Omascribe

A keyboard-driven TUI for recording, transcribing, and summarising meetings on Linux.

Built specifically for Omarchy Quattro — integrates natively with the Quickshell bar, desktop notifications, and SUPER+M keybinding.

Features

• Record — mic + system audio (PipeWire/PulseAudio)
• Transcribe — local Whisper (CPU, privacy-first)
• Summarise — cloud LLM (OpenAI, Anthropic, OpenRouter) or local Ollama
• Write notes — add your own context during recording for better AI summaries
• Keyboard-driven — Lazygit-inspired layout, no mouse required
• Omarchy-native — bar status, notifications, app menu, and SUPER+M out of the box

Quick Start

git clone https://github.com/jamespember/omascribe.git
cd omascribe
./setup.sh

On Omarchy Quattro this adds:

• SUPER + M — launch or focus
• Apps menu entry
• Omascribe control panel — bar widget with live recording status, quick actions, and recent meetings
• Desktop notifications for recording events

The control panel plugin lives at integrations/omarchy/omascribe-control/ and is installed by ./setup.sh into ~/.config/omarchy/plugins/.

Usage

omascribe

Key	Action
r	Start recording
s	Stop and process
x	Cancel recording
o	Open in editor
e	Edit title
t	View transcript
T	Manage tags
d	Delete
,	Settings
A	Audio test
q	Quit
j/k or ↑↓	Navigate
/	Search
1 / 2	Focus Meetings / Note pane

During recording, write notes in the text area — they're fed to the AI as extra context.

AI Setup

Cloud (fast, recommended):

./setup_cloud.sh
# or press `,` in the app and pick a provider

Local (free, private, slower):

curl -fsSL https://ollama.com/install.sh | sh
ollama pull llama3.2:3b

Or skip AI entirely — set ai_provider: none in settings for transcription-only.

Output

Notes are saved as markdown in notes/:

---
title: "Sprint Planning"
date: 2026-08-18
duration_seconds: 1860
word_count: 4230
tags: [meeting, auto-generated]
---

# Sprint Planning

**Date:** August 18, 2026 at 2:30 PM  
**Duration:** 31 minutes  
**Words:** 4,230

## AI Summary
...

### Action Items
- Sarah to send preview link by tomorrow morning

Full transcripts with timestamps are saved separately in transcripts/.

Audio

Recording modes: combined (mic + system, default), mic, system

Device selection: Pick specific mic and output devices in Settings → Audio, or use system default.

Audio Test (A from main view) records a 5-second clip and diagnoses whether your meeting app's audio is actually hitting the captured sink. Catches common traps like Zoom routing to a different output.

Configuration

Settings are stored in ~/.config/omascribe/config.yaml:

ai_provider: anthropic        # none | openai | anthropic | openrouter | local
ai_model: haiku               # haiku/sonnet | mini/standard | cheap/balanced/premium
whisper_model: base           # tiny | base | small | medium | large
whisper_device: cpu           # cpu | cuda | auto
recording_mode: combined      # mic | system | combined
editor: nvim
notes_dir: notes
transcripts_dir: transcripts

Development

python -m venv venv
source venv/bin/activate
pip install -e ".[all,dev]"
pytest          # 127 tests
ruff check omascribe/ tests/

License

MIT