Ownscribe is a new open-source tool that handles meeting transcription and summarization entirely on your machine. No cloud APIs, no data uploaded anywhere, no bots joining your calls. For anyone wary of sending meeting recordings to Otter.ai or Zoom's cloud transcription, this is a privacy-first alternative worth a look. Apfel also offers a similar local-first approach for on-device AI tasks.
The tool uses WhisperX for speech-to-text with word-level timestamps and supports speaker diarization through pyannote.audio. Summarization runs through local LLMs, defaulting to Microsoft's Phi-4-mini which downloads automatically on first run, or you can point it at Lemonade, any OpenAI-compatible local server. On macOS 14.2+, it captures system audio directly via Core Audio Taps, so it works with Zoom, Teams, Meet, or any app without virtual audio drivers.
It also supports customizable summarization templates for different use cases like lectures or quick briefs, plus silence auto-stop so you don't end up with hours of dead air if you forget to hit stop.
The catch: it's macOS-only for system audio capture right now. Other platforms can use it with an external audio source via sounddevice, but the tight Core Audio integration is a Mac feature. You'll need Python 3.12+, uv, and ffmpeg installed. Developer paberr built this as a CLI-first tool, so there's no GUI, but the tradeoff is a lightweight setup that respects your data sovereignty by default.