AI music prompt generator: pick six things, copy one line.

The three lines under the prompt are variations: same song, one thing changed. If the first result is close but not right, try one of those before you rewrite everything.

An AI music engine fills every gap you leave. Leave the tempo out and it picks one. Leave the voice out and it picks one. Sometimes it guesses well. The point of a prompt is to stop guessing where it matters to you.

• Genre sets the drum pattern, the chord vocabulary and the song shape. It is the biggest single decision.
• Mood steers melody and harmony. “Dark trap” and “happy trap” are different songs with the same drums.
• Tempo decides how the words fit. A ballad at 75 BPM has room for long lines. At 145 the same lines get swallowed.
• Voice is what people notice first. Say who sings and how: warm, raspy, soft, rapped.
• Instruments give the arrangement a face. Two or three is enough. Ten is a list nobody can follow.
• Production is the room the song lives in: polished and modern, dusty and lo-fi, live and loose.

Then the topic. One concrete thing beats a theme. “Love” is a theme. “Driving back to her street after three years” is a song.

The median prompt behind 4,628 finished songs on Octa was 14 words. Short and specific wins. See the data.

Each of these follows the builder’s pattern. Copy one, change the topic, and it is yours.

• upbeat pop song about leaving a small town, nostalgic and happy, around 120 BPM, warm female vocal, acoustic guitar and piano, polished modern mix, big singalong chorus
• slow R&B song about a text you should not send, romantic and sad, around 75 BPM, soft female vocal, piano and synth pads, stripped back
• fast trap song about working double shifts, dark and energetic, around 145 BPM, rapped male vocal, 808 bass and drum machine, polished modern mix
• mid-tempo country song about a truck that will not start, nostalgic, around 95 BPM, deep male vocal, acoustic guitar and live drums, live band in a room
• instrumental lo-fi track for studying, chill and dreamy, around 75 BPM, piano and vinyl crackle, lo-fi tape warmth, no vocals
• upbeat house track about the last night of summer, energetic and happy, around 120 BPM, soft female vocal, synth pads and drum machine, 80s synths
• fast rock song about quitting a job by email, angry and energetic, around 145 BPM, raspy male vocal, electric guitar and live drums, live band in a room, big singalong chorus
• slow acoustic song about a grandmother's kitchen, nostalgic and romantic, around 75 BPM, warm female vocal, acoustic guitar and strings, stripped back

For more, sorted by genre, see 36 AI music prompts that work.

• Naming an artist. Octa blocks artist names in prompts. Describe what you like about the sound instead: the tempo, the voice, the instruments.
• Stacking genres. “Pop rock trap jazz” is four half-songs. Pick one genre and let the mood and instruments add the colour.
• Writing a paragraph. The engine weighs every word. A long story in the prompt pulls it in ten directions. Put the story in the lyrics.
• Leaving out the voice. If you care who sings, say so. It is the first thing a listener hears.

Copy the line, open Octa in the Android app or the web studio, paste it and press create. About a minute later you have a full track with vocals, mix, master and cover art. If you already wrote the words, paste them too and check them first with the lyric checker. Lyrics can be written in many languages; sung results vary by language, so listen to the first take before you make more.

Have a reference track and want to match it? Run it through the key and BPM finder first and put the numbers in the prompt.