What is voicelogger-cli?

voicelogger-cli is a command line tool that records audio from your microphone and turns it into a text note. It transcribes speech locally on your own machine using whisper.cpp, a fast speech recognition engine, so the raw recording never leaves your computer. Optionally, it can then send the transcript to an AI model to clean it up: removing filler words, fixing terminology, and organizing the content into a tidier note. The original raw transcript is always kept as a separate file, and nothing is uploaded unless you choose to run that optional cleanup step. Setting it up involves installing Node.js version 20 or newer, cloning the repository, and running a couple of install commands so the voicelogger command becomes available in your terminal. You also need ffmpeg to capture microphone audio and whisper-cpp to run the local transcription, both installable through Homebrew on macOS. Linux and Windows microphone support is described as experimental. A one-time download fetches the speech recognition model, which takes about 30 seconds. If you want the automatic cleanup feature, a setup wizard walks you through choosing an AI provider such as Anthropic, OpenRouter, or Ollama, the last of which runs entirely on your own machine with no API key or internet connection needed. You can skip this step entirely and just use raw transcripts. Once set up, the main workflow is simple: run a record command, speak, then press Enter to stop. The tool prints the cleaned note directly in your terminal and saves both the raw and cleaned versions to files. Other commands let you list past recordings, show a specific one, manually clean a raw transcript later, tag recordings with a project name, and even automatically copy a finished note into a specific project folder on your computer. Configuration such as where logs are saved, which AI provider to use, and which model handles cleanup can all be changed through simple commands or environment variables.