1---2name: audio3description: Process, enhance, and convert audio files with noise removal, normalization, format conversion, transcription, and podcast workflows.4---5
6## Requirements
7
8**Required:**
9- `ffmpeg` / `ffprobe` — core audio processing
10
11**Optional (for advanced features):**
12- `sox` — additional noise reduction
13- `whisper` — local transcription (or use API)
14- `demucs` — stem separation
15
16## Quick Reference
17
18| Situation | Load |
19|-----------|------|
20| FFmpeg commands by task | `commands.md` |
21| Loudness standards by platform | `loudness.md` |
22| Podcast production workflow | `podcast.md` |
23| Transcription workflow | `transcription.md` |
24
25## Core Capabilities
26
27| Task | Method |
28|------|--------|
29| Convert formats | FFmpeg (`-acodec`) |
30| Remove noise | FFmpeg filters or SoX |
31| Normalize loudness | `ffmpeg-normalize` or `-af loudnorm` |
32| Transcribe | Whisper → text, SRT, VTT |
33| Separate stems | Demucs (vocals, drums, bass, other) |
34
35## Execution Pattern
36
371. **Clarify goal** — What format? What loudness? What platform?
382. **Analyze source** — `ffprobe` for codec, sample rate, channels, duration
393. **Process** — FFmpeg/SoX for transformation
404. **Verify** — Check output plays, meets specs, sounds correct
415. **Deliver** — Provide file to user
42
43## Common Requests → Actions
44
45| User says | Agent does |
46|-----------|------------|
47| "Convert to MP3" | `-acodec libmp3lame -q:a 2` |
48| "Remove background noise" | Apply highpass/lowpass or dedicated denoiser |
49| "Normalize for podcast" | `-af loudnorm=I=-16:TP=-1.5:LRA=11` |
50| "Transcribe this" | Whisper → output SRT/VTT/TXT |
51| "Extract audio from video" | `-vn -acodec copy` or re-encode |
52| "Make it smaller" | Lower bitrate: `-b:a 128k` or `-b:a 96k` |
53| "Speed up 1.5x" | `-af atempo=1.5` |
54
55## Format Quick Reference
56
57| Format | Use Case | Quality |
58|--------|----------|---------|
59| WAV | Master, editing | Lossless |
60| FLAC | Archive, audiophile | Lossless compressed |
61| MP3 | Universal sharing | Lossy, 128-320 kbps |
62| AAC/M4A | Apple, podcasts | Lossy, efficient |
63| OGG/Opus | WhatsApp, Discord | Lossy, very efficient |
64
65## Quality Defaults
66
67- **Podcast:** -16 LUFS (Spotify), -19 LUFS (Apple)
68- **Music:** -14 LUFS (Spotify), -16 LUFS (Apple Music)
69- **MP3 quality:** VBR `-q:a 2` (~190 kbps) or CBR `-b:a 192k`
70- **Sample rate:** 44.1kHz for music, 48kHz for video sync
71
72## Scope
73
74This skill:
75- Processes audio files user explicitly provides
76- Runs FFmpeg commands on user request
77- Does NOT access cloud services without user knowing
78- Does NOT store files persistently (user manages their files)