music playing in the background.
• Sing da-da or la-la rather than humming with your mouth closed: every note gets a clear start, so repeated notes are not merged into one long one.
• Keep a steady pace. Tap your foot if it helps; a melody with a pulse is easier to arrange than one that floats.
• Hum one phrase and repeat it once or twice. A few clear seconds beat thirty uncertain ones.
• Stay in one range. Big jumps are fine; switching octaves mid-phrase is read as a slip and corrected, which may not be what you meant.

You do not need perfect pitch. Slight sharpness or flatness is normalised; what matters is the shape of the melody — which notes go up, which go down, and how long each one lasts.

The hum carries the melody; the words you type carry everything else. A short style line — “warm acoustic pop, female vocal, hopeful” — decides the instruments, the tempo feel and the singer. Leave the lyrics box empty and the engine writes lyrics that fit your melody; paste your own and they are sung to it, line by line. Tag sections with [Verse] and [Chorus] if you want the arrangement to follow a structure you already have in mind. The lyrics guide shows how to write lines that sit well on a tune.

Describe the style rather than naming an artist — artist names are blocked by the content filter and the credit is returned. More working style lines, sorted by genre, are in AI music prompts that work.

Humming already starts from your voice; My Voice lets the finished song end there too. Record a short guided sample in the Android app, and the vocal on any song you generate — including the ones you hummed — can be sung by you. It is the difference between “a song built on my melody” and “my melody, sung by me, produced like a record”. Consent rules, limits and the tuning model behind it are on the AI singing voice generator page.

Every hum-to-song result includes the full track, the lyrics as sung, cover art and a shareable page, in your library on the phone and in the web studio at the same time. Songs made from your own melody, prompts and lyrics are yours to use under the app’s terms — streaming, video, a demo for your band. Every uploaded recording is screened against protected works before anything is generated, so humming a famous chorus may be refused; when that happens the credit is returned.