Humla
Privacy-first macOS meeting transcription and AI summarization tool
Humla is a meeting-notes app for macOS that records, transcribes, and summarizes meetings using local or cloud AI. It captures microphone and system audio as separate streams to ensure clean transcripts and supports local Whisper models, speaker identification, and customizable summary presets.
- Built with
- Whisper-1Large v3Nova-3 / Nova-2NB Whisper LargeStrong
The developer provides detailed documentation on supported AI engines in the repository README, specifically naming OpenAI's Whisper-1 and reasoning models.
README explicitly lists "OpenAI — whisper-1" for transcription and "gpt-5.x reasoning models" for summaries.; source explicitly credits Speechmatics (Mentioned in the 'About' section and title: 'OpenAI / Speechmatics / on-device Whisper'.); Deepgram Nova-3 / Nova-2 (README lists 'Deepgram — Nova-3 / Nova-2' under transcription options.); NB Whisper Large (README mentions support for 'NB Whisper Large from Nasjonalbiblioteket' for Norwegian transcription.); pyannote community-1 (README credits 'pyannote community-1' for the offline speaker identification pipeline.)
- Posted by
- We haven't verified this person made this build — attribution is our best guess from public sources.
- Discovered
- Discovered onX
- Source date
- Published on X Jul 4, 2026
- Listed
- Added to Dropday 2h ago
- Evidence
- Strong
The project is a comprehensive open-source repository with over 300 commits, 72 releases, and a detailed build process for a native macOS application.