prompting.

in the plug-in you tap a start point, up to three genres and a few sound words — the part, the tempo and your track’s key are already set. this line is right there under them, and it’s editable: change a word in it and it’s sent as you leave it.

• TrackType: Instrument asks stable audio 3 for an isolated part, not a full mix.
• the part is a short head per role — Drum loop, Bass groove loop, solo bass — and your words hang off it.
• genres — up to three, in the order you pick them: all of them say the style (funk, soul and disco style), the first one leads the Genre: field. every part plays the same genres, so it sounds like one band — or give one part its own.
• Instruments follows your sound words: organ on keys is an organ, flute on a lead is a flute, upright on bass is an upright bass.
• BPM is logic’s tempo. the length is asked for in whole bars (one bar extra, cut back on the downbeat), so the part lands on your grid without stretching.
• Key is your track’s, detected as it plays (or the one you pick). it’s a strong hint, not a lock — it lands on your key or a fifth away; re-roll if it clashes. never sent for drums.

stable audio 3 reads genre, instrument and rhythm first, then mood and energy, then how it’s recorded. stems already supplies the part, the genre and the tempo — your words are the middle: how it plays and how it sounds.

good words:

• dusty breakbeat with ghost notes
• round fingered bass, sparse, laid back
• wurlitzer stabs, offbeat, warm tape
• wide analog pad, slow swells
• plucky arp, bright, short delay

keep it short. under a dozen words. the whole line is kept under 45 words; past that, your words give way from the end — never the structure around them. long prompts drift.

in the plug-in these are chips — sound, playing, feel, recording — tap to add, tap again to take away. each part has a few start points (funk pocket, 808 trap sub, neo soul rhodes…) that set the genres and words in one tap, for a good first take.

or edit the line itself. it’s sent as written — stems only takes vocal words out, keeps it under 45 words (from the end of your words, never the key and bpm), and keeps its Key: and BPM: in step with your track. take the bpm out and it tells you what that costs, with one tap to put it back. rebuild goes back to the built line.

• singing, lyrics, vocals. it can’t — it makes vocal-like textures, never words. the plug-in takes vocal words out of your prompt before it sends it.
• a key in your words. stems already sends your track’s key in its own field — set it with the key chip instead.
• “no …”. negative words don’t do anything on this model. say what it plays instead.

genres it knows by name: house · deep house · techno · minimal techno · drum and bass · dubstep · uk garage · breakbeat · trap · hip hop · boom bap · lo-fi hip hop · trip hop · r&b · neo soul · soul · funk · disco · jazz · jazz fusion · blues · latin · afrobeat · reggae · dub · reggaeton · pop · synthwave · indie pop · indie rock · rock · garage rock · punk · post-punk · shoegaze · dream pop · metal · country · folk · ambient · cinematic.

• feel — off · low · mid · high. off writes from words alone. turned up, the part starts from your track itself so it sits on your groove; high follows it closest. if it starts sounding like the whole band, say what it plays alone (solo bass) and keep feel low.
• takes — 1 · 2 · 3 per press, each about five seconds. re-roll keeps your words and changes the seed. takes of one prompt are close cousins — change a word to get somewhere new.
• steps — 8 · 6 · 4. the model is built for 8; fewer is faster and rougher.