music-sfx
Installation
SKILL.md
music-sfx
Run every graph with the bundled client: python scripts/comfy_client.py --url $COMFY_URL run workflows/<file> --set NODE.INPUT=VALUE --out out/. Upload input files first with python scripts/comfy_client.py upload <file>. Before the first run on a server, check it with python scripts/preflight.py --url $COMFY_URL workflows/<file>. Node packs, model files, VRAM and weight licenses for this skill are in references/requirements.md. Use open-weights models only, and never swap in a hosted or closed model.
| Need | Workflow | Set |
|---|---|---|
| Music, or a song with lyrics | music_ace_step15.api.json |
4.tags (genre, instruments, BPM, mood), 4.lyrics (or [instrumental]), 6.seconds, 4.seed |
| Alternate music model | music_minimax_music3.api.json |
13.caption, 13.lyrics |
| SFX / ambience | sfx_stable_audio_open.api.json |
3.text, 5.seconds (up to about 47 s) |
| SFX, newer model | sfx_stable_audio_3.api.json |
6.text, 11.seconds |
Prompt rules
Condensed from Fuobe's audio rules; the full copy is in references/fuobe-audio-prompt-rules.md.
- Music models take a style description, not a scene: genre, tempo or BPM, key instruments, mood, structure. Lyrics go in the lyrics field only, never in the tags.
- SFX models take one sound per prompt in concrete physical words (material, action, space, distance): "heavy wooden door slams in a stone hall, short reverb". Don't send music words to an SFX model.
- Set the duration to the cut length plus about 1 s for tails. A bed under narration should be sparse, with no vocals.
- Ignore the closed or hosted model rows in the reference file; this skill is open weights only.
- Weight licenses: check Stable Audio Open's license (believed to be the Stability AI Community License) before any commercial use. The MiniMax Music 3 license is unverified.