How to write AI music prompts

An AI music prompt is not a list of keywords, nor is it a paragraph of adjectives that gets better the longer it becomes. It is more like a short production brief: it tells Mureka what genre the song is, what emotion it expresses, who sings it, which instruments are used, how it develops, and what kind of final sound texture you want to hear.

A good Prompt does not have to be long, but every important description should correspond to a musical result that can be heard and compared. Instead of writing “premium, good-sounding, professional,” it is better to specify whether the verse should be restrained, whether the chorus should open up, whether the vocal should feel close, whether the drums should sit back, and which sounds should leave room for the main melody.

The 7 elements of an AI music Prompt

A Prompt that is easy to control is usually organized in this order:

Genre + emotion + rhythmic feel + vocal direction + core instruments + section structure + production texture

For example:

Dreamy alternative R&B, intimate and nocturnal, slow groove, clear lead vocal with close harmonies, electric piano and muted guitar, restrained verse that opens into a spacious chorus, warm low end and clean modern mix.

This prompt explains the genre, atmosphere, speed, vocal, instruments, section contrast, and mixing texture in order. The model receives a prioritized musical map, not a string of competing tags.

Write the creative intent first, then add sound details

Before you start listing instruments, answer three questions:

1. What theme does this song want to express?

2. What emotional change should the listener feel?

3. What scenario will the song be used for?

You can first write a brief for yourself:

A reflective indie-pop song about starting over in a new city, written for a late-night drive. The mood moves from uncertainty to quiet confidence.

Then expand it into a music Prompt:

Reflective indie pop, mid-tempo and cinematic, intimate lead vocal, clean electric guitar, warm piano and subtle synth pads, restrained verse, rising pre-chorus, open singable chorus, gradual emotional lift, polished but natural mix.

The first line explains “why this song is being written.” The second explains “what you want it to sound like.” When the two work together, it becomes easier for Mureka to turn an abstract theme into concrete choices in melody, vocals, arrangement, and structure.

How to write genre: from broad category to audible traits

Writing only a broad category, such as pop or rock, is usually not enough to define the full sound. You can add the most important rhythm, instrument, vocal, and production traits of that genre.

The three-layer structure of a genre Prompt

Layer 1: genre identity

State the musical direction of the song, such as indie pop, dream pop, alternative R&B, West Coast G-Funk, or minimal electronic pop.

Layer 2: audible markers

Add elements that can be heard, such as reverb-rich guitars, warm Rhodes, sliding synth bass, restrained drums, or crystalline synth pulses.

Layer 3: overall character

Explain how those elements work together, such as hazy and intimate, laid-back and nocturnal, or spacious and contemplative.

Example:

West Coast G-Funk, laid-back and nocturnal, sliding synth bass, Moog whistle lead, talkbox atmosphere, warm Rhodes, muted wah guitar, behind-the-beat drums, raspy male vocal, smoky lowrider night-cruise mood.

This is more specific than “G-Funk, cool, atmospheric,” because every phrase corresponds to audible information about arrangement, rhythm, vocal, or space.

How to write emotion: describe change, not just adjectives

“Sad,” “happy,” and “emotional” can be starting points, but they do not explain how the emotion develops in the song. A more useful way is to describe the emotional curve and the role of each section.

For example:

Begin intimate and uncertain, build tension through the pre-chorus, release into a hopeful chorus, then return to a quiet and resolved ending.

You can also break emotion into actions:

• restrained verse: keep the verse restrained;

• rising pre-chorus: gradually increase tension in the pre-chorus;

• open sustained notes in the chorus: use more open and extended notes in the chorus;

• stripped bridge: reduce density in the bridge to create contrast;

• gradual dynamic lift: gradually lift the overall energy.

This way of writing translates emotion into structure and vocal action, making it easier to verify in the generated result.

How to write vocals: from “voice” to “singing style”

A vocal Prompt does not only describe gender or range. It can also specify pronunciation, distance, emotion, and singing method.

Common dimensions for vocal description

Timbre and range: warm low register, clear bright tone, soft breathy texture.

Sense of distance: close and intimate vocal, spacious vocal with gentle reverb.

Singing style: conversational phrasing, restrained vocal runs, confident sustained notes.

Lyric expression: clear diction, emphasize the central phrase, keep the verse conversational.

Harmony arrangement: close harmonies in the chorus, stacked backing vocals in the final chorus.

