Local audio notes: record/import → local Whisper transcript → 3-layer LLM summary → cross-meeting knowledge graph. Self-hosted, no cloud. MCP server + shared graph viewer.
View on GitHub ↗Home → a meeting's 3-layer summary → the cross-meeting knowledge graph. (Placeholder demo data. ▶ full-quality MP4.)
Standalone local audio notes extracted from type-by-voice.
This tool records or imports audio, transcribes it locally with faster-whisper, and writes:
- transcript.md - notes.md - summary.md - audio.wav - meta.json
It does not do global push-to-talk dictation, tray UI, or paste text into the focused app. Those remain the responsibility of type-by-voice.
One-command install. It creates a Python venv, downloads the local Whisper transcription model, installs Ollama + a small local summary model, and writes a config — so transcription and summaries run fully locally, no cloud, out of the box.
Lighter machines can use a smaller model: VOICENOTESWHISPERMODEL=small ./install.sh (and VOICENOTESLLM=llama3.2:3b).
Layer 2/3 summaries need an LLM — run Ollama locally, or point openaiurl at an OpenAI-compatible server (see below).
The default page is . It can start/stop a recording, upload audio files, import macOS Voice Memos, and download the generated files. The generated notes are plain Markdown in a local folder — keep them in your own repo, share them, or read them from Claude via the MCP server.
The three layers run strictly in sequence — Layer 1 → Layer 2 → Layer 3:
- Layer 1: raw chunked transcript, written live as each audio segment finishes. Nothing else runs until the whole transcript is done. - Layer 2: once Layer 1 completes, the ENTIRE transcript is handed to the local Ollama model in one go, producing a detailed full-text summary grouped by topi
From the project README.
Add the radar badge to your README — it shows your project was picked up by MCP Radar and links to this page:
[](https://mcp.liqiwa.com/s/nikoloside--voice-notes.html)
Stop your AI from making things up — it proposes, deterministic tools decide, every claim checked against ground truth with evidence. Grounded facts and context survive resets. Reverse engineering is
zhuyansen/new-term-radarDaily radar for new AI terms: names of new models, products and repos, caught on X, Hugging Face, GitHub and sitemaps before keyword tools list them.
yzg1226/mcd-party-officer🍟 麦当劳门店派对成团调度 Skill — 把派对从「策划」推进到「办得成」：成团概率雷达 / 包场vs拼团 / 择日黄历 / 儿童营养指数 / 招募卡与邀请函
fellowgeek/mcp-memoryAn OKF-backed Model Context Protocol (MCP) server delivering persistent long-term memory and SQLite FTS5 search for AI agents.
hirotomasato/yowesGenerate realistic teacher documents (ID cards, licenses, letters) for 13 countries via MCP.
hujjun/mcd-saver-strategist麦当劳省钱点餐助手（麦门精算师）｜麦当劳优惠券组合优化 Skill：拉取麦当劳菜单与全部可用优惠券，算出最省的点餐组合，巨无霸套餐 ¥47 → ¥21.9（省 53%）。McDonalds MCP（麦当劳 MCP）coupon optimizer · McDonalds China MCP Server · Agent Skill for WorkBuddy · 麦当劳程序员创意开发大赛参赛作品
The top new MCP servers of the week, every Monday. No spam, unsubscribe anytime.