菜单
Use when polishing AI-generated audio — per-stem processing with SoX cleanup, EQ, compression, and remix. Processes stems or full mixes into polished WAVs ready for mastering.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
| name | daw-master:mix-engineer |
| description | Use when polishing AI-generated audio — per-stem processing with SoX cleanup, EQ, compression, and remix. Processes stems or full mixes into polished WAVs ready for mastering. |
| version | 0.1.0 |
| author | Ported from bitwize-music-studio |
| license | MIT |
| metadata | {"hermes":{"tags":["audio","mix","polish","stems","sox","ffmpeg"],"related_skills":["daw-master:sox-engine","daw-master:ffmpeg-audio","daw-master:audio-analyzer","daw-master:mastering-engineer"]}} |
Polishes raw AI-generated audio (Suno, etc.) by processing per-stem WAVs with targeted cleanup, EQ, and compression using SoX + FFmpeg engines, then remixes into a polished stereo WAV ready for mastering.
Role: Per-stem processing, noise reduction, frequency cleanup, dynamic control, stem remixing.
Not role: Creative production, lyrics, generation.
Primary tools: sox-engine.transform, ffmpeg-audio.transform, audio-analyzer.analyze
stems/ directory is present, process each stem independently for targeted processing impossible on a full mix.polished/ — originals never modified.sox-engine — single-track effects (EQ, compand, normalize, fades, highpass/lowpass/bandpass)ffmpeg-audio — multi-track mixing, loudnorm, advanced dynamicsaudio-analyzer — feature extraction for analysis decisionsNo external Python packages beyond those already installed.
{audio_root}/artists/[artist]/albums/[genre]/[album]/
├── stems/
│ ├── 01-track-name/
│ │ ├── 0 Lead Vocals.wav
│ │ ├── 1 Backing Vocals.wav
│ │ ├── 2 Drums.wav
│ │ ├── 3 Bass.wav
│ │ └── ... (up to 12 stem types)
├── polished/ ← output
└── mastered/ ← mastering-engineer output
analyze_mix_issues(album_path: str) -> DictScan for stems or full mix and detect issues.
{
"source_mode": "stems" | "full_mix",
"tracks": [...],
"issues": {
"noisy_tracks": [...],
"muddy_tracks": [...],
"harsh_tracks": [...],
"clicky_tracks": [...],
"sub_bass_rumble": [...]
}
}
polish_audio(album_path: str, genre: str = "default", dry_run: bool = False) -> DictProcess stems or full mix with genre-appropriate settings. Writes to polished/.
{
"mode": "stems" | "full_mix",
"tracks_processed": 5,
"output_dir": "/path/to/polished",
"commands": [...],
"success": bool
}
polish_album(album_path: str, genre: str = "default") -> DictEnd-to-end: analyze → polish → verify.
_generate_noise_profile(input_wav: str) -> strAuto-generates SoX noise profile from quietest 100ms window.
_detect_clicks(audio_data, sr) -> List[float]Detects transient clicks (> 6σ amplitude spikes).
Full chains for all 12 stem types (Vocals, Backing Vocals, Drums, Bass, Guitar, Keyboard, Strings, Brass, Woodwinds, Percussion, Synth, Other): references/stems.md
Quick reference:
| Stem | Key steps |
|---|---|
| Lead Vocals | NR → +2dB @ 3kHz → -2dB @ 7kHz → compress |
| Backing Vocals | NR → +1dB @ 3kHz → -2.5dB @ 7kHz → compress tighter |
| Drums | Click removal → compress gentle |
| Bass | Highpass 30Hz → mud cut 200Hz → compress |
| Guitar | Highpass 80Hz → mud cut 250Hz → +1.5dB @ 3kHz → compress |
| Keyboard | Highpass 40Hz → mud cut 300Hz → +1dB @ 2.5kHz → compress |
| Strings | Highpass 35Hz → +1dB @ 3.5kHz → wide stereo → very light compress |
| Brass | Highpass 60Hz → +1.5dB @ 2kHz → compress |
| Woodwinds | Highpass 50Hz → +1dB @ 2.5kHz → compress |
| Percussion | Highpass 60Hz → click removal → +1dB @ 4kHz → stereo width |
| Synth | Highpass 80Hz → +1dB @ 2kHz → stereo width → compress |
| Genre | Vocal | Bass | Drums | High-mid |
|---|---|---|---|---|
| Default | 0 dB | 0 dB | 0 dB | –2 dB @ 7 kHz |
| Hip-Hop/Rap | +1 dB | +1 dB | +0.5 dB | aggressive |
| Rock/Metal | 0 dB | 0 dB | +0.5–1 dB | –2.5 to –3 dB |
| EDM/Electronic | 0 dB | +0.5–1 dB | +0.5–1 dB | lighter |
| Folk/Country | +0.5 dB | 0 dB | +0.5 dB | –1.5 dB |
| Ambient/Lo-Fi | 0 dB | 0 dB | 0 dB | –1.5 dB; lighter NR |
from daw_master.mastering_engineer import master_audio
master_audio(album_path, source_subfolder="polished", genre=genre)
Before handoff: no clipping (peak < –0.1 dBFS), all samples finite, noise floor reduced, no artifacts, files in polished/.
Use when working with the Sleepy Circuits Hypno 2 video synthesizer/resampler — 2-channel video mixer/looper, shader engine, MIDI CC mapping, CV/modulation, sampler. Firmware v0.0.163+.
Hardware instrument skills — reference guides for each device in the music studio setup.
GLSL fragment shader techniques adapted for the Hypno 2 video synthesizer — single-pass generative visuals and video effects optimized for Raspberry Pi 5, 5-uniform limit, CC-mapped parameters
Korg KAOSS DJ — USB DJ controller with X-Y touchpad, Serato DJ Intro integration, KAOSS effects, EQ, crossfader, and MIDI on Linux
Korg KAOSSILATOR dynamic phrase synthesizer (2007 model) — X-Y touchpad, 100 programs, gate arpeggiator, phrase loop recording, USB MIDI on Linux
Use when working with the Korg kaossilator 2S dynamic phrase synthesizer — X-Y touchpad, 150 programs, 50 arp patterns, loop recording, audio player, master recorder, microSD storage, and USB MIDI on Linux.