oatda-generate-speechlisted

# OATDA Speech Generation Generate spoken audio from text through OATDA's unified audio API. ## When to Use Use this skill when the user wants to: - Convert text to speech or audio - Create voiceovers, announcements, narration, or accessibility audio - Use TTS models from OpenAI, xAI, Google and others through OATDA - Use the OATDA `generate_speech` capability ## HTTP vs MCP (read this first) | Path | Response | Use when | |------|----------|----------| | **MCP** `generate_speech` | Text with **`AUDIO_URL:`** (HMAC `?exp=&sig=`, ~1h TTL) + metadata in `structuredContent` (`download_url`, `format`, `duration_seconds`, `costs`) | Already connected to OATDA MCP (Cursor, etc.) | | **HTTP** `POST /api/v1/llm/speech` | Raw audio bytes (`Content-Disposition: attachment`) | Shell/scripts: `curl --output speech.mp3` | **Upstream providers (OpenAI TTS, xAI Grok TTS) do not return a hosted download URL** for speech. xAI returns raw bytes (`curl ... --output hello.mp3` in [their docs](https://docs.x.ai/docs/guides/voice)). OATDA MCP mirrors TTS into a short-lived signed download so editors can fetch without a browser session. **MCP agents:** After `generate_speech`, extract `AUDIO_URL:` from the tool text (or `structuredContent.download_url`) and download with curl - **no login required**. Do **not** expect MCP `AudioContent` / base64 audio blocks (removed). URL expires in ~1 hour. **Shell/scripts:** Prefer HTTP with `--output` (see step 4), or MCP signed URL + curl. ## Prerequi