Script Submit contract guide

Final Audio Script format

Submit one JSON object using the final_audio_script_2 schema. It needs a title, language, performance, ordered units, and a cast row for every speaker ID used in speech. Earlier final_audio_script_1 files still work with Normal delivery.

• Each spoken line is {"kind":"speech","speaker_id":"...","text":"...","delivery":"urgent"}. Choose a delivery ID below or omit it for Normal. Its text is recorded exactly. scene and section headings and stanza_break are silent.
• Set delivery_catalog_version to delivery_catalog_2. Delivery is a best-effort performance direction; the cast profile remains the same person on every line.
• Each cast row needs speaker_id, name, story_function (narrator, other, or antagonist), and one inline profile. The profile name must match the cast name.
• The profile uses play_cast_profile_1 with gender, exact Dali voice ID, age ID, personality ID, and language. Use en for an English script profile; use und for other script languages. Saved-person references are not admitted here yet.

Defaults and required fields

For final_audio_script_2, omit a speech line's delivery to use normal. You may also omit content_form when no form applies. An explicit null delivery is not a valid choice.

There are no automatic defaults for the title, language, performance, spoken text, speaker IDs, or cast. Every spoken speaker needs a matching cast row and a complete inline voice profile. The schema_version, catalog_version, and delivery_catalog_version are required. Validation identifies missing or invalid fields before audio creation.

Example cast row from the current server contract:

Available system voices and cast dimensions

Download contract schema

Back to Script Submit