A Go TUI application for real-time meeting transcription using local speech-to-text. Works with any meeting application by capturing system audio.
- Captures system audio using PulseAudio/PipeWire monitor sources
- Works with any meeting application (Zoom, Meet, Teams, Discord, etc.)
- Fully local transcription using whisper.cpp - no API calls, no data sent anywhere
- Real-time transcription display with audio level visualization
- Save transcripts to text files
- Beautiful TUI interface built with Bubble Tea
# Fedora
sudo dnf install portaudio-devel pipewire-pulseaudio gcc-c++ make
# Ubuntu/Debian
sudo apt install portaudio19-dev pulseaudio-utils build-essential
# Arch Linux
sudo pacman -S portaudio pulseaudio-utils base-develYou need to install whisper.cpp for the transcription engine:
# Clone and build whisper.cpp
git clone https://github.com/ggerganov/whisper.cpp
cd whisper.cpp
make
# Install the binary
sudo cp build/bin/whisper-cli /usr/local/bin/whisperDownload a Whisper model from Hugging Face:
# Create models directory
mkdir -p ~/.rekord/models
# Download base English model (recommended default)
wget -P ~/.rekord/models https://huggingface.co/ggerganov/whisper.cpp/resolve/main/ggml-base.en.binFor more context regarding model selection, see the whisper.cpp model files instructions.
# Install via Go
go install github.com/exler/rekord/cmd/rekord@latestDownload a prebuilt binary from GitHub Releases:
https://github.com/exler/rekord/releases
Or build from source:
# Clone the repository
git clone https://github.com/exler/rekord
cd rekord
# Build
go build -o rekord ./cmd/rekord# Run with default settings (uses default audio monitor and base model)
rekord
# Specify a different model
rekord -model ~/.rekord/models/ggml-small.en.bin
# Specify a specific audio device
# Use `pactl list sources short` to find the correct monitor source for your system
rekord -device alsa_output.pci-0000_00_1f.3.analog-stereo.monitorgo run cmd/rekord/main.goEnvironment variables:
WHISPER_PATH: Path to the whisper.cpp executable (default: searches in PATH and common locations)
Command-line flags:
-model: Path to the Whisper model file-device: Audio device name (usepactl list sources shortto list)-output: Output directory for saved transcripts
Rekord is under the terms of the MIT License, following all clarifications stated in the license file.
