Generate Music (MiniMax)

Generate music synchronously and save the resulting audio.

Instructions

1. Parse $ARGUMENTS. Use the free text as prompt; accept optional --lyrics, --model, --format, --response, --instrumental, --region, and --output values.
2. Select the endpoint from --region:
   • global (default): https://api.minimax.io/v1/music_generation
   • cn: https://api.minimaxi.com/v1/music_generation
3. Read the Bearer token from MINIMAX_API_KEY. Never print the token.
4. Default model to music-3.0. Accepted generation models are music-3.0, music-2.6, music-3.0-free, and music-2.6-free.
5. Build an application/json request with model, prompt, lyrics, stream: false, output_format, audio_setting, lyrics_optimizer, and is_instrumental. The audio format may be mp3, wav, or pcm. For the China endpoint, include aigc_watermark only when explicitly requested.
6. POST the request and check base_resp.status_code == 0.
7. Read data.status: 1 means processing and 2 means completed. Because the API is synchronous, report a non-completed response instead of polling a separate endpoint.
8. On completion, read data.audio. For output_format: url, download the URL promptly because it expires after 24 hours. For output_format: hex, decode it with xxd -r -p into the requested output path.

Request

# region=global -> api.minimax.io; region=cn -> api.minimaxi.com
curl -sS -X POST "https://api.minimax.io/v1/music_generation" \
  -H "Authorization: Bearer $MINIMAX_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "music-3.0",
    "prompt": "Warm cinematic strings that build toward a hopeful finale",
    "lyrics": "",
    "stream": false,
    "output_format": "url",
    "audio_setting": { "format": "mp3" },
    "lyrics_optimizer": true,
    "is_instrumental": true
  }'

Response

{
  "data": {
    "status": 2,
    "audio": "https://example.invalid/generated-music.mp3"
  },
  "base_resp": { "status_code": 0, "status_msg": "success" }
}