# MiniMax Music 2.6 - Core.Today AI API

> Generate full-length songs or instrumentals (up to ~6 minutes) from a text prompt. 99%+ accurate BPM/key control, 14+ structure tags, auto-generated lyrics, and instrumental-only mode.

- **Provider**: MiniMax
- **Model ID**: minimax/music-2.6
- **Category**: Audio & TTS
- **Credits**: 350 per generation (1 audio file)
- **Speed**: Medium
- **Quality**: High

## Features
- Full-length songs up to ~6 minutes (most 2-4 min)
- 99%+ accurate BPM and key control via prompt
- 14+ structure tags: [Intro], [Verse], [Chorus], [Hook], [Drop], [Bridge]...
- Auto lyric generation from the prompt (lyrics_optimizer)
- Instrumental-only mode (is_instrumental)
- Realistic vocal timbre with breathing and pitch transitions
- Up to 44.1kHz / 256kbps / WAV output

## Use Cases
- Original song production from lyrics
- Background music with exact BPM for video editing
- Game and app soundtracks
- Jingles and brand music
- Demo tracks for songwriters

## API Endpoint

Base URL: https://api.core.today/v1
Create Prediction: POST /predictions
Get Status: GET /predictions/{job_id}

## Authentication
Header: X-API-Key: YOUR_API_KEY

## Input Parameters

### Optional
- **prompt**: string - Style description (max 2,000 chars): key, BPM, genre, mood, vocal style, instruments, production. e.g. 'E minor, 90 BPM, acoustic guitar ballad, male vocal, emotional'. Required when is_instrumental is true
- **lyrics**: string - Lyrics (max 3,500 chars). Use \n for line breaks, \n\n for pauses between sections. Structure tags: [Intro], [Verse], [Pre Chorus], [Chorus], [Hook], [Drop], [Bridge], [Solo], [Build Up], [Inst], [Interlude], [Break], [Transition], [Outro]
- **is_instrumental**: boolean (default: false) - Generate instrumental music without vocals. Prompt is required, lyrics are not needed
- **lyrics_optimizer**: boolean (default: false) - Automatically generate lyrics from the prompt when lyrics are empty — create a complete song from just a style description
- **sample_rate**: integer (default: 44100) - Sample rate for the generated music
  Options: 16000, 24000, 32000, 44100
- **bitrate**: integer (default: 256000) - Bitrate for the generated music
  Options: 32000, 64000, 128000, 256000
- **audio_format**: string (default: mp3) - Output format. Use wav for production work
  Options: mp3, wav, pcm

## Examples

### Acoustic Ballad from Lyrics
Emotional ballad with exact key/BPM and structured lyrics
```json
{
  "model": "minimax/music-2.6",
  "input": {
    "prompt": "E minor, 90 BPM, acoustic guitar ballad, male vocal, emotional, intimate studio feel",
    "lyrics": "[Verse]\n밤하늘의 별을 세다가\n너의 이름을 떠올렸어\n\n[Chorus]\n아직도 기억해 그날의 노래\n바람에 실려 오는 멜로디\n\n[Outro]\n그날의 노래...",
    "audio_format": "mp3",
    "bitrate": 256000,
    "sample_rate": 44100
  }
}
```

### Full Song from One Line
Auto-generate lyrics and a complete song from a style prompt only
```json
{
  "model": "minimax/music-2.6",
  "input": {
    "prompt": "Upbeat K-pop, 120 BPM, bright female vocal, catchy synth hook, summer vibes",
    "lyrics_optimizer": true
  }
}
```

### Instrumental Background Track
Vocal-free lo-fi track for videos or study sessions
```json
{
  "model": "minimax/music-2.6",
  "input": {
    "prompt": "Lo-fi hip hop, 80 BPM, warm vinyl texture, mellow piano, relaxed study vibe",
    "is_instrumental": true,
    "audio_format": "mp3"
  }
}
```

## Response Format
```json
{
  "job_id": "abc123",
  "status": "pending | processing | completed | failed",
  "result": "URL or data (when completed)"
}
```

## Usage Flow
1. POST /predictions with model and input -> receive job_id
2. GET /predictions/{job_id} -> poll until status is completed or failed
3. Result contains output URL(s)

## Tips
- Specify exact key and BPM in the prompt (e.g. 'E minor, 90 BPM') — output matches 99%+ of the time
- Order your prompt: key, BPM, genre, mood, vocal style, instruments, production character
- Use 14+ structure tags ([Verse], [Chorus], [Hook], [Drop], [Bridge]...) and \n\n for pauses between sections
- Set lyrics_optimizer: true to create a complete song from just a style description
- English and Mandarin vocals are strongest; other languages work with less consistent pronunciation
- Use wav + 44100/256000 for production; each generation is unique even with identical inputs

## Documentation
https://replicate.com/minimax/music-2.6