Quickstart Guide

Get dictating in under 2 minutes.

1. Install

curl -fsSL https://get.omniscribe.dev/install.sh | sh

The installer checks your system, installs dependencies, downloads the binary, and adds it to your PATH.

Requires: ydotool for text injection. The installer offers to install it automatically.

2. Start the Daemon

Open a terminal and run:

omniscribe-core daemon

The daemon runs in the background, listening for your hotkey and processing audio.

You'll see:

[INFO] Config loaded. Hotkey: F9
[INFO] Audio device: Your Microphone
[INFO] Global hotkey registered: F9
[INFO] Daemon running. Press F9 to start dictation.

Leave this terminal running. Open another for the TUI.

3. Open the TUI (Optional)

In a second terminal:

omniscribe-core tui

The TUI shows your tier, word usage, dictation mode, sign-in status, and settings. If you prefer just the widget, skip this — the daemon works independently.

4. Start Dictating

  1. Open any text field (browser, editor, terminal, IDE)
  2. Press F9 (default hotkey)
  3. Start speaking
  4. Press F9 again to stop
  5. Text appears where your cursor is

That's it. No cloud, no latency — your voice becomes text instantly.

5. Sign In for Pro Features

Free tier: 500 words per 5-hour window, requires internet for word cap sync.

Upgrade to Pro ($5/month) for:

  • Unlimited words
  • Works fully offline
  • No word cap checks

Sign in from the TUI: press s → browser opens → authenticate → done. Or sign up at omniscribe.dev/signup.

6. Polish Modes

OmniScribe has three polish modes that format your dictation:

ModeWhat it doesExample
VoiceFixes grammar, removes filler words"uh, I think we should deploy today" → "I think we should deploy today"
NotesStructures into Markdown bullet pointsMeeting notes → ## Action Items\n- Deploy\n- Review
EmailFormats as professional emailRaw dictation → "Subject: ...\n\nDear ...,\n\n..."

Switch modes in the TUI or configure per-mode hotkeys.

7. Configuration

Config file: ~/.config/omniscribe/config.toml

[hotkey]
combo = "F9"

[stt]
model_override = "ggml-base"

[polish]
enabled = true
mode = "voice"

[recording]
mic_device = "default"
vad_threshold = 0.02
silence_duration_secs = 1.5

Run calibration to auto-tune for your mic:

omniscribe-core calibrate

8. Uninstall

omniscribe-core clear && rm ~/.local/bin/omniscribe-core

Troubleshooting

Text doesn't appear

Install ydotool and start the daemon:

sudo apt install ydotool        # Debian/Ubuntu
sudo dnf install ydotool        # Fedora
sudo pacman -S ydotool          # Arch

sudo ydotoold &                 # start the daemon

No input device

Check your microphone:

arecord -l                      # list capture devices
pactl list sources | grep Name  # PulseAudio sources

Set a specific mic in config.toml.

Hotkey not working

  • Make sure no other app is capturing F9
  • Try a different hotkey: F10, F11, F12
  • Some Wayland compositors restrict global hotkeys — try X11

Model download failed

Check your internet connection. Models are ~75 MB (tiny) to ~1.5 GB (large).

omniscribe-core install-models          # recommended for your hardware
omniscribe-core install-models ggml-base # specific model

Pricing & plans →Report an issue →