OpenReader help
Frequently asked questions.
The short version for privacy, platforms, exports, voices, and how to get help.
Is OpenReader free?
Yes. OpenReader is open source under the AGPL-3.0 license. Anyone can fork it, modify it, or self-host it. A separate commercial license is available for teams who can't comply with AGPL terms, write through the contact page.
Does it work on Intel Macs?
No. The desktop app is Apple Silicon only (M1, M2, M3, and newer). The local TTS models run on MLX, which targets the Apple Neural Engine and unified memory. There is no Intel fallback planned.
When will Windows or Linux come?
No firm date. The app is built on Tauri, so a port is technically viable, and PyTorch gives us a path off MLX for non-Apple hardware. Sequencing depends on demand. If you want it, say so on the GitHub Discussions tracker.
How does the App Store edition work?
OpenReader's App Store edition is sandboxed, and all executable inference code ships inside the signed app bundle. Voice model weights download on first use because they are data, not executable code.
Is my reading private?
Yes. The desktop app generates audio locally on your device. The text you load and the audio it produces never leave your device. Voice models download from Hugging Face on first use. User-initiated article imports, RSS/OPDS feeds, and remote catalog reads connect only to the source you choose.
The web library at openreader.io uses Resend only for the optional email signup you submit explicitly. We collect zero analytics. See the privacy policy for the full picture.
How do I uninstall?
Drag OpenReader.app to the Trash. Optionally remove the app data directory at ~/Library/Application Support/OpenReader and the Hugging Face model cache at ~/.cache/huggingface/. That's it, no background services, no launch agents, no leftover daemons.
Can I add my own TTS model?
Yes. OpenReader's backend is structured around a model registry. See CONTRIBUTING.md and the existing entries under apps/desktop/backend/models/ for the shape, then open a PR.
What audio formats can I export?
Projects export as a .zip archive containing MP3 chunks plus a manifest with word timings. Re-import it elsewhere, stitch the chunks together with ffmpeg, or stream the result into a podcast app.
Which voices and languages are supported?
Apple system voices work immediately on macOS. Two natural local model families are also available on demand: Kokoro-82M offers 54 voices across 8 languages and is fast enough for real-time use, while Qwen3-TTS-1.7B offers 9 voices across 10 languages with higher fidelity at the cost of speed. Their model files download only after you choose them, and you can mix voices per project.
How do I report a bug or request a feature?
File an issue or start a thread on GitHub. Bugs go in Issues, requests and longer discussions go in Discussions.