HumlaGlossary
The words behind private meeting notes.
Plain-English definitions of the terms that come up around local, no-bot, on-device meeting notes.
Free · Open source · macOS 13+ (Apple Silicon)
All terms
Meeting-notes glossary.
- Speaker diarization — Speaker diarization is the process of splitting an audio recording into segments by who is speaking, so a transcript can show “who said what.”
- Local-first software — Local-first software keeps your data on your own device as the primary copy, working without a server and syncing only if and when you choose.
- On-device transcription — On-device transcription converts speech to text entirely on your own computer, with no audio sent to a cloud service.
- Meeting bot — A meeting bot is an automated participant that joins a video call to record and transcribe it, appearing in the attendee list like another guest.
- System audio capture — System audio capture records the sound your computer plays — any app's output — rather than only what a microphone hears.
- Whisper (speech recognition) — Whisper is OpenAI's open-source speech-recognition model, widely used to transcribe audio and runnable entirely on your own hardware.
- Two-party consent — Two-party consent (or all-party consent) is a legal rule requiring everyone in a conversation to agree before it can be recorded.
- GDPR & meeting recording — Under the EU's GDPR, a meeting recording is personal data, so it needs a lawful basis (such as consent), data-minimisation, and care about where it's stored.
- Bring your own key (BYOK) — Bring-your-own-key means an app uses your own API key for a service, so you pay the provider directly and control the account instead of the app reselling access.
- Offline transcription — Offline transcription turns speech into text with no internet connection, using a model that runs locally on your device.
- AI notetaker vs AI notepad — An AI notetaker records and transcribes a meeting for you; an AI notepad is one you write in yourself while the AI augments your notes with the transcript.
- Voice activity detection (VAD) — Voice activity detection is the technique of detecting when someone is actually speaking in an audio stream, as opposed to silence or background noise.
- RAG (retrieval-augmented generation) — RAG (retrieval-augmented generation) is a technique where an AI first retrieves relevant documents, then writes its answer grounded in what it found — ideally citing the sources it used.
- Semantic search — Semantic search finds text by meaning rather than exact keywords, using embeddings to match passages that are about the same thing even when they share no words.
- Embeddings — Embeddings are numeric fingerprints of meaning — lists of numbers that place similar pieces of text near each other, so software can measure how related two passages are.
- Agentic retrieval — Agentic retrieval is when an AI doesn't just run one search but loops — searching, reading a result, then trying a new query — until it has enough to answer.