AI game music generator: one track per level, cut to WAV

An AI game music generator writes a track from a text brief, so you can make a menu theme, level music and a boss fight without a composer: one brief per scene, with the mood, tempo, key and instruments that scene needs. You then cut each track to length and export it in the format your engine imports; loop points are yours to set.

On Sume the Music Router generates one track per request at $0.125 per audio generation, plus a 5.5% agent fee by default. The brief format is from Music 1.0 and the WAV step from Timeline audio, read on 2026-09-29.

How do I write a brief for a level, menu or boss track?

Write one brief per scene. The seven-axis brief from the Music docs (mood, genre, tempo, key, instruments, one named moment, era) is laid out with examples in AI music prompt examples; the docs call the axes creative directions, not guaranteed settings, so listen to each take. For a game:

• Make areas distinct: for contrasting scenes, the docs say to change the genre family, the tempo (at least 12 BPM apart) and the lead instrument.
• Keep one theme recognizable. When you want a consistent score, keep those axes close and change only the arc.
• Name the moment the energy lifts, such as a boss's second phase: "drums enter at 0:40".
• Send concept art as image_url (a public HTTPS image) to condition the track on the scene.
• Close with "Instrumental, no vocals." Put anything else to avoid in the prompt too; a non-empty negative_prompt returns 400.

curl -X POST https://api.sume.com/v1/music-router/generate \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: level-3-ice-caves-take-1" \
  -d '{
    "prompt": "Cold, curious and slightly tense chiptune ambient, 96 BPM, E minor. Square-wave arpeggio, soft noise hats, glassy pad. Calm first half; bass pulse enters at 0:45. A 2-minute track. Instrumental, no vocals.",
    "image_url": "https://example.com/concept/ice-caves.png"
  }'

Can an AI game music generator make seamless loops or stems?

Not on Sume. The request's creative fields are a prompt, an optional image_url and an optional model; the docs list no loop, stem or duration field, and the result is an audio artifact with no loop points or separate stems documented. So:

• Set loop points yourself, in the engine's audio settings or an audio editor, on a bar line.
• Ask for the length in the prompt ("a 2-minute track"). duration and duration_seconds are rejected, and tracks run up to a few minutes.
• There is no layered or adaptive music: for a calm and an intense version of a level, generate two tracks with the same tempo and key.
• There is no audio input or lyric field, so you can't extend or remix an existing track.

How do I get WAV files for my game engine?

The generated track is typically audio/mpeg on media.sume.com. A Timeline audio split takes that Sume-hosted file, cuts it into 1 to 20 ranges, and returns each range as its own file:

How much does AI game music cost?

Every generation is a flat $0.125 per audio generation, whatever the prompt length, and every take counts. The track counts and takes below are example assumptions, with one split per kept track.

On rights: the Sume Terms of Service say Sume does not claim ownership of generated outputs and that paid plans include commercial use as described on the pricing page. That is what the terms say, not legal advice. Art for the same game is covered in AI sprite generator.