Hosted music and sound effects

Create music or a sound effect from a prompt with the app's own generators, paid with StillMade credits.

Use runtime: "capability" with operation: "audio.music" or operation: "audio.sfx" to create music or a sound effect with the app's own generators: ACE-Step music and StillMade sound effects, the same ones the Editor uses. The person running the Block pays with StillMade credits at the app's prices (music by the minute, one flat price per sound effect). Block code never sees keys, endpoints or payment.

The manifest declares one text, script or brief input, one audio output, and permissions.capabilities: ["audio.music"] (or ["audio.sfx"]):

{"schemaVersion":1,"operation":"audio.music","prompt":{"$input":"brief"},"output":"music","settings":{"durationSec":30}}

settings.durationSec is optional and only suggests a length. Music runs 5 to 240 whole seconds; sound effects run 0.5 to 22 seconds in tenths. The person running the Block can change the length in the dialog, and the quote shows the price for that length before anything is generated.

The output is a measured MP3 or WAV reference:

{"kind":"audio","assetId":"audio:run:0","versionId":"<sha256>","url":"/api/media/owner/capaud3-....mp3","mimeType":"audio/mpeg","duration":30.02,"bytes":481254,"sampleRate":44100,"channels":2}

Fixture expectations use format any, mp3 or wav with duration and byte bounds (up to 1,800 seconds and 128 MiB):

{"music":{"kind":"audio","format":"any","minDuration":1,"maxDuration":241,"minBytes":45,"maxBytes":134217728}}

The offline SDK includes packages/block-sdk/audio-generation-example.js, exporting audioMusic and audioSfx. Review, listening and acceptance follow the same flow as hosted speech: Review cost never generates, each fixture and the sample is quoted and charged once, and a provider refusal before generation is refunded.