A prompt is a brief, the same kind of brief you would give a music director. If you tell a composer 'make something nice for my sister's wedding', you will get something, but probably not the song in your head. If you say 'a playful Hindi-Punjabi sangeet number, female lead, dhol and tumbi with a modern beat, fast and flirty', the first draft lands much closer.
AI works the same way. It fills every gap you leave with a guess, and guesses drift toward the most generic version of a song. This guide gives you a formula you can reuse, shows where each part goes in Autunes, and lists the mistakes that waste credits.
The seven-part prompt formula
You do not need all seven parts every time, but every part you skip is one the model decides for you. Read it as a checklist, not a template you must fill in order.
- Genre: the style family, as specific as you can make it. 'Sufi rock' beats 'Indian music'; 'lo-fi hip-hop' beats 'chill'.
- Mood: how it should feel to listen. Wistful, defiant, devotional, flirty, triumphant, hushed.
- Instruments: two or three you actually want to hear. Dhol, tabla, harmonium, bansuri, acoustic guitar, 808 bass, strings.
- Voice: male or female, solo, duet or choir, and the delivery. Soft, powerful, raspy, rap.
- Tempo: a BPM number if you know one, or plain words such as slow, mid-tempo, brisk, fast dance.
- Language: the language the vocal should be sung in. On Autunes this also goes in the language picker.
- Use: what the song is for. A first dance, a Diwali reel, a temple aarti, a gym playlist, a birthday surprise.
Where each part goes in Autunes
If you start from an idea, type it into Describe your song and Autunes writes the lyrics for you. Put the whole formula in that one box: 'A Marathi Ganpati aarti, male and female chorus, dholki and taal, brisk and joyful, for a home pandal.'
If you write your own lyrics, the lyrics go in the Lyrics box and the sound goes in Style Tags. Keep them apart: style words typed into the lyrics box are not lyrics. Autunes now catches lines like 'Mood: festive' or 'Music: dhol, shehnai' in the lyrics box and moves them into the style instead of singing them, but it is cleaner to put them where they belong from the start.
Three more controls do part of the prompt's job for you. The language picker sets the singing language and defaults to Hindi; if your description names an Indian language, the picker follows it until you change the picker yourself, and your own choice in the picker always wins. The vocal tags (male vocal, female vocal, duet, choir, rap, soft vocals, powerful vocals) set the voice in one click. And the BPM field pins the tempo when words like 'fast' are not precise enough. If you want no singing at all, switch on Instrumental Only rather than writing 'no vocals'.
Example prompts you can copy
Each of these names the sound, the voice, the tempo and the purpose. Swap the details to fit your own song.
- Romantic Hindi acoustic ballad, soft female vocal, fingerpicked guitar and light strings, slow 72 BPM, for a first-dance video.
- Punjabi wedding bhangra, dhol and tumbi, powerful male lead with group chorus, fast 140 BPM, loud and celebratory.
- Bhojpuri festive folk-pop, male and female duet, dholak and harmonium, bouncy mid-tempo, Chhath celebration.
- Evening Krishna bhajan in Hindi, gentle female vocal, bansuri and tanpura drone, slow and meditative, for a home satsang.
- Tamil folk-pop, spirited female vocal, mridangam and violin over a modern beat, 118 BPM, festive and proud.
- Hindi gully rap, boom-bap drums, gritty male rap vocal, fast tight flow, story of a first job in a big city.
- English indie pop, airy male vocal, jangly electric guitar and soft synth pads, 96 BPM, bittersweet, about leaving college.
- Lo-fi instrumental for studying, mellow electric piano, vinyl crackle, soft brushed drums, 70 BPM, calm and warm.
Sound words beat feeling words
The most common weak prompt is a stack of adjectives: 'beautiful, emotional, amazing, catchy, viral song'. None of those words describes a sound, so the model has nothing to act on. Replace each feeling word with something you could point to in a recording.
Instead of 'emotional', try 'piano and cello, restrained vocal, slow'. Instead of 'energetic', try 'dhol, claps, fast 140 BPM'. Instead of 'catchy', give the song a short chorus line that repeats in your lyrics. Mood words are still useful, just pair each one with an instrument or a tempo so it has something to hold on to.
Naming a famous singer is another weak spot. Autunes is not built to copy real artists' voices, and a name tells the model less than a plain description of the sound you like about that singer: 'deep, warm baritone, slow and conversational' is more useful than any name.
How long should the prompt be?
One or two sentences, roughly 15 to 40 words, is enough for almost every song. Shorter than that and you are leaving most of the formula to chance. Much longer, and you tend to add contradictions, such as 'calm' and 'high energy' in the same brief, which pull the song toward a muddy middle.
If a long prompt keeps producing confused results, cut it back to genre, two instruments, voice and tempo, then add one detail at a time.
Common prompt mistakes
Most disappointing first takes come from one of these.
- Leaving the language picker on Hindi while writing English or Tamil lyrics.
- Asking for two clashing genres at once, such as 'heavy metal lullaby', without saying which one leads.
- Writing 'no vocals' in the prompt instead of using Instrumental Only.
- Listing six or seven instruments. Two or three clear ones produce a cleaner mix.
- Putting style notes, titles or credits inside the lyrics box.
- Changing five things between takes, so you cannot tell which change helped.
When the first take is close but not right
Do not rewrite everything. If the song is nearly there, make a variation: a fresh take of the same song for 5 credits (one free on the Free plan, unlimited on Pro and Creator). If one element is wrong, change only that part of the prompt and generate again, so you learn what each word does.
Engine choice matters too. Zori5 Turbo costs 10 credits a song and suits quick drafts; Zori5.5 XL costs 15 and gives a richer result. On the Free plan you get XL when you paste your own lyrics, which is a good reason to write or paste lyrics rather than only describing an idea once you know what you want.