Home

YazSes Wiki

Hold a key → speak → release. YazSes is offline, on-device voice dictation for Linux, macOS, and Windows — it types your speech into any app, plus voice commands and macros. No cloud, no API key, no subscription. Audio never leaves your machine.

Users: the documentation site has install + usage. This Wiki is the contributor & developer knowledge base — how it works inside, where it's going, and how to make your first pull request.

---

🚀 New here?

pipx install yazses            # or: pip install yazses
yazses quickstart              # 3-step, machine-tailored setup
yazses doctor                  # self-check: mic, hotkey, model, permissions

Then hold your hotkey, speak, release — it types.

🧭 Wiki map

Page	What's in it
Feature Catalog	All 139 capabilities across 9 areas, with a "good ways to contribute" note per area.
How It Works	The dictation pipeline, the daemon, and the module map — read this before touching code.
Roadmap & Vision	Where YazSes is heading (a voice-first, 100%-offline interaction layer) and how to help.
Contributing — Start Here	Dev setup, your first PR, good first issues, how review works, recognition.
FAQ	Common questions — privacy, accuracy, languages, Wayland, comparisons.

🙋 Want to contribute?

We're actively looking for contributors and we review fast.

• 👉 Good first issues — pick one; many are 10–30 minutes.
• 💡 Ideas & feature voting — suggest and upvote what to build next.
• 🙏 Q&A — ask anything.

🧱 What makes it solid

• 100% on-device. Transcription runs locally via faster-whisper (CPU int8). Nothing is uploaded, ever.
• Cross-platform by design. One daemon, an OS-abstraction layer (platform/) with Linux/macOS/Windows backends behind shared Protocol interfaces.
• Tested. A large, green test suite runs on Linux × macOS × Windows (Python 3.11 & 3.12) on every push.
• Opt-in everything. Advanced features ship off by default — installing YazSes never changes your setup until you turn something on.