---
name: audio-and-music-production
description: "Umbrella for songwriting, AI music generation, local song models, and audio/spectrogram analysis workflows."
version: 1.0.0
author: Hermes Agent
license: MIT
metadata:
  hermes:
    tags: [audio, music, songwriting, ai-music, spectrogram, analysis, generation]
    related_skills: [audiocraft, edge-tts, spotify]
---

# Audio and Music Production

Use this class-level skill when the user asks to write songs, craft lyrics, create AI-music prompts, generate songs locally, or analyze audio features/visualizations.

## Pick the workflow

- **Songwriting craft** — lyrics, rhyme, meter, emotional arc, structure, genre adaptation, parody, Suno-style prompt writing.
- **AI song generation** — local/open-source models such as HeartMuLa or other text/lyrics-to-song systems.
- **Audio analysis/visualization** — spectrograms, mel/chroma/MFCC/loudness/tempogram plots, and feature screenshots via `songsee` or similar tools.

## Songwriting and AI music prompts

Start with intent: story, point of view, genre, emotional arc, and performance style. Prefer singable meter and strong concrete imagery. For AI music systems, separate lyrics from style tags and keep instrumental/production directions concise.

## HeartMuLa / open-source song generation

Use HeartMuLa when the user wants a local/offline Suno-like model from lyrics + tags. Watch for Python and dependency compatibility issues; preserved details include RoPE cache and HeartCodec loading patches.

## Audio visualization with `songsee`

```bash
songsee track.mp3 -o spectrogram.png
songsee track.mp3 --viz spectrogram,mel,chroma,hpss,selfsim,loudness,tempogram,mfcc,flux
songsee track.mp3 --start 12.5 --duration 8 -o slice.jpg
```

Use visualizations to inspect arrangement, tempo, loudness, similarity, and spectral content.

## Preserved source details

Full absorbed source skills are preserved in `references/absorbed-*.md` for detailed songwriting guidance, install commands, compatibility patches, and CLI options.