⚡ Short answer
Describe the sound, not just the genre. The best Muse prompts name the mood, instruments, tempo, vocals, and structure in one or two sentences — then you generate 3–4 variations and keep the best. Formula and examples below.
What Muse does
Muse is an AI song generator: you describe the music you want in plain language, it composes a complete track — arrangement, mix and all — in seconds. You then refine it, generate variations, and export when you're happy. No instruments or theory required. The workflow is always the same: describe → generate → refine. Your leverage is entirely in step one: the prompt.
The 7-part prompt formula
A strong Muse prompt answers seven questions. You don't need all seven every time, but the more you give, the less generic the result:
- Genre — lo-fi hip-hop, cinematic orchestral, synth-pop…
- Mood — melancholic, euphoric, tense, cozy…
- Instruments — fingerpicked acoustic guitar, warm Rhodes piano, 808s…
- Tempo / energy — "slow 70 BPM", "driving", "builds to a big chorus"
- Vocals — instrumental, soft female vocal, male spoken word…
- Structure — verse-chorus, ambient loop, builds then drops
- Reference feeling — "like a rainy Sunday morning", "like a movie trailer climax"
Tip: describe the feeling of a reference instead of naming an artist — it guides the model without boxing it into imitation.
7 copy-paste prompts
1. Lo-fi study beats
Lo-fi hip-hop, cozy and calm. Warm Rhodes piano chords, soft vinyl crackle, mellow drums at 76 BPM. Instrumental, loopable, no sudden changes. Like rain on a window while studying.
2. Cinematic trailer build
Cinematic orchestral, tense and epic. Starts with low strings and a heartbeat pulse, builds over 60 seconds to a huge brass and choir climax, then cuts to silence. No vocals.
3. Upbeat pop demo with vocals
Upbeat synth-pop, joyful summer energy. Bright synths, driving 122 BPM beat, catchy chorus hook. Female vocal, verse-chorus-verse-chorus structure. About road trips with friends.
4. Ambient focus music
Ambient electronic, weightless and calm. Slow-evolving pads, soft sub-bass, sparse piano notes with long reverb. Instrumental, no percussion, seamless loop for deep work.
5. Acoustic singer-songwriter
Intimate acoustic folk, bittersweet. Fingerpicked guitar, soft male vocal, minimal production, room reverb. Verse-chorus, 90 BPM. About missing home.
6. Energetic workout track
High-energy EDM, aggressive and motivating. Punchy 128 BPM four-on-the-floor, distorted bass drop after a 16-bar build, stadium synth stabs. Instrumental.
7. Jazz café background
Smooth bossa nova jazz, relaxed and warm. Nylon-string guitar, brushed drums, upright bass, soft female vocal humming the melody. 110 BPM, like a Sunday afternoon café.
5 mistakes that ruin AI songs
- One-word genre prompts. "Make jazz" gives you elevator music. Add mood, instruments and tempo.
- Contradictory moods. "Happy but sad and epic but chill" confuses the model — pick one emotional core.
- Keeping the first generation. Always generate 3–4 variations of the same prompt; the best one is rarely the first.
- Overstuffing. More than ~10 descriptors and the model starts ignoring half of them. Two focused sentences beat a paragraph.
- Skipping the refine step. A good prompt gets you 80% there; the last 20% is variations and mix tweaks.
A practical workflow
- Write your prompt with the 7-part formula (two sentences max).
- Generate 3–4 variations. Pick the strongest 30 seconds of each.
- Re-prompt around the winner: "same vibe, warmer bass, slower build".
- Export the keeper. Done — most good tracks take 10–15 minutes total.
Note: if you plan to release AI-generated music commercially, check Muse's current licensing terms and your distributor's AI-content policy first — both change often.