Dawnstar ReadAloud — Highlight, Hotkey, Hear

If you prefer listening to text, Dawnstar ReadAloud makes it effortless on Linux. Many TTS options today sound robotic, are paid, or require copying text into a browser. I wanted a faster, more natural workflow: highlight any text, press a hotkey, and hear neural voices instantly. Here's how I built it

The Goal

A simple TTS system that:

• Uses neural voices (not robotic synthesis)

• Works with global keyboard shortcuts in any app

• Caches audio for instant replay of repeated phrases

• Handles files, PDFs, EPUBs, and URLs

• Works on both X11 and Wayland

The Solution

A Python CLI that leverages Microsoft Edge TTS (Azure neural voices, free) with MD5-based caching and desktop integration.

How It Works

1. Text Input: Accepts direct text, files, URLs, or the clipboard via xclip (X11) or wl-paste (Wayland)

2. Chunking: Splits text at sentence boundaries (. ! ?) for natural speech rhythm

3. TTS Engine: Primary Edge TTS, with gTTS and eSpeak as fallbacks

4. Caching: Stores audio in ~/.cache/tts_app/ keyed by md5(text + lang + speed)

5. Playback: Auto-detects player (mpg123 → paplay → cvlc → ffplay)

The Setup

#!/bin/bash
# Install Dawnstar ReadAloud

# System dependencies
sudo apt install mpg123 xclip poppler-utils

# Clone and install
git clone https://github.com/emeeran/Dawnstar-ReadAloud.git tts && cd tts
python3 -m venv .venv
./.venv/bin/pip install -e .

# Test it
./tts "Hello! This sounds surprisingly natural."

# Set up global hotkeys
python3 configure.py

Setting Up Keyboard Shortcuts

The configure.py Script auto-detects your desktop and registers hotkeys. For GNOME via gsettings:

# Speak selection: Ctrl+Alt+S
gsettings set org.gnome.settings-daemon.plugins.media-keys.custom-keybinding:/org/gnome/settings-daemon/plugins/media-keys/custom-keybindings/custom0/ name 'TTS Speak'
gsettings set org.gnome.settings-daemon.plugins.media-keys.custom-keybinding:/org/gnome/settings-daemon/plugins/media-keys/custom-keybindings/custom0/ command '/home/you/.local/bin/tts-speak'
gsettings set org.gnome.settings-daemon.plugins.media-keys.custom-keybinding:/org/gnome/settings-daemon/plugins/media-keys/custom-keybindings/custom0/ binding '<Ctrl><Alt>s'

# Stop speaking: Ctrl+Alt+Q
gsettings set org.gnome.settings-daemon.plugins.media-keys.custom-keybinding:/org/gnome/settings-daemon/plugins/media-keys/custom-keybindings/custom1/ name 'TTS Stop'
gsettings set org.gnome.settings-daemon.plugins.media-keys.custom-keybinding:/org/gnome/settings-daemon/plugins/media-keys/custom-keybindings/custom1/ command '/home/you/.local/bin/tts-stop'
gsettings set org.gnome.settings-daemon.plugins.media-keys.custom-keybinding:/org/gnome/settings-daemon/plugins/media-keys/custom-keybindings/custom1/ binding '<Ctrl><Alt>q'

# Register them
gsettings set org.gnome.settings-daemon.plugins.media-keys custom-keybindings "['/org/gnome/settings-daemon/plugins/media-keys/custom-keybindings/custom0/', '/org/gnome/settings-daemon/plugins/media-keys/custom-keybindings/custom1/']"

On X11, the primary selection (highlighted text) is prioritized over clipboard—you don't need to copy, just highlight.

The Result

Highlight any text, press Ctrl+Alt+S, and hear it spoken:

# Cache stats after using the hotkey
./tts --cache-stats

Cache Statistics:
  Files: 12
  Size: 4.2 MB / 500 MB
  Location: /home/you/.cache/tts_app/

First playback takes ~1 second (network request), subsequent plays of the same text are instant.

Usage Examples

./tts "Direct text"              # Speak directly
./tts document.txt               # Read a file
./tts book.epub                  # Read EPUB (skips front matter)
./tts https://example.com        # Read a webpage
cat notes.txt | ./tts -          # Read from stdin
./tts -l en-uk -s slow file.txt  # British English, slow speed

Requirements

• Python 3.12+

• mpg123 (recommended audio player)

• xclip or wl-clipboard (for keyboard shortcuts)

• poppler-utils (PDF support, optional)

# Ubuntu/Debian
sudo apt install mpg123 xclip poppler-utils

Why This Matters

The best TTS system is the one you actually use. By reducing friction to a single hotkey, I find myself listening more—documentation while cooking, proofreading by ear, articles during walks. The neural voices make it pleasant rather than tolerable.

The modular architecture makes it easy to extend: add languages in core/constants.py, swap backends in core/engine.py. Full documentation at USER_MANUAL.md.

Happy listening!