What problem does it solve? Producing natural-sounding Mandarin voiceovers with accurate subtitle timing is difficult when TTS APIs have confusing authentication schemes and no built-in timing metadata. This Skill standardizes calls to the Volcengine Doubao Speech 2.0 API, handles the correct new-console API key authentication, and returns character-level timestamps needed for subtitle generation. ## Core Features & Use Cases - Mandarin and Multilingual Narration: Generate voiceovers via the async submit/query endpoints of Doubao Speech 2.0, with configurable voice, speech rate, and sample rate. - Word-Level Timestamp Metadata: Retrieve sentences[].words[] timing data from the query response to build accurate subtitles grouped by Chinese semantic phrases. - Safe Approval Workflow: Generate a short sample for user approval of voice, accent, and speed before committing to a full paid narration. - Use Case: You are producing a Chinese explainer video. Use this Skill to generate the narration MP3, keep the query JSON as the timing source of truth, and build captions aligned to the actual spoken words. ## Quick Start Set DOUBAO_SPEECH_API_KEY in your .env file, then ask the agent to generate a short Mandarin narration sample with Doubao TTS and save it to your project's audio assets folder.
