AI Song Prompts: 30 Examples That Actually Work

Quick answer: The AI song prompts that work best follow one pattern: genre + tempo + 2 to 3 instruments + vocal + one emotion + a concrete subject, in 15 to 45 words. Genre, tempo and vocal words change the output the most. Stacked adjectives, mixing jargon and artist names change it the least or cause problems. Below are 30 tested-style examples with notes on why each works.

This page is a working library for anyone who has typed a prompt into a song generator and got something bland back. It is for creators, hobby songwriters and anyone using an AI song generator with vocals who wants to steer the result instead of rolling dice.

Each example comes with a short note on what it is doing. Read the notes, not just the prompts: once you see which details the model responds to, you can write your own prompts for any idea. At the end you will find a rewrite clinic that takes four weak prompts and fixes them.

What actually changes the output

Before the examples, here is how much different kinds of words tend to move a result. This is based on how text-to-music models are generally built and what you hear across many generations, not a measured benchmark. Use it as a priority list when a prompt is getting long.

The subject ranks lower for the sound because it mostly steers the lyrics, not the arrangement. That is why story belongs in the lyrics when you write your own.

Pop and dance prompts (1 to 6)

1. "Bright synth-pop at 118 BPM, punchy drums, analog synth bass, airy female vocal, euphoric, about dancing alone in the kitchen at 2 a.m." The BPM sits in the classic dance-pop pocket and "alone in the kitchen" gives the lyrics a scene.
2. "Mid-tempo pop ballad at 76 BPM, grand piano and soft strings, male vocal that builds into a big final chorus, bittersweet, about selling a childhood home." "Builds into a big final chorus" is one of the few arrangement instructions models follow reliably.
3. "Upbeat pop at 122 BPM with handclaps and a whistled hook, cheerful female vocal, about the first warm day of spring." A whistled hook is a distinct sound that shows up clearly.
4. "Dark pop at 96 BPM, sub bass, sparse snaps, breathy female vocal, confident and cool, about walking away from a group chat." Sparse arrangements need the word "sparse" or "minimal", or the model fills space.
5. "80s-inspired synth-pop at 112 BPM, gated drums, bright arpeggiated synths, male vocal, nostalgic, about a mixtape found in a glovebox." An era cue does a lot of work here: it implies drum sound, synth choice and reverb.
6. "Dance-pop at 124 BPM, four-on-the-floor kick, piano chords, powerful female vocal, joyful, a wedding party song about finally saying yes." Four-on-the-floor is genre vocabulary the model understands.

For deeper pop technique, including where hooks should land, see making pop songs with AI.

Acoustic, folk and country prompts (7 to 12)

7. "Fingerpicked indie folk at 88 BPM, acoustic guitar and cello, soft male vocal, tender, about my grandmother's handwriting on recipe cards." Two acoustic instruments plus "soft" gives an intimate, close sound.
8. "Warm country at 100 BPM, acoustic guitar, fiddle and pedal steel, friendly baritone vocal, about a dog who waits by the door every evening." Fiddle and pedal steel are strong country markers. Our country song guide has more.
9. "Campfire singalong folk at 108 BPM, strummed guitar, stomps and claps, group vocals on the chorus, warm, about old friends on a lake trip." "Group vocals on the chorus" often produces gang-vocal choruses.
10. "Slow Americana at 72 BPM, harmonica and upright bass, weathered male vocal, lonely, about a diner that stays open all night." "Weathered" is a vocal texture word that tends to register.
11. "Upbeat bluegrass at 132 BPM, banjo, mandolin and fiddle, bright female vocal, playful, about a chicken that escapes every Sunday." Bluegrass runs fast; a slow BPM here would fight the genre.
12. "Acoustic singer-songwriter at 80 BPM, nylon guitar only, whispery female vocal, calm, about a rainy morning in a small apartment." "Only" is a useful limiting word.

