stable-audio-3
Adapt music, stem, ambience, Foley, sound effect, audio-to-audio, inpainting, or continuation cues into Stable Audio 3 prompts with source, action, production character, duration, and model mode.
SKILL.mdskills/stable-audio-3/SKILL.md
GitHubStable Audio 3 Prompting
Choose model path
medium: higher-quality music and general audiosmall-music: fast musicsmall-sfx: sound effects
Music
Describe genre, important instruments, mood and energy, and BPM when useful. Stable Audio 3 is for instrumental or non-intelligible vocal textures, not dialogue.
Stem or solo instrument
Start with TrackType: Instrument when isolation matters. State instrument, genre context, technique, mood, tempo, recording environment, and effects.
SFX
State:
- physical source
- exact triggering action
- duration or decay behavior
- distance and microphone perspective
- room or exterior character
- processing only when needed
TrackType: SFX is useful for isolated effects.
Existing audio
For audio-to-audio, inpainting, or continuation, preserve source-file identity and record the intended edit region or influence strength as execution metadata. Keep the prompt plausible relative to surrounding audio.
Output
Save under 04_generation/prompts/stable-audio-3/<cue-id>.md with MODEL, MODE, DURATION, PROMPT, and optional NEGATIVE_PROMPT.
Done
The prompt names a physically or musically coherent source and duration, and does not ask Stable Audio 3 to produce intelligible spoken dialogue.