MiniMax Music 2.6 on Sume: vocals from a 10-2,000 character prompt

MiniMax Music 2.6 is a Music Router model on Sume with the id minimax-music-2.6. It makes a song with vocals: MiniMax writes the lyrics from your prompt (Sume sends lyrics_optimizer: true), so you send a style brief and no lyrics field. The prompt must be 10 to 2,000 characters, the row takes no duration_seconds, and it bills one flat price per track.

Fal, which serves the model, describes it as creating complete tracks with singing, backing music and arrangements, lists $0.15 per track, allows commercial use, and returns MP3 (fal MiniMax Music 2.6, read 2026-10-14). A separate post covers MiniMax closing its own music API to new users.

What the row takes

Source: the Music Router docs and the Sume pricing code on main, read 2026-10-14.

Cost

Fal lists $0.15; at Sume's list times 1.25 that is $0.1875 per track. Confirm the credits on GET /v1/music-router/models or the API rate card.

Write the brief like a style line

Since the lyrics come from the prompt, name genre, tempo, mood, vocal type and the topic of the words in one line. Short prompts under 10 characters fail. To make the length predictable, use a row that takes duration_seconds, such as Stable Audio 2.5, though that one has no vocals. For a flat clip see Lyria 2.

curl -X POST https://api.sume.com/v1/music-router/generate \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "minimax-music-2.6",
    "prompt": "City pop, 80s retro, groovy synth bass, warm female vocal, 104 BPM, nostalgic urban night, lyrics about a last train home"
  }'