Version2.36.0
Revision416
Size444.6 MB
Licenseunset
Confinementstrict
Basecore24

Local, offline voice dictation for Linux

ScreenshotScreenshotScreenshotScreenshotScreenshot

YazSes is offline, on-device voice dictation. Hold a key (Space by default), speak, release — your words are transcribed locally and typed into whatever app has focus. No cloud, no API key, no subscription; nothing leaves your machine.

Features:
  • Offline speech-to-text with faster-whisper (CPU, int8) — no GPU needed
  • Voice commands for editors and terminals: "undo that", "save file", "go to line 42", "run the tests", "rename this to user_id"
  • Personal vocabulary and multi-step macros
  • Works in any focused text field — editors, browsers, terminals, chat

INSTALL — all four commands are required. A snap cannot connect its own interfaces, and the daemon starts and looks healthy without them; it simply never hears you, or never sees the key.

sudo snap install yazses
sudo snap connect yazses:audio-record
sudo snap connect yazses:raw-input
yazses setup

  • audio-record — the microphone. Without it, dictation hears nothing.
  • raw-input — the hold-to-talk key, read from /dev/input/event*. Without it the hotkey does nothing at all.
  • yazses setup — provisions the rest, and re-checks both interfaces.

THEN, TO START DICTATING

yazses doctor (reports anything still missing)
yazses mic-level --set (tune the silence gate to your voice, ~4 s)
yazses start (hold Space, speak, release)

First run downloads the speech model (~250 MB–1 GB): one internet connection, then everything after that is offline. Config and models live under ~/snap/yazses/. To remove: sudo snap remove yazses

BEFORE YOU INSTALL
  • Architecture: stable is amd64 today. On arm64, run "sudo snap install yazses --edge" (one release behind) or install with pipx.
  • Session: this is a strictly-confined snap and keystroke injection works on X11 sessions. For Wayland, or for full input access, install via pipx instead: https://pypi.org/project/yazses/
  • A snap ships a fixed set of libraries, so features cannot be added to it later. Included: dictation, voice commands, tray, overlay, Meeting Mode with speaker diarization, the Parakeet English engine, spoken read-back. Not included, and needing a pipx install instead: Cocktail Filter, offline LLM cleanup, Glance-Type, prosody-aware punctuation.

Docs: https://mskazemi.com/yazses/
Issues: https://github.com/MSKazemi/yazses/issues

Update History

2.35.0 (415)2.36.0 (416)
30 Aug 2026, 19:45 UTC
2.34.0 (412)2.35.0 (415)
27 Aug 2026, 21:15 UTC
2.33.0 (411)2.34.0 (412)
27 Aug 2026, 18:30 UTC
2.32.0 (409)2.33.0 (411)
27 Aug 2026, 16:45 UTC
2.32.0 407 → 409
26 Aug 2026, 14:30 UTC
2.31.0 (388)2.32.0 (407)
26 Aug 2026, 13:45 UTC
2.29.0 (348)2.31.0 (388)
26 Aug 2026, 11:30 UTC
2.20.0 (154)2.29.0 (348)
23 Aug 2026, 02:30 UTC
2.19.0 (149)2.20.0 (154)
23 Aug 2026, 01:00 UTC
2.18.2 (118)2.19.0 (149)
14 Aug 2026, 15:45 UTC
2.18.0 (115)2.18.2 (118)
13 Aug 2026, 17:00 UTC
2.17.0 (60)2.18.0 (115)
13 Aug 2026, 14:15 UTC
2.16.0 (54)2.17.0 (60)
9 Aug 2026, 10:30 UTC
2.15.1 (25)2.16.0 (54)
9 Aug 2026, 02:30 UTC
2.15.0 (24)2.15.1 (25)
7 Aug 2026, 20:45 UTC
2.14.0 (23)2.15.0 (24)
7 Aug 2026, 19:30 UTC
2.13.0 (22)2.14.0 (23)
7 Aug 2026, 04:30 UTC
2.12.0 (21)2.13.0 (22)
7 Aug 2026, 04:00 UTC
2.12.0-dev.4 (20)2.12.0 (21)
31 Jul 2026, 22:00 UTC
2.11.0-dev.10 (17)2.12.0-dev.4 (20)
30 Jul 2026, 06:00 UTC

Published31 May 2026, 23:07 UTC

Last updated30 Aug 2026, 19:12 UTC

First seen1 Jun 2026, 00:00 UTC