Sound Effects

Generate original sound effects and ambience from a text brief, fully offline. A thin wrapper drives Stable Audio Open Small (a diffusion text-to-audio model Stability trained to be "better at sound effects and field recordings than music"); you (the model) shape the brief into a concrete prompt, generate a couple of candidates, and let the user pick.

Pairs with audio-theater (which cues SFX into a radio drama / podcast) and bg-music (music beds). For music, use bg-music; this skill is for non-musical sound.

flowchart LR
  Brief["brief: what sound, materials, space"] --> Gen["generate_sfx.py<br/>Stable Audio Open Small (MLX), 8 steps"]
  Gen --> Out["candidate mp3/wav (<=11s stereo)"]