How Loudly takes direction
The song path takes a style description of up to about 1000 characters, lyrics of up to about 3000, a vocal gender and an instrumental switch. The bed path takes genre, mood, tempo and length. Both live on the same platform, and the instrumental switch defaults to on — so if you want a vocal you have to ask for one rather than assume it.
Layer by layer
- Choose the model first. MANTA 1 for a song with vocals, VEGA for an instrumental bed. The rest of the brief follows from that one choice.
- The instrumental switch is on by default. Leaving it on while the lyrics box is full asks the platform to ignore one of your two instructions, and it will pick for you.
- Write the style field as sound, not story. Genre, tempo, instruments, vocal character. The lyrics carry the story; the style carries what it sounds like.
- Tag the lyric sections. A hook marked [Chorus] is placed where you wrote it. Long lyrics without markers get arranged for you.
- For beds, name the job and the exit. "Underscore for a product demo, steady energy, ends on a resolved chord" does more than "nice background music".
- Set length from the edit. Generate the duration you will actually use — trimming a bed usually removes the ending you asked for.
Loudly setups you can paste
One song on MANTA 1, two beds on the VEGA path. Paste the block for the model you picked.
Indie pop song · 108 BPM · vocals
Product demo bed · 100 BPM · 90s
Podcast intro · 88 BPM · 30s
More ready-made prompts across moods and engines: AI music prompt examples.
What to leave out
- Leaving the instrumental switch on with lyrics written. Two instructions, one outcome — and it is not the one you wanted.
- Writing a song brief for a bed. Verses, choruses and a big hook fight the thing the music is meant to support.
- Dense arrangements in a bed. Three melodic parts in the same octave is how a bed starts competing with dialogue.
- Ignoring the tail. An abrupt cut mid-phrase is the most common giveaway of a stock bed.
- Artist names. Genre plus mood plus instrumentation describes the track better than any reference.
The rule underneath all of them: describe the music, never the record. Reference an artist or a track and you inherit someone else's rights problem along with a vaguer brief.
Build a Loudly brief from a reference track
The hard part is the honest numbers: tempo, key, brightness, dynamics, and the instruments actually carrying the groove. Music DNA Studio measures them from any MP3, WAV, FLAC or MP4 — locally in your browser, nothing uploaded — then outputs a Loudly-shaped prompt alongside versions for the other fourteen engines. You make the creative calls; the spec underneath is measured, not guessed.
Both paths start from the same measured facts — tempo, brightness, dynamics, instrumentation — which is what makes switching between a song and a bed cheap.
Guides for the other engines
Eleven Music · Suno · Udio · Mureka · MiniMax Music · YuE2 · Soundverse · Treblo · Stable Audio · Soundraw · Mubert · Beatoven · AIVA · Loudly · MusicGen
Build The Loudly Brief
Drop in a reference, pick Loudly, copy the style block for MANTA 1 or the settings for a VEGA bed. Free, no signup, nothing leaves your device.
ANALYSE YOUR AUDIO — FREE →FAQ
Does Loudly generate vocals? ▾
Yes, on MANTA 1 — its song model produces full tracks with lyrics and a chosen vocal gender. The VEGA models are instrumental, and the instrumental switch defaults to on, so a sung track has to be asked for explicitly.
Which Loudly model should I use? ▾
MANTA 1 for a song with vocals; VEGA 2 for a production-quality bed and VEGA 3 when you want the arrangement decisions handled in-generation. The model choice decides which fields you fill, so make it first.
How much can I write in each field? ▾
Roughly 1000 characters of style and 3000 of lyrics. The style field is a description of the sound, not a brief — genre, tempo, instruments and vocal character fit comfortably in two sentences.
Can I use Loudly tracks commercially? ▾
Paid plans carry commercial rights; the free tier is a demo tier with short previews, one download a day and non-commercial terms. Check the licence attached to your plan before delivery, and note that Loudly keeps the copyright on generated tracks.
Why does my bed fight the voiceover? ▾
Density and midrange. Music that fills 200 Hz–2 kHz competes directly with speech; ask for sparse midrange, one lead instrument and no vocal-like textures.