MeetMind
Turns recorded conversations into structured, searchable meeting information, with the transcription and analysis pipeline running locally rather than outsourced to a hosted API.
Overview
From a raw recording to a structured workspace
Users upload recorded audio; MeetMind processes it into a timestamped transcript, an extractive summary, decisions, topics, action items and overall sentiment.
The current workflow is recorded-audio processing rather than live meeting capture: users upload existing recordings in MP3, WAV or M4A. A major part of the engineering story is that the intelligence pipeline runs locally rather than outsourcing the entire workflow to a hosted meeting-summary API.
Processing Pipeline
Audio in, structured meeting out
Audio Upload
MP3 · WAV · M4A
Local Whisper Transcription
faster-whisper
Speaker Separation
Feature clustering
Meeting Analysis
Local NLP
Structured Workspace
Summary · Actions · Transcript
Transcription uses faster-whisper running locally, producing timestamps with generic speaker labels. Speaker separation is approximate: audio features and clustering group speech into generic speakers such as Speaker 1 and Speaker 2, not identity recognition.
Meeting Intelligence
What comes out of the analysis
Produced by local heuristics, not an LLM: an automatic title, a structured extractive summary, and more.
Summary
Automatic title when appropriate, extractive summary and overall discussion description.
Discussion
Critical points and speaker-associated discussion points.
Decisions & actions
Important decisions and extracted action items.
Topics
Topics and keyphrases surfaced from the transcript.
Sentiment
Overall positive / neutral / negative reading with a short explanation.
Workspace
Action items become a living checklist
Three primary views per meeting: AI Summary, Action Items, Transcript.
Action items persist as an interactive checklist users can mark complete, giving a processed meeting a useful follow-up layer beyond a static transcript. Each meeting's detail view also tracks title, filename, upload date, duration, processing status, topics and sentiment.
History & Export
Searchable history, exportable reports
Authenticated users have persistent meeting history, searchable across titles, filenames, transcripts, summaries and topics. Each meeting can also generate a PDF report containing its metadata, topics, sentiment, summary, action items and full timestamped transcript.
Stack
Technology
Frontend
Backend
Database
Audio / ML
Reports & Auth