Speech AI is a complete speech processing platform with three APIs: • Pronunciation Assessment — phoneme-level scoring, 17MB model, beats human experts • Speech-to-Text — word timestamps + confidence scores, same 17MB model • Text-to-Speech — 12 English voices via Kokoro-82M (#1 TTS Arena, Apache 2.0) All three ship as an MCP server with 8 tools, so AI agents can assess, transcribe, and speak in one integration. REST API and Azure Marketplace also available.