Text-to-Music Prompting: Genre, Mood and Tempo Words

Quick answer: A strong text-to-music prompt fills six slots in about 20 to 40 words: genre, instruments, texture, mood, energy and tempo, plus where the music will be heard. Use concrete sound words like "brushed snare," "tape-saturated piano" or "sub bass" instead of adjectives like "epic," and put an exact BPM when timing matters.

This is a reference page. Keep it open while you write prompts for any AI music tool, including the MusicVivo AI music generator. It is aimed at people who already know the basic workflow and want better words, not more words.

After reading you will have a six-slot formula, word banks for each slot, a list of words that do nothing, and several before and after rewrites you can adapt directly.

The six-slot formula

Every effective music prompt we write follows the same skeleton. You do not need every slot every time, but checking each one before you render catches most weak prompts.

Assembled, that reads: "dusty boom bap instrumental, chopped soul piano, upright bass and crunchy drums, vinyl crackle, slightly detuned, nostalgic late-night feel, laid-back head-nod groove, 88 BPM." That is 28 words and every one of them is doing work.

Then add context at the end if the track serves a purpose: "for a coding livestream," "under a podcast interview," "for a game's pause menu." Context is the slot most people skip and the one that most often fixes arrangement problems.

Slot 1: genre and style words

Genre is the heaviest word in the prompt. It sets drums, instruments, tempo range and mix character at once. Be as specific as you can while staying with a style that is widely recognized.

Pairing two genres works when they share a tempo range and rhythmic feel: "lo-fi jazz," "folk pop," "cinematic synthwave." Pairing genres that clash, like "death metal bossa nova," usually produces a muddle unless that is the joke you want.

For deeper genre guides with full prompt sets, see synthwave and retro electronic, jazz and cinematic music.

Slot 2 and 3: instruments and texture

Instruments say what plays. Texture says how it sounds. Together they do most of the fine steering.

Instruments: name two or three. The model has to give each named part space. Beyond four, parts start fighting and the mix gets crowded.

Useful instrument words with built-in character:

• Keys: Rhodes, Wurlitzer, felt piano, upright piano, honky-tonk piano, Hammond organ
• Guitars: nylon-string guitar, fingerpicked acoustic, clean Strat-style electric, fuzz guitar, slide guitar, twelve-string
• Bass: upright bass, round Motown-style bass, sub bass, 808, slap bass, synth bass
• Drums: brushed snare, tight electronic drums, gated reverb drums, live room kit, hand percussion, shaker, rimshots
• Synths: analog pad, arpeggiated synth, supersaw lead, plucky synth, warm Juno-style chords
• Orchestral: solo cello, staccato strings, French horn swells, choir pad, taiko drums, celesta

Texture words: one or two. These are adjectives that describe recordings rather than feelings, which is why they work.

• tape-saturated, lo-fi, vinyl crackle, warbly, detuned
• dry and close, roomy, cavernous reverb, airy
• crisp, punchy, gritty, distorted, glassy, muffled

The "don't" column is not wrong in intent. It just carries no information the model can act on. "Professional" and "high quality" are what every render already tries to be.

Slot 4: mood words that actually change the music

Mood words work best when they imply musical choices. "Melancholy" suggests minor harmony and slower phrasing. "Triumphant" suggests brass, major keys and big drums. "Happy" is so broad it barely moves anything.

Here is a bank sorted by emotional direction. Choose one, or two from the same row.

Contradictions confuse. "Calm but aggressive" usually averages into neither. If you want contrast, put it in time instead: "starts calm and sparse, becomes aggressive at the drop." Now the model has a structure to follow rather than two forces to blend.

Slot 5 and 6: energy and tempo

Energy and tempo are different. A track can be slow and heavy (a 70 BPM doom groove) or fast and light (a 170 BPM ambient drum and bass with sparse pads). Treat them as two dials.

Energy words, low to high: sparse, minimal, laid-back, steady, driving, pounding, relentless.

Tempo: use numbers. "Upbeat" might land anywhere from 100 to 140 BPM. "112 BPM" lands at 112. Here are the commonly cited ranges for popular styles, so you can pick a number that fits the genre you named.

Trap is often written at 130 to 160 BPM but feels like half that speed, because the snare lands on beat three. If you want a slow, heavy trap vibe, the number still belongs in that higher range. For more on how tempo and key affect editing and layering, read setting BPM and key.

Before and after: four rewrites

These are the kinds of prompts people actually type, and the rewrites that work better. Each "after" stays under 40 words.

