Built for visually impaired gamers and computer users— works over ANY application, including fullscreen games with high quality voices.
Click to watch HockEye Detailed Demo playing Life Is Strange and reading aloud dynamic content across various applications.
HockEye is a free, fully local screen reader and select to speak overlay for Linux. Press a hotkey, drag to select any area of your screen, and it instantly reads the text aloud — over games, videos, apps, anything.
No subscriptions. No internet. No API keys. Runs entirely on your machine.
Built by a gamer who just wanted to play Life is Strange and hated the non-existence of any select to speak like application on computers. So, just decided to create it :))
Most screen readers don't work over fullscreen games. Commercial alternatives are expensive and Windows-only. No free select-to-speak overlay existed that worked seamlessly for Linux gamers in fullscreen.
HockEye was built specifically to fix that — starting with Life is Strange, expanding to everything.
- Press Numlock → drag to select any screen region → text is read aloud instantly
- GPU accelerated OCR via EasyOCR — handles stylized game fonts, mixed backgrounds
- Natural female voice via Piper TTS — runs fully offline
- Smart sentence chunking — minimal delay before first word is spoken aloud
- Magnifier Support in fullscreen Games! - While in Selection Mode Magnifier can be used to zoom in on content even in fullscreen which was not possible before on most computers
- Works over any application — games, browsers, terminals, anything
- 100% free and local — no API, no cost, no internet required
- Linux with X11 display server
- Python 3.10+
- NVIDIA GPU recommended (falls back to CPU)
pipandvenvtkinterandALSA utilities (For Aplay)
git clone https://github.com/PoeticKiler/HockEye.git
cd HockEyepython3 -m venv hockeye-env
source hockeye-env/bin/activatesudo apt install python3-tk alsa-utilspip install -r requirements.txt
pip install torch torchvision --index-url https://download.pytorch.org/whl/cu121Note: torch is installed separately to explicitly target the CUDA 12.1 build without bloating the standard requirements file (If your system runs an older NVIDIA driver requiring an older runtime, you can change
cu121tocu118in the URL above.)
mkdir voices
cd voices
wget https://huggingface.co/rhasspy/piper-voices/resolve/main/en/en_US/lessac/high/en_US-lessac-high.onnx
wget https://huggingface.co/rhasspy/piper-voices/resolve/main/en/en_US/lessac/high/en_US-lessac-high.onnx.json
cd ..Open main.py and update this line with your actual path:
VOICE_MODEL = '/your/path/to/voices/en_US-lessac-high.onnx'python main.py- Run
python main.py— wait for "HockEye is running!" message - Press Numlock at any time — even inside a fullscreen game
- The tool is Ready! — drag to select any text on screen
- Release mouse — text is read aloud instantly
- Overlay closes — you're back in your game
| Component | Time |
|---|---|
| OCR (NVIDIA RTX 3050) | ~0.5 seconds |
| TTS first syllable | ~0.3 seconds |
| Total | ~0.8 seconds |
CPU fallback is slower (~3-4 seconds for OCR).
- Phase 1 — Select to Speak overlay
- Phase 2 — Real time object/enemy detection and highlighting (YOLOv8)
- Phase 3 — AI Driven visual to audio gaming
- Ongoing phases side by side
- Windows support (Very Soon... Please reach out for more info)
- Wayland support
Contributions welcome! This project exists to help visually impaired people access games and applications. If you can improve it, please do.
- Fork the repo
- Create your branch (
git checkout -b feature/AmazingFeature) - Commit your changes (
git commit -m 'Add AmazingFeature') - Push to branch (
git push origin feature/AmazingFeature) - Open a Pull Request
If you represent an accessibility organization, gaming company, or want to collaborate:
GitHub Issues — for bugs and features
Email — adarshchaturvedi236536@gmail.com
Location — PEC Chandigarh, India
This project was built for the visually impaired gaming community. If you can help spread it, please do!
- X11 only — Wayland not supported yet
- First run — EasyOCR downloads models (~2GB) on first launch. Subsequent runs are instant
- GPU Recommended - CPU Only users without dedicated GPU will face OCR delays.
MIT License — see LICENSE for details.
Free to use, modify, and distribute. Just keep the attribution.
Built with ❤️ for the Linux accessibility community.
"Hearing a 🐧 is always easier than spotting one."