Example:

Clear female lead vocal with conversational phrasing, precise diction, restrained verses, stronger sustained notes in the chorus, and close harmonies added only in the final section.

If you need pronunciation for a specific person or proper noun, write the pronunciation requirement separately. Do not use a long string of conflicting vocal adjectives, or the model will have trouble deciding which one matters most.

How to write instruments: explain roles, not just a list

Listing more than ten instruments does not mean the arrangement will be richer. A more effective approach is to describe the role each core instrument plays in the song.

For example:

Fingerpicked acoustic guitar carries the verse, warm bass supports the groove, soft strings widen the pre-chorus, and muted drums keep the chorus moving without masking the vocal.

This Prompt explains that the guitar carries the verse foundation, the bass supports the groove, the strings widen the pre-chorus, and the drums move the chorus forward without covering the vocal.

A shorter version also works:

Piano-led verse, subtle synth texture underneath, fuller drums and layered guitar in the chorus, leave space around the lead vocal.

Three questions to consider when writing instruments

1. Which instrument carries the main melody or main motif?

2. Which instrument handles rhythm, low end, or space?

3. Which instruments enter only in specific sections?

Assigning roles to instruments usually helps the model create a layered arrangement better than simply adding more instrument names.

Use section structure to control energy and contrast

Structure is one of the easiest parts of a Prompt to overlook, but also one of the most useful. It tells the model that the song is not one sound that stays the same from start to finish, but a complete form that develops.

Common structure terms

[Verse 1] restrained and intimate
[Pre-Chorus] gradual lift and rising tension
[Chorus] open, memorable, and emotionally released
[Verse 2] add a new melodic detail
[Bridge] stripped back with a different texture
[Final Chorus] fuller harmony and wider arrangement

If you do not want to write lyric labels, you can put it directly into the Prompt:

Sparse verse, gradual pre-chorus lift, wide memorable chorus, stripped bridge, final chorus with added harmonies and percussion.

The value of structure terms is that they help explain why a certain sound is needed, instead of asking every second of the song to stay equally “full.”

Use space and density to make the arrangement clearer

A common misunderstanding is to treat “more sounds” as “better arrangement.” Mureka V9.5 can follow the idea that arrangement density serves the task of each section, while parts leave room for the main melody, vocals, and emotional change.

You can add these phrases to the Prompt:

• sparse verse: keep the verse sparse;

• leave room for the vocal: leave space for the vocal;

• instruments enter gradually: let instruments enter gradually;

• clear separation between parts: keep the parts clearly separated;

• wider arrangement in the chorus: widen the arrangement in the chorus;

• restrained low end: control low-end density.

Example:

Keep the verse sparse and close, let the vocal lead, introduce percussion gradually, and open the stereo field in the chorus without making the arrangement crowded.

These descriptions do not weaken the completeness of the song. Instead, they help make the chorus’s energy change clearer.

When should you use negative instructions?

Negative instructions are useful for solving a specific problem you have already heard. They are not suitable for listing everything you dislike.

A targeted way to write them:

No spoken intro, avoid excessive vocal runs, no abrupt genre change.

It is not recommended to write a full paragraph of prohibitions, because too much negative content may compete with the positive creative direction. First make clear what you want to hear, then add one to three negative instructions to fix the most obvious problems.

Prompt length: be specific, but do not lose priority

A Prompt can contain a lot of information, but the most important elements should appear first. The recommended priority is:

1. genre and overall identity;

2. emotion and rhythm;

3. vocal direction;

4. core instruments;

5. sections and dynamics;

6. production texture and a small number of constraints.

If the generated result moves away from the direction, do not immediately make the Prompt longer. First remove unnecessary descriptions, then check whether the three to five most important musical traits are clear enough.

How to iterate a Prompt after generation

The real value of a Prompt is that it lets you explain what result a change produced. A useful method is to change only one major variable at a time.

A practical iteration loop

First version: define the identity.

Write only the genre, emotion, rhythm, vocal, and two or three core instruments first, and confirm the overall direction of the song.

Second version: adjust the structure.

Add contrast between verse, chorus, and bridge, and check whether the energy develops.

Third version: adjust the vocal.

Modify pronunciation, distance, singing style, harmony, or how key phrases are emphasized.

Fourth version: adjust the arrangement.

Explain which parts should sit back, which instruments should enter in the chorus, and how the low end and space should change.

Fifth version: solve specific problems.

