Music Generation Skill (Lyria)

Generate instrumental music from text prompts using Google Lyria API via Vertex AI.

Capabilities

• Generate 30-second WAV instrumental tracks optimized for short-form content
• Support style, mood, key, and instrument specifications
• Save generated music to workspace with custom or timestamped filenames
• Iterate based on user feedback

Use Cases

• TikTok/YouTube Shorts — background music for 15-30s videos
• Instagram Reels — quick musical intros/outros
• Video transitions — short audio bridges
• Loops — repeating segments for longer content

Limitations

• 30 seconds max per generation (Lyria constraint)
• Short-form only — designed for TikTok/Reels/Shorts, not full songs
• Instrumental only (no vocals/lyrics)
• Bearer token expires hourly (requires periodic refresh)
• English prompts recommended

---

First-Time Setup

When using this skill for the first time on a machine, follow these steps:

Step 1: Create Directory Structure

# Create the lyria folder structure in workspace
mkdir -p ~/.openclaw/workspace/lyria/generated_music

Step 2: Install gcloud CLI (if not installed)

Check if gcloud is installed:

gcloud --version

If not installed, install Google Cloud SDK:

macOS (Homebrew):

brew install --cask google-cloud-sdk

Linux:

curl https://sdk.cloud.google.com | bash
exec -l $SHELL

Verify installation:

gcloud --version

Step 3: Authenticate with Google Cloud

gcloud auth login

Follow the browser prompt to authenticate.

Step 4: Get Your Configuration Values

Get project_id:

gcloud config get-value project

Get location:

gcloud config get-value compute/region

If no region is set, use: us-central1 (recommended for Lyria)

Get bearer_token:

gcloud auth print-access-token

⚠️ Note: This token expires in approximately 1 hour. You'll need to refresh it periodically.

Step 5: Create Config File

Create ~/.openclaw/workspace/lyria/config.json with your values:

{
  "project_id": "your-project-id-here",
  "location": "us-central1",
  "bearer_token": "ya29.a0AfH...your-token-here",
  "output_dir": "~/.openclaw/workspace/lyria/generated_music"
}

Replace the placeholders with your actual values from Step 4.

---

Usage

Quick Generate (Shell Wrapper)

The shell wrapper is the recommended way to generate music. It reads all configuration from the config file automatically.

./scripts/music-gen.sh "<prompt>" [name]

Arguments:

• prompt (required): Text description of the music you want
• name (optional): Custom filename (without extension). If omitted, uses timestamp.

Examples:

# With custom name:
./scripts/music-gen.sh "chill lo-fi piano C minor" "my_relaxing_track"
# Output: ~/.openclaw/workspace/lyria/generated_music/my_relaxing_track.wav

# With timestamp (default):
./scripts/music-gen.sh "energetic rock guitar solo"
# Output: ~/.openclaw/workspace/lyria/generated_music/music_20260302_143022.wav

Direct Python Usage

For more control or integration with other tools, call the Python script directly:

python3 scripts/music-gen.py <config_file> "<prompt>" [name]

Arguments:

• config_file (required): Path to config.json
• prompt (required): Text description of the music
• name (optional): Custom filename (without extension)

Example:

python3 scripts/music-gen.py ~/.openclaw/workspace/lyria/config.json "jazz saxophone smooth" "evening_jazz"

---

Prompt Guidelines

Good prompts include:

• Genre/Mood: "chill lo-fi", "energetic rock", "melancholic jazz", "epic orchestral"
• Key: "C minor", "F major", "E Phrygian"
• Instruments: "piano, strings, soft drums", "electric guitar, bass, drums"
• Tempo/Feel: "slow and relaxing", "fast and driving", "mid-tempo groove"

Example prompts:

• "A calm acoustic folk song in C minor with gentle guitar melody and soft strings, no drums"
• "Upbeat electronic dance music with strong synth bass and driving beats, 128 BPM feel"
• "Melancholic jazz piano in F minor with soft brush drums and upright bass"
• "Epic cinematic orchestral with brass and strings, heroic and uplifting mood"

---

Refreshing Your Bearer Token

Since tokens expire hourly, you may need to refresh during long sessions:

1. Get new token:

   gcloud auth print-access-token
   

2. Update config.json with the new token

3. Continue generating music

---

Workflow for Agents

When a user asks you to generate music:

1. Check if setup is complete:

   • Verify ~/.openclaw/workspace/lyria/config.json exists
   • If not, guide user through First-Time Setup above

2. Check if token is valid:

   • If generation fails with 401/403, token expired
   • Guide user to refresh token (see above)

3. Refine the prompt:

   • Ask user for style, mood, instruments, key if not provided
   • Help craft a descriptive prompt

4. Generate:

   ./scripts/music-gen.sh "<refined_prompt>" "<descriptive_name>"
   

5. Deliver:

   • Send the generated .wav file to user
   • Confirm success

6. Iterate (if needed):

   • Ask: "Want any changes? Faster? Different instruments?"
   • Update prompt, regenerate, deliver

---

Error Handling

Error	Cause	Solution
Config file not found	First-time setup incomplete	Follow First-Time Setup steps
401 Unauthorized	Bearer token expired or invalid	Refresh token: gcloud auth print-access-token
403 Forbidden	No access to Lyria API	Enable Vertex AI API in Google Cloud Console
404 Not Found	Project or location incorrect	Verify project_id and location in config
No predictions	API issue or invalid prompt	Retry with modified prompt

---

File Structure

~/.openclaw/workspace/
└── lyria/
    ├── config.json              # User credentials and settings
    └── generated_music/         # All generated audio files
        ├── music_20260302_143022.wav
        ├── my_relaxing_track.wav
        └── ...

---

API Reference

• Model: lyria-002
• Endpoint: https://{location}-aiplatform.googleapis.com/v1/projects/{project}/locations/{location}/publishers/google/models/lyria-002:predict
• Docs: https://ai.google.dev/gemini-api/docs/music-generation
• Pricing: $0.06 per 30 seconds of generated music

---

Best Practices

1. Use specific prompts — "jazz piano in F minor" works better than "jazz music"
2. Specify what to exclude — "no drums" or "no vocals" helps steer output
3. Generate variations — Create 2-3 versions and let user pick favorite
4. Monitor costs — At $0.06 per 30s, 10 generations = $0.60
5. Refresh token proactively — If session >45 min, get fresh token before generating

---

Example Complete Session

User: Generate some lo-fi music for studying

Agent: I'll create a chill lo-fi track for you. First, let me check your setup...
[Checks config exists - yes]
[Checks token validity - yes]

Generating: "Chill lo-fi hip hop with soft piano melody, warm vinyl crackle, relaxed beats, no vocals"
...
Generated: ~/.openclaw/workspace/lyria/generated_music/lofi_study_track.wav
[Sends file]

Done! 30 seconds of chill lo-fi. Want me to make it longer by generating a continuation, or try a different vibe?