SongGeneration Studio

AI Song Generation with Full Style Control using Tencent AI Lab's SongGeneration (LeVo) model.

Features

• Full Song Generation: Generate complete songs with vocals and instrumental tracks
• Style Control: Control gender, timbre, genre, emotion, instruments, and BPM
• Lyrics Support: Input your own lyrics with section markers (intro, verse, chorus, bridge, outro)
• Reference Audio: Use a reference audio file to guide the style of generation
• Multiple Output Modes: Generate mixed, vocals only, instrumental only, or separate tracks
• Web UI: Easy-to-use web interface for song creation

Requirements

• GPU: NVIDIA GPU with at least 10GB VRAM (24GB+ recommended for full quality)
• Storage: ~25GB for models and dependencies
• OS: Windows, macOS, or Linux

Installation

1. Open Pinokio
2. Find "SongGeneration Studio" in the app list
3. Click "Install"
4. Wait for the installation to complete (this may take a while due to large model downloads)

Usage

1. Click "Start" to launch the server
2. The Web UI will open automatically in your browser
3. Create your song:
   • Enter a song title
   • Add song sections (intro, verse, chorus, etc.) with lyrics
   • Configure style settings (genre, emotion, voice, instruments)
   • Optionally upload a reference audio file
4. Click "Generate Song" and wait for the generation to complete
5. Play and download your generated song

Song Sections

The following section types are supported:

Section	Description
intro-short / intro / intro-long	Song introduction
verse / verse-short / verse-long	Main verses
chorus / chorus-short / chorus-long	Song chorus
bridge / bridge-short / bridge-long	Bridge section
outro-short / outro / outro-long	Song ending
break	Musical break
solo	Instrumental solo

Style Settings

• Voice Gender: Female or Male
• Timbre: Bright, Dark, Soft, Powerful, Warm, Clear, Raspy, Smooth
• Genre: Pop, Rock, Metal, Jazz, R&B, Folk, Dance/Electronic, Reggae, Chinese Style
• Emotion: Happy, Sad, Energetic, Romantic, Angry, Peaceful, Melancholic, Hopeful
• Instruments: Specify instruments (e.g., "piano and drums", "electric guitar, bass, drums")
• BPM: Tempo (60-200)

API Documentation

Generate a Song

// JavaScript
const response = await fetch('/api/generate', {
  method: 'POST',
  headers: { 'Content-Type': 'application/json' },
  body: JSON.stringify({
    title: "My Song",
    sections: [
      { type: "intro-short", lyrics: null },
      { type: "verse", lyrics: "Walking through the city lights" },
      { type: "chorus", lyrics: "We are alive, we are on fire" },
      { type: "outro-short", lyrics: null }
    ],
    gender: "female",
    timbre: "bright",
    genre: "Pop",
    emotion: "happy",
    instruments: "piano and drums",
    bpm: 120,
    output_mode: "mixed"
  })
});
const { generation_id } = await response.json();

# Python
import requests

response = requests.post('http://localhost:8000/api/generate', json={
    'title': 'My Song',
    'sections': [
        {'type': 'intro-short', 'lyrics': None},
        {'type': 'verse', 'lyrics': 'Walking through the city lights'},
        {'type': 'chorus', 'lyrics': 'We are alive, we are on fire'},
        {'type': 'outro-short', 'lyrics': None}
    ],
    'gender': 'female',
    'timbre': 'bright',
    'genre': 'Pop',
    'emotion': 'happy',
    'instruments': 'piano and drums',
    'bpm': 120,
    'output_mode': 'mixed'
})
generation_id = response.json()['generation_id']

# cURL
curl -X POST http://localhost:8000/api/generate \
  -H "Content-Type: application/json" \
  -d '{
    "title": "My Song",
    "sections": [
      {"type": "intro-short", "lyrics": null},
      {"type": "verse", "lyrics": "Walking through the city lights"},
      {"type": "chorus", "lyrics": "We are alive, we are on fire"},
      {"type": "outro-short", "lyrics": null}
    ],
    "gender": "female",
    "timbre": "bright",
    "genre": "Pop",
    "emotion": "happy",
    "instruments": "piano and drums",
    "bpm": 120,
    "output_mode": "mixed"
  }'

Check Generation Status

curl http://localhost:8000/api/generation/{generation_id}

Download Generated Audio

curl http://localhost:8000/api/audio/{generation_id}/0 --output song.wav

API Endpoints

Endpoint	Method	Description
/api/health	GET	Health check and GPU status
/api/models	GET	List available models
/api/presets	GET	Get available style presets
/api/generate	POST	Start a new song generation
/api/generation/{id}	GET	Get generation status
/api/generations	GET	List all generations
/api/audio/{id}/{track}	GET	Download generated audio track
/api/upload-reference	POST	Upload a reference audio file

Credits

• SongGeneration (LeVo): Tencent AI Lab
• Pinokio Launcher: This installer

License

This launcher is provided for educational and research purposes. Please refer to the original SongGeneration repository for licensing information.