Record. Transcribe. Summarize.

A Linux desktop app that records meetings, transcribes them, and generates structured notes — all in a few clicks.
Supports cloud and local AI processing, with 100+ providers via LiteLLM.
Arch Linux · PipeWire + Wayland · GTK3

# Install from AUR
yay -S meeting-recorder    # or: paru -S meeting-recorder

# Launch
meeting-recorder

Open Settings, pick your transcription and summarization providers, add API keys, and you're ready to record.

For fully offline use, select Whisper (transcription) + Ollama via LiteLLM (summarization) — no API keys needed.

• Two modes — Headphones (mic + system audio) or Speaker (mic only, avoids echo)
• Pause / Resume mid-recording
• Separate audio tracks — optionally save mic and system audio as independent files for better diarization
• Transcribe existing files — drop in any audio/video file for transcription without recording

Mix and match freely — e.g. Whisper + Ollama runs fully offline with no API keys.

• Per-monitor Wayland-native capture via gpu-screen-recorder
• Merge with audio — optionally combine screen recording + audio into a single video
• Night light inhibition — automatically pauses KDE night light during recording for accurate colors

• Browse all recorded meetings in a searchable list
• AI-generated titles — auto-name meetings from their notes, or generate titles manually
• Inline rename — double-click to rename any meeting
• Open folders, delete recordings, bulk-select

• Call detection — monitors PipeWire for active calls and notifies you to start recording
• Auto-title — AI generates a short title from your meeting notes after processing
• Artifact cleanup — choose which output files to keep (audio, transcripts, screen recordings, notes)
• GPU memory management — automatically unloads models between pipeline steps to prevent OOM on limited VRAM
• System tray — StatusNotifierItem on KDE (pystray fallback), configurable click actions, background job status
• Autostart — optionally launch on login

1. Click Record (Headphones) or Record (Speaker) to start
2. Pause / Resume as needed — a timer shows elapsed time
3. Click Stop — transcription and summarization run automatically
4. Browse results in the app or open the output folder

Each session saves to a dated hierarchy:

~/meetings/
└── 2026/
    └── March/
        └── 04/
            └── 14-30_Standup/
                ├── recording.mp3              # Combined audio
                ├── recording_mic.mp3          # Mic track (if separate tracks)
                ├── recording_system.mp3       # System track (if separate tracks)
                ├── screen-eDP-1.mp4           # Screen recording (if enabled)
                ├── screen-eDP-1_merged.mp4    # Screen + audio merged (if enabled)
                ├── transcript.md
                └── notes.md

LiteLLM routes to providers via the model string prefix:

gemini/gemini-2.5-flash              # Google Gemini
ollama/phi4-mini                     # Local Ollama
openai/gpt-4o                        # OpenAI
anthropic/claude-sonnet-4-latest     # Anthropic
openrouter/anthropic/claude-sonnet-4 # OpenRouter
groq/whisper-large-v3                # Groq (transcription)

Select from curated lists in Settings, or type any provider/model string.

yay -S meeting-recorder
# or
paru -S meeting-recorder

git clone https://github.com/AJV009/meeting-recorder.git
cd meeting-recorder
./install.sh

Installs system deps via pacman, sets up a Python venv via uv, and auto-installs gpu-screen-recorder from AUR if yay/paru is available.

./uninstall.sh

Your recordings (~/meetings/) and config (~/.config/meeting-recorder/) are preserved.

• Arch Linux with PipeWire (tested on KDE Plasma 6 / Wayland)
• ffmpeg, pipewire, pipewire-pulse, wireplumber, Python 3, GTK3
• Optional: gpu-screen-recorder (AUR) for screen recording

Open Settings from the gear icon or system tray menu. Settings are organized into tabs:

pactl load-module module-echo-cancel aec_method=webrtc noise_suppression=true

pulse.cmd = [
  { cmd = "load-module" args = "module-echo-cancel aec_method=webrtc noise_suppression=true" flags = [] }
]

systemctl --user restart pipewire pipewire-pulse

~/.local/share/meeting-recorder/meeting-recorder.log

ffmpeg -i corrupted.mp4 -c copy recovered.mp4

git clone https://github.com/AJV009/meeting-recorder.git
cd meeting-recorder
python3 -m venv .venv --system-site-packages
.venv/bin/pip install -e ".[dev]"
PYTHONPATH=src python3 -m meeting_recorder

Run tests:

python -m pytest tests/ -v

MIT