sonote

What problem does it solve?

This skill helps you turn real-time Korean speech from meetings or YouTube live streams into accurate, agent-ready transcripts without manual transcription effort.

Core Features & Use Cases

• Real-time Korean STT CLI: Transcribe microphone input or YouTube live audio with low-latency streaming and structured output for downstream automation.
• Speaker diarization (optional): Separate speakers using HF_TOKEN-powered diarization to make meeting transcripts easier to understand and index.
• LLM post-processing: Automatically refine transcription and generate summaries/to-dos using supported LLM backends (Codex/Gemini/Ollama), producing more useful meeting intelligence.

Use case: You run an agent-driven workflow for a professional meeting—capture speech, get speaker-separated transcripts in JSON, then automatically produce a structured summary and action items for follow-up.

Quick Start

Run the skill in meeting mode with diarization and JSON output by starting a local server for real-time subtitles at 8000 using: sonote meeting --json.