Skip to content

How-To Guides

Practical, copy-pasteable recipes for common YazSes tasks. Each guide is self-contained — pick the one that matches what you want to do.

Guide What it covers
Add words to your personal dictionary Teach STT names, jargon, and acronyms it keeps mis-hearing.
Create spoken macros and snippets Say a trigger phrase to expand canned boilerplate or templates.
Apply a house style to your dictation Keep terminology and spellings consistent from a rules file.
Set a tone per application Casual in Slack, formal in email, verbatim in a terminal — chosen by the window you speak into.
Jump to a symbol by voice Say "go to line 240" or "jump to function tokenize" and the editor moves there.
Change the hold-to-talk key Pick a hotkey (and a dedicated command key) that never clashes with typing.
Dictate into a remote SSH host Type with your voice into a shell or editor on another machine.
Aim dictation with your gaze Glance at a window and hold-to-talk — the text lands in the pane you looked at (X11).
Set up hands-free input One preset that composes gaze, head pointer, face switch and chords; what it sets, what it needs, every way to stop it, and how it recovers.
Stop hands-free input by voice Say a short phrase and the pointer, chords and face switch pause, with no key and no pointer.
Use a muscle, blink or jaw-clench switch Squeeze a muscle to hold-to-talk, or double-blink for a command: configure it, calibrate it to you, and keep its false-positive budget in check (experimental).
Hardware for the biosignal switch A DIY serial muscle sensor or a BrainFlow board, which channels are read, the mains setting, and the safety rules.
Click anywhere by voice Say "mouse grid", cell numbers, then "click" — pointer control on any pixels (experimental; X11 or macOS).
Transcribe recordings to text Turn an existing audio/video file into text — offline, with speaker labels.
Tune for speed and accuracy Choose the STT model, set the VAD threshold, and manage CPU load.
The update said it worked but the version did not change A package manager can exit 0 without upgrading anything — the reinstall command per install method.
Stop a misheard command from running Hold rm -rf, mkfs, curl \| sh until you say confirm — on in a fresh install.
Over SSH and in Remote-SSH editors Why dictation into an SSH session needs no setup, and when the remote agent is needed.
Connect YazSes to an MCP client Let an agent transcribe files on your machine, and optionally ask you a question out loud: the one-click bundle or the manual config.
Install only what you need Read the DOWNLOAD column before you turn a capability on, and keep an install small.
Choose a model on a low-RAM machine Measured memory per Whisper model, and which features load a model of their own.
Reduce CPU use and battery drain What dictation costs on a laptop, and the settings that actually lower it.
Run fully air-gapped Which files YazSes downloads, how to carry them to an offline machine, and how to check nothing phones home.
Fix "Silent audio -- discarding" Set the voice-activity threshold with yazses mic-level, or find the muted microphone.
Dictation types into the wrong window The causes, the no-text-target guard, and precise detection with AT-SPI on Linux.
The tray icon does not appear The AppIndicator requirement, the tray's own error log, and what each desktop needs.
Dictation stopped after a distro upgrade A moved Python, a switch to Wayland, lost input-group access, or a stopped ydotoold.

See also