audio_generate
Generate spoken audio from written text for voiceovers, narration, ad reads, or character lines. Create one or multiple clips per request using your selected voice.
Instructions
For a user's voiceover request, load the generating-audio skill BEFORE calling this — it picks the right model and voice and prepares the script for reading aloud (this tool does none of that, and calling it raw gives a flat, mispronounced read). Turn written text into spoken audio: voiceovers, narration, ad reads, character lines, or any script read aloud. Pass requests: ONE object per clip (wrap even a single clip — { requests: [ { text } ] }); add more objects (up to 10) to generate DIFFERENT lines in one call — a single approval covers the batch. Each result carries the spoken audio plus a local file path, or a structured error with a hint. This generates speech and nothing else: no sound effects, music, or ambience, no re-voicing an existing recording, and no dubbing a video. If the user asks for one of those, say so plainly rather than substituting a different tool. Every request needs a voice — the voice_id of a row from list_voices. There is no default voice. Models differ in expressiveness, language coverage, speed, price, and per-request character limit — call list_audio_models to compare them. Set dry_run=true to preview the exact requests without generating (no credits spent).Each entry in results is one of three things: finished media; a {status:"pending", ...} job handle to rejoin with job_status; or a failure carrying ok: false and an error. A failed entry is terminal — report its error and never poll or re-submit it. Read every entry rather than the top-level counters alone.
Input Schema
| Name | Required | Description | Default |
|---|---|---|---|
| dry_run | No | If true, return the requests that would be sent (keys masked), make no API call. | |
| requests | Yes | One object per audio clip (wrap even a single clip); add more objects to generate different lines in one call (up to 10). |