1. Study music

• Before: "chill music for studying"
• After: "lo-fi hip hop instrumental for a late-night study stream, mellow Rhodes chords, soft boom bap drums, vinyl crackle, calm and steady, 76 BPM, no vocals"

2. Product video

• Before: "upbeat corporate music"
• After: "bright modern pop instrumental for a 30-second app launch video, plucky synths, clean electric guitar, claps on two and four, confident and optimistic, 116 BPM, builds to a clear ending"

3. Horror game

• Before: "scary music"
• After: "dark ambient track for a horror game corridor, low cello drones, distant metallic scrapes, sparse heartbeat pulse, uneasy and cold, very slow, no melody"

4. A song with vocals

• Before: "love song"
• After: "warm acoustic pop ballad about meeting again after years apart, fingerpicked guitar and soft piano, gentle female lead vocal, bittersweet and hopeful, 72 BPM, big final chorus"

For vocal prompts specifically, including how to ask for harmonies and delivery styles, the AI song prompt tips article goes further.

Words that do nothing, or do harm

Some words appear in almost every beginner prompt and contribute nothing. Others actively hurt.

Empty words: amazing, awesome, high quality, professional, best, perfect, beautiful, nice, cool, epic (on its own).

Harmful patterns:

• Living artist names. Reputable tools will not imitate a specific artist, and you should not try. Describe the sound instead: "breathy, close-miked female vocal over sparse electronic beats."
• Copyrighted song titles as references. Same issue. Describe tempo, instruments and mood.
• Too many instruments. A list of seven instruments becomes a crowded mix.
• Contradictory moods stacked together.
• Instructions the tool cannot follow. "Make it go viral" is not a sound.

Context phrases: the slot that fixes arrangements

Context tells the model what the music has to make room for. These phrases reliably change density and the space left in the midrange:

• "under a voiceover" or "under a podcast conversation": fewer lead parts, steadier dynamics
• "for a game menu" or "loopable": no big ending, consistent energy
• "for a trailer" or "builds to a hit": rising intensity, clear climax
• "for a wedding first dance": slower, warmer, emotional peak in the last third
• "background for a cafe": gentle, unobtrusive, no sudden changes

If you make a lot of content beds, the background music for videos guide pairs these phrases with mixing advice for speech.

Structure words: shaping the track over time

Most prompts describe a track as if it were a single moment. Music happens over time, and a few structural words let you shape that arc without writing out sections.

• "starts sparse, builds gradually": a slow rise in density, good for openers and reveals
• "drops out to just piano before the final section": a breakdown that makes the return feel bigger
• "steady throughout, no big changes": ideal for loops, beds and anything under speech
• "short intro, straight into the groove": useful when an edit starts on action
• "ends on a sustained chord" or "clean ending on the downbeat": avoids abrupt cut-offs
• "loopable, no intro or outro": for game levels and streams

Use one structural phrase per prompt. Two can work if they describe different moments, such as "short intro, ends on a sustained chord." Three or more starts to read like a list of wishes.

For vocal songs, structure is better set with section tags in pasted lyrics than with words in the prompt. The song structure guide for AI generation shows how verse, chorus and bridge tags map to bar counts.

Building a personal prompt library

The fastest way to improve is to keep the prompts that work. A plain text file with one entry per successful render is enough. Record four things:

1. The full prompt text.
2. The settings: genre preset, tempo, energy, vocals and length.
3. One sentence on what the track was used for.
4. One sentence on what you would change next time.

After a dozen entries you will notice patterns in your own taste: the instruments you return to, the tempo where your edits feel natural, the context phrases that fix your recurring problems. Those patterns become templates. A template is a prompt with blanks, such as "lo-fi hip hop instrumental for a ___ stream, ___ chords, soft boom bap drums, vinyl crackle, ___ BPM, no vocals," which you fill in seconds.

Group the library by use rather than by genre. When a new job arrives you think "I need something for a product video," not "I need synth pop," and the use-based grouping matches how you will search.

A printable prompt checklist

Run this before you press generate. It takes ten seconds and catches most weak prompts.

When a prompt passes the check but still misses, change one slot at a time. Our practical workflow for generating music explains how to run those one-change comparisons efficiently.

People also ask

What makes a good text-to-music prompt? A good prompt fills clear slots: genre, two or three instruments, a texture, a mood, energy and tempo, plus where the music will be heard. Concrete sound words steer far better than vague adjectives.

Should I put the BPM in a text-to-music prompt? Yes, when timing matters. A number is more precise than words like upbeat, and it helps with editing and layering. Keep it in the usual range for the genre.

Which words should I avoid in music prompts? Skip empty intensifiers like amazing, contradictory mood pairs, long instrument lists and the names of living artists. Replace them with specific instruments, textures or moods.

How long should a text-to-music prompt be? One or two sentences, around 20 to 40 words. For more control, use the dedicated settings rather than piling on adjectives.

Related reading

• How to Generate Music with AI: A Practical Workflow
• How to Make Instrumental Music with AI
• Setting BPM and Key When Generating AI Music
• AI Song Prompt Tips
• Lo-fi Sound Textures
• Making Synthwave and Retro Electronic Music with AI