Meetily: 100% Local AI Meeting Recorder, Real-Time Transcription & Diarization

Boardroom Privacy: Local AI Meeting Recording and Summarization

Cloud-based meeting recorders such as Otter.ai, Fireflies.ai, and Zoom AI Companion have become standard corporate tools, but they introduce profound security risks. Feeding executive strategy discussions, patient consultations, and intellectual property reviews into third-party cloud infrastructure is an unacceptable exposure for security-conscious organizations. Meetily delivers a robust, 100% local alternative.

Engineered with a high-performance Rust audio processing core, Meetily pairs on-device speech transcription engines (Whisper and NVIDIA Parakeet) with local LLM summarization via Ollama. It transcribes discussions in real time, separates speaker identities, and synthesizes structured meeting minutes without sending a single audio packet over the internet.

What It Is

Meetily is a self-hostable, open-source AI meeting assistant available as both a native cross-platform desktop client (macOS, Windows, Linux) and a centralized team server deployable via Docker.

It captures dual audio channels simultaneously (microphone input and system speaker loopback), generates synchronized timestamps with speaker diarization, and automatically produces structured Markdown meeting minutes featuring executive summaries, key decisions, and assigned action items.

Why You Have Not Heard Of It

Developed by Zackriya Solutions, Meetily was created to satisfy strict compliance frameworks (GDPR, HIPAA, SOC 2) in legal, medical, and defense environments where third-party AI bots are strictly barred from video calls. Built without venture-backed PR campaigns, the project gained swift adoption across security engineering teams and privacy-conscious homelabs.

How It Works

Meetily’s pipeline is cleanly divided into three distinct offline stages:

Rust Audio Capture & VAD Engine: High-fidelity dual-stream audio capture combined with Voice Activity Detection (VAD) to filter background noise and isolate speech segments with minimal CPU load.
Accelerated Neural Transcription: Leverages quantized Whisper.cpp and Parakeet models utilizing Apple Silicon Metal, NVIDIA CUDA, and AMD ROCm hardware acceleration for low-latency transcription.
Local Ollama Integration: Upon meeting completion, full transcripts are dispatched locally to models like Llama 3, Qwen, or Mistral running on an internal Ollama instance to generate concise action points.

Running It

Meetily can be run directly as a native application or deployed as a self-hosted team backend using Docker Compose:

services:
  meetily-server:
    image: ghcr.io/zackriya-solutions/meetily:latest
    container_name: meetily
    restart: unless-stopped
    ports:
      - "3000:3000"
    volumes:
      - ./recordings:/app/recordings
      - ./transcripts:/app/transcripts
    environment:
      - OLLAMA_HOST=http://host.docker.internal:11434
      - DEFAULT_MODEL=llama3:8b
    extra_hosts:
      - "host.docker.internal:host-gateway"

Launch the container:

docker compose up -d

What It Replaces

SaaS Meeting Recorders (Otter/Fireflies): Eliminates per-seat monthly subscription fees and removes the risk of vendor data harvesting.
Intrusive Third-Party Bots: Records system audio natively at the driver level, avoiding awkward external bot avatars entering client meetings.
Manual Note-Taking: Produces consistent, timestamped documentation in seconds post-call.

Limitations

Real-time transcription and prompt summarization require adequate local compute (recommended: 16 GB of unified RAM or a dedicated 8 GB GPU). On older, CPU-only hardware, batch post-meeting transcription is recommended over live streaming modes.

Who It Is For

Legal, healthcare, and financial professionals obligated by law to protect confidential client communications.
Engineering leaders discussing sensitive architectural roadmaps and zero-day vulnerability mitigations.
Remote teams seeking a unified, private meeting archive under their complete organizational control.

Conclusion

Meetily proves that local AI tools can match—and often exceed—the capabilities of proprietary SaaS products while restoring absolute data sovereignty. For teams that refuse to compromise on privacy, it is the premier open-source meeting assistant.