Rock, punk and soul prompts (13 to 18)

13. "Garage rock at 140 BPM, fuzzy guitars, loose drums, shouted male vocal, rebellious, about getting kicked out of a parking lot." "Loose" and "fuzzy" push toward a raw, unpolished sound.
14. "Pop-punk at 168 BPM, power chords, fast drums, energetic male vocal, about the last day before summer break." Pop-punk lives above 150 BPM.
15. "Stadium rock at 120 BPM, big guitars, pounding toms, soaring female vocal, triumphant, about a small town team winning the final." "Soaring" and "triumphant" point in the same direction, so they reinforce instead of cancel.
16. "Retro soul at 96 BPM, horns, Hammond organ and tight drums, gritty male vocal, joyful, about getting paid on a Friday." Horns plus organ is a strong era signal.
17. "Slow blues shuffle at 70 BPM, electric guitar and piano, raspy male vocal, weary, about a car that will not start in winter." Describing the voice ("raspy") beats naming a singer.
18. "Neo-soul at 84 BPM, warm Rhodes, laid-back drums, smooth female vocal, about a Sunday with no plans." "Laid-back" shifts drum feel behind the beat.

Hip hop, electronic and lo-fi prompts (19 to 24)

19. "Boom bap hip hop at 90 BPM, dusty drums, jazzy piano sample feel, confident male rap vocal, about growing up above a laundromat." For more rap-specific prompting, see how to make a rap song with AI.
20. "Trap at 140 BPM, rolling hi-hats, 808 bass, melodic male vocal, determined, about working two jobs."
21. "Lo-fi hip hop at 78 BPM, vinyl crackle, soft Rhodes, mellow female vocal, sleepy, about studying in a library at closing time." Texture words like vinyl crackle are well understood in lo-fi.
22. "Synthwave at 100 BPM, retro synths, gated reverb drums, male vocal, night-drive mood, about neon signs reflected on wet streets."
23. "Melodic house at 122 BPM, warm pads, plucked synth, soft female vocal chops, uplifting, about a sunrise after a long night."
24. "Chill electronic at 95 BPM, glitchy percussion, deep bass, airy vocal, calm, about watching a city from a train window."

Special-use prompts (25 to 30)

25. "Gentle lullaby at 64 BPM, music box and soft strings, hushed female vocal, peaceful, about a baby named Ada falling asleep." Names in the prompt usually make it into the lyrics.
26. "Funny country ballad at 80 BPM, acoustic guitar and pedal steel, overly dramatic male vocal, about the last slice of pizza." "Overly dramatic" is a performance cue that often comes through.
27. "Uplifting worship at 72 BPM, piano and swelling pads, warm female lead with choir on the chorus, about gratitude in hard seasons." Our worship song guide covers congregational range.
28. "Short pop hook at 120 BPM, big drop after the first line, catchy female vocal, about Monday mornings." Sized for vertical video.
29. "Educational pop at 110 BPM, bright ukulele and claps, friendly female vocal, about the planets in order from the sun." For teaching songs, see educational rap songs for a related approach.
30. "Cinematic instrumental at 70 BPM, piano, strings and slow-building drums, no vocals, hopeful, for a travel video ending." "No vocals" plus vocals set to none removes any chance of a stray voice.

The rewrite clinic: four weak prompts fixed

Most bad results come from one of four problems. Here they are with the fix.

Problem 1: Too vague

Before: "A song about love." After: "Slow R&B at 72 BPM, warm keys and soft drums, smooth male vocal, tender, about cooking dinner for someone after a long shift." The topic is still love, but now it has a scene.

Problem 2: Adjective soup

Before: "Epic emotional powerful dark uplifting cinematic song." After: "Cinematic pop at 90 BPM, piano and big drums, powerful female vocal, determined, about running a first marathon." One emotion, one subject.

Problem 3: Too many instruments

Before: "Guitar, piano, violin, cello, trumpet, saxophone, synth, banjo, drums." After: "Piano and cello with light drums." Name the two or three you care about. The model will add support instruments that fit the genre anyway.

Problem 4: Story in the wrong place

Before: A 120-word prompt narrating the whole plot. After: A 25-word style prompt, with the story moved into your own lyrics, tagged with [Verse] and [Chorus]. If you need help with that step, see how to format lyrics for AI and our guide on song structure tags.

