Crisp Video Transcript Summaries

Замовник: AI | Опубліковано: 03.01.2026

Project Title: AI Meeting Bot – Fireflies-Like Meeting Assistant (Local, Free & Private) --- Project Description: I am looking for a developer (or team) to build a fully functional AI meeting bot that can automatically join online meetings (Google Meet or Zoom) through a meeting link, record the audio, and generate timestamped transcripts with speaker names — similar to Fireflies.ai, but running entirely on local and free tools (no paid APIs). The goal is to develop a self-hosted, privacy-friendly meeting assistant that operates fully offline. --- Core Features 1. Auto-Join Meetings The bot automatically joins Google Meet or Zoom using Selenium or a headless browser. Option to use an existing Chrome profile (logged-in bot account). 2. Audio Recording (Local) Capture complete meeting audio using FFmpeg or similar open-source tools. Save all recordings locally — no cloud upload. 3. Automatic Transcription Convert recorded audio to text using Whisper or WhisperX (OpenAI’s open-source models). Store transcripts in structured JSON format. 4. Speaker Detection (Diarization) Automatically detect speaker changes using pyannote.audio or WhisperX diarization. No prior training or voice samples required. 5. Speaker Name Recognition (OCR) Automatically identify on-screen speaker names during meetings using OCR (Tesseract). Map diarized audio segments to the visible meeting participant names. 6. Transcript Output Merge transcript text, timestamps, and detected speaker names into a clean, formatted output. Export final results as both PDF and JSON files (with timestamps, speaker names, and spoken text). 7. Local-Only Operation Everything runs on the local machine. No cloud services, no paid APIs, and complete data privacy. --- Tech Stack (Open Source Only) Python 3.9+ Selenium – for browser automation (meeting join) FFmpeg – for system audio capture Tesseract OCR – for on-screen name recognition Whisper / WhisperX – for transcription pyannote.audio – for speaker diarization ReportLab – for PDF generation --- Expected Deliverables 1. Fully working project with complete setup instructions 2. Modular Python scripts for: Meeting joining Audio recording Transcription Speaker mapping PDF/JSON generation 3. Configurable settings (meeting details, file paths, output preferences) 4. Well-documented README and in-code comments 5. Tested on at least one platform (Google Meet or Zoom browser version) --- Bonus (Optional) Simple web dashboard (Streamlit or Flask) to review transcripts and rename speakers Start/stop meeting recording through a local UI Multi-platform support (e.g., Microsoft Teams, Webex) --- Skills Required Python automation (Selenium, subprocess, threading) Speech-to-text (Whisper / WhisperX) Audio processing (FFmpeg) OCR (Pillow + pytesseract) PDF generation (ReportLab) Basic AI pipeline design and integration --- Why This Project Is Interesting This project aims to create a Fireflies.ai alternative that is completely free, offline, and privacy-focused. It combines AI, automation, and speech technologies into a single open-source meeting assistant that anyone can self-host and use without cloud dependency.