Add only a small number of negative instructions to solve the most obvious repetition, crowding, or abrupt changes heard during listening.

Save every version of the Prompt and its corresponding audio, so next time you know which description truly changed the result.

Four Prompts you can directly adapt

1. Emo Rap

https://www.mureka.ai/song-detail/156330950197249?from=mine_generation_song

Emo rap with restrained midtempo half-time pull, intimate and emotionally exposed. Use a 2-bar intro: a detuned glassy synth phrase through chorus and reverse reverb, then enter the vocal. Build from a soft clipped kick, dry snare, sparse hats and rounded sub-bass that follows the vocal cadence. Let a blurred synth pad and distant formant-vocal texture answer the main three-note motif. Keep verses thin and conversational; lift the hook with doubled lead vocals, a higher answer and wider synth tails, not extra drums. Feature a young male voice moving between weary melodic rap and cracked singing, with clear words, small pitch scoops and restrained Auto-Tune. Use concrete breakup details and self-contradicting thoughts. Keep the mix close, hazy and low-end controlled. Avoid piano, strings, guitars, brass, trap hat barrages, sliding 808s, bright pop gloss and oversized vocal stacks.

2. Retro Synthwave

https://www.mureka.ai/song-detail/157034838228993?from=mine_generation_song

Dark Korean girl-group synth-pop with a slower pulse, deep minor-key tension and a dangerous sensual mood. Open with a 2-bar close-miked female pickup over a low, filtered analog bass. Keep verse one sparse: heavy muted kick, dry snare, minimal hats and a cold repeating synth-bass figure with wide gaps. Gradually deepen the low end and let a detuned synth motif emerge beneath intimate vocals. Build the pre-chorus through rising melodic tension, darker filters and held silences. Open the hook into broader sub-bass, restrained group harmonies and a hypnotic descending melody, never a bright release. Let verse two grow more intense; save the thickest bass and vocal layers for the final chorus. Feature mature female voices, smoky midrange, controlled breathiness, crisp Korean diction and confident lower-register phrasing. Avoid piano, bells, disco grooves, busy percussion and fast tempos.

3. Girl-Crush Hip-Hop

https://www.mureka.ai/song-detail/156995917316098?from=mine_generation_song

Sleek girl-group hip-hop fused with smooth R&B: confident, sultry and midtempo with a half-time bounce. Use a 2-bar intro: breathy female tag, filtered synth pulse and one clipped 808 hit, then enter verse one. Build from deep sub-bass, syncopated kicks, dry snaps, crisp hats and muted electronic accents. Add warm electric keys, glossy detuned synths and chopped vocal flickers. Feature contrasting female voices: one husky, nasal rapper with crisp consonants, playful swagger, Korean-English phrases and elastic rhythmic phrasing; others deliver airy R&B lines, soft falsetto touches and close layered doubles. Move from stripped rap verses into a smooth melodic pre-chorus, then a short addictive group hook with breathy responses and stronger bass. Keep the mix polished, intimate and fashion-forward. Avoid rock guitars, brass, aggressive EDM drops, shouted choruses and childish cute vocals.

4. UK Garage × Deep House K-Pop

https://www.mureka.ai/song-detail/157064986886146?from=mine_generation_song

Korean girl-group UK garage and deep-house pop with relaxed, slower midtempo swing, cool confidence and subtle Y2K club character. Open with a 2-bar low female pickup and a warm filtered bass pulse, then settle into an unhurried two-step groove. Use soft syncopated kicks, a dry spacious snare, sparse shuffled hats, rounded bass notes and one muted synth stab. Keep rhythmic subdivisions minimal, leaving generous space between drum hits. Rotate female voices between smooth midrange talk-singing, lightly husky melodies, easy rhythmic rap and airy upper-register touches. Let members trade short, relaxed phrases with subtle tuning, close doubles and small vocal chops. Build the hook through a simple repeated melody, gentle bass movement and restrained group responses. Keep it clean, sensual and composed. Avoid fast tempos, double-time drums, busy hats, piano, bells, marimba and EDM drops.

These examples highlight a sense of space, localized genre detail, and instrumental atmosphere respectively. You can replace the theme, rhythm, vocal, and instruments while keeping the order of “genre — emotion — sound — structure — production.”