Words that usually get ignored

Some prompt words feel helpful but rarely change anything you can hear:

• Mixing and mastering jargon like "sidechain at 4:1", "LUFS -14" or "Neumann U87". The model mixes in one pass and does not expose those controls.
• Key signatures in passing, like "in F sharp minor", tend to be loose suggestions in most text-to-music tools. Tempo is far more reliable.
• Quality words like "high quality", "professional", "masterpiece". Every output aims for that already.
• Negative lists like "no guitars, no drums, no reverb, no harmony". One clear negative ("no vocals") works; a long list of exclusions is less reliable than describing what you want.

Mixing details aside, it is worth knowing how tempo and key fit into music prompts generally. Our explainer on BPM and key in AI music covers when each matters.

A template you can reuse

Copy this skeleton and fill in the brackets:

[Genre] at [BPM] BPM, [instrument 1] and [instrument 2], [vocal texture] [gender] vocal, [one emotion], about [concrete subject with one physical detail].

Then run a quick self-check before you generate:

If a prompt passes all seven checks and the result still misses, generate again before rewriting. Two takes of the same prompt can sound quite different. If the vocal is the problem, our guide on choosing vocal gender, style and language is the next stop.

Worked example: building a prompt from zero in three passes

Here is how one prompt evolves when you follow the priority list above, starting with only an idea: a song for a friend who just opened a tiny bakery.

Pass 1, the idea only. "A song about my friend's new bakery." This generates, but the model has to guess the genre, tempo and voice. In practice you get a generic mid-tempo pop song with lyrics about "dreams coming true", which could be about anything.

Pass 2, add the three strongest levers. "Acoustic pop at 104 BPM, warm female vocal, about my friend's new bakery." Now the sound has a direction. The tempo feels like a morning walk, and the acoustic label keeps it friendly rather than glossy. The lyrics are still vague because the subject is still a label, not a scene.

Pass 3, add instruments, one emotion and a physical detail. "Acoustic pop at 104 BPM, ukulele and light hand percussion, warm female vocal, proud and cheerful, about my friend Lena opening her tiny bakery at 5 a.m., flour on the doorframe." Thirty words. Every part pulls the same way. The name and the flour on the doorframe give the lyric writer something real to build verses from, and "proud" gives the chorus a feeling to state.

If the third pass still feels close but not quite right, change one lever per take. Swap ukulele for piano to make it gentler, or move to 112 BPM to make it more of a celebration. Changing one thing at a time is how you learn what each word does in your favorite genres.

How prompts and settings interact

On MusicVivo you can set genre, mood, energy, tempo and vocals with controls as well as in text. When they agree, they reinforce each other. When they disagree, results get unpredictable. If your prompt says "slow ballad at 70 BPM" and the tempo control is set high, you are sending two different instructions.

A simple habit avoids this: decide whether the text or the controls are your source of truth for each lever. Many people let the controls handle vocals and length, since those are clear switches, and let the text handle genre, instruments and mood, since words carry more nuance. Whatever you choose, make sure the two never contradict.

People also ask

What is a good prompt for an AI song? One genre, a tempo, two or three instruments, the vocal type, one emotion and a concrete subject. For example: mid-tempo indie pop at 104 BPM, jangly guitar and tambourine, bright female vocal, hopeful, about repainting a first apartment.

How long should an AI song prompt be? Roughly 15 to 45 words. Move story details into the lyrics box when the prompt gets longer than that.

Can I use an artist's name in an AI song prompt? It is better to describe the sound instead: vocal texture, instruments, tempo and era. That avoids imitating a real person and usually gives clearer results.

Why does the AI ignore parts of my prompt? Too many requests or conflicting ones. Genre, tempo and vocal words carry the most weight, so trim the rest.

Related reading

• How to Write a Song with AI: A Step-by-Step Guide
• Making Pop Songs with AI: Hooks, Tempo and Polish
• Verse, Chorus, Bridge: Structuring Songs for AI
• Text-to-Music Prompting Guide
• Song Lyrics Prompt Examples
• Why Do Some AI Songs Sound Robotic, and How to Fix It