Music & Audio Generation

Overview

ToRun can generate music and other audio assets the same way it generates images and video: you describe what you want, ToRun routes the request to a capable provider model, and you receive a finished asset. Generation is billed per asset produced, with the full cost shown before and after the run.

Music and audio generation are exposed as platform capabilities (music-gen, text-to-speech) and run through the same capability-first routing as every other mode: your selected mode resolves to the required capability set, ToRun filters to models that support it, applies your subscription tier, and ranks the remaining models by price.

What you can produce

• Music — text-to-music clips from a prompt (genre, mood, instrumentation, tempo).
• Speech / narration — see the separate Voice (TTS/STT) doc for spoken-word audio across 29 languages.
• Workflow-driven audio — any audio capability can be used as a step inside a no-code workflow (for example, generate a jingle, then save it to your CDN library).

How to generate music

1. Open the audio/music mode from the chat or media surface.
2. Write a prompt describing the piece. Be specific about genre, mood, length, and instrumentation.
3. (Optional) Pick a model if you want a specific provider; otherwise routing selects the best-priced capable model.
4. Run. The cost estimate appears before you confirm, and a BillingRecord is written when the asset is produced.

Cost and metering

• Each generated asset creates exactly one BillingRecord with a full pricing snapshot.
• Audio jobs are tracked as job entities (for example AudioGenJob) so you can see status, inputs, and the resulting asset.
• If a job runs local compute (such as audio composition/merging), that compute is metered separately as compute-seconds and billed at the published compute rate.

Storage and reuse

Generated audio is uploaded to your media library on Bunny CDN so the URL stays stable. Provider-side URLs typically expire within hours; ToRun re-hosts the asset so your history and shares keep working. In workflows, use the artifact-save step to persist a transient provider URL to a permanent CDN URL.

Tips

• Keep prompts concrete — vague prompts produce generic results and still cost a full asset.
• For repeatable production (e.g. a jingle per blog post), wrap the audio step in a workflow so you get consistent settings and automatic CDN archiving.
• Check your remaining allowance or wallet balance before batch runs; each asset is a separate billable call.