How to write Prompts for multilingual songs

If the song uses non-English lyrics, it is recommended to clearly state the target language and separate the language requirements from the music requirements:

Spanish lyrics, clear pronunciation and natural syllable stress, modern Latin pop, warm acoustic guitar, mid-tempo groove, intimate verse and a bright singable chorus.

For multilingual songs, you can further explain which language appears in which section, and which shared melody or emotional traits should be maintained. After generation, focus on checking proper nouns, stress, connected pronunciation, and repeated phrases.

Mureka’s materials show that multilingual capability and evaluation systems are important parts of model evolution. Language changes are not only a matter of lyric translation; they also affect syllable count, stress position, melodic landing points, and vocal expression. For that reason, language requirements should be included in the listening and revision process.

How to judge whether a Prompt is effective

Do not only check whether the model “understood a few keywords.” Listen for whether those requirements run through the complete song. You can use this checklist:

• Is the genre identity clear in the verse, chorus, and bridge?

• Does the emotion develop in the direction described in the Prompt, instead of reaching the highest point at the start?

• Does the vocal match the specified timbre, distance, and singing style?

• Do the core instruments play clear roles?

• Does the arrangement add density when needed and leave space when needed?

• Do lyric keywords land in suitable sections and melodic positions?

• Does the ending complete the emotional resolution of the song?

In the same-standard evaluation of 100 songs provided in the materials, Mureka V9.5 reached a 97.0% Prompt-control pass rate and a 95.7% full genre-expression rate. These data points show that Prompt direction and genre expression are key capabilities of V9.5, but the evaluation results come from a specific sample and cannot replace listening judgment for a specific song.

Common mistakes and how to improve them

Mistake 1: only writing the genre name

Problem: Pop song does not explain emotion, rhythm, vocal, or structure.

Improvement: Add rhythmic feel, core instruments, vocals, and section contrast.

Mistake 2: too many adjectives, too few musical actions

Problem: beautiful, amazing, emotional, professional is difficult to turn into audible choices.

Improvement: Rewrite them as concrete descriptions such as restrained verse, rising pre-chorus, clear diction, warm low end.

Mistake 3: asking every section to be “big”

Problem: If every section is full, the chorus has no room to lift.

Improvement: Clearly write sparse verse, fuller chorus, and stripped bridge.

Mistake 4: changing too much at once

Problem: If you change genre, lyrics, vocal, rhythm, and instruments at the same time, it is hard to know what caused the change.

Improvement: Save versions and adjust only one major variable at a time.

Mistake 5: turning constraints into a negative list

Problem: Too many prohibitions can distract the model from the creative goal.

Improvement: First describe the sound you want clearly, then add a small number of targeted constraints.

Common questions

Should AI music Prompts be written in my native language or in English?

Both work. Music terms in English examples are relatively concentrated, which can make it convenient to express genre, arrangement, and production details. No matter which language you use, the key is to describe clear musical actions and section relationships.

Does a longer Prompt produce better results?

Not necessarily. The information in a Prompt needs priority. Long and conflicting descriptions may make it hard for the model to identify the core direction. Use a concise Prompt to define the identity first, then add details based on what you hear.

Why does the same Prompt generate different results?

The result can be affected by lyrics, language, structure, and the generation process. Keep effective Prompts, fix most descriptions, and change only one or two details at a time to find a more stable creative direction.

How can I make the chorus more memorable?

In addition to writing memorable chorus, you can say that the chorus needs a more open melody, longer sustained notes, a repeated central phrase, more harmonies, or a wider arrangement. Then listen to whether the chorus truly contrasts with the verse.

How can I keep the vocal from being covered by the accompaniment?

You can write leave room for the lead vocal, clear vocal center, or accompaniment should support rather than mask the vocal in the Prompt, while reducing unnecessary instruments and low-end stacking.

When should I rewrite a Prompt?

If the song’s core genre, emotion, and structure have all moved away from the target, rewriting the Prompt is more efficient. If only vocal pronunciation, one instrument, or one section has a problem, prioritize local editing or follow-up creation tools such as Remix, Extend, and Studio.

Turn the Prompt into an executable music brief

A good AI music prompt does not stuff every imagined detail into one text. It breaks creative intent into audible choices: genre identity, emotional curve, vocal style, instrument roles, section structure, and production space.

In Mureka, you can first use a concise Prompt to generate a direction, then gradually add details through listening and version iteration. Treat every revision as a production decision, and the Prompt will no longer be a random trial-and-error keyword set, but a reusable, comparable, and sustainable creative method.