User Guide

Last updated: February 26, 2026

Getting Started

Activation

Hold the Fn (Globe) key to start dictating. Release to stop and insert text into the active app. You can also use the Right Option key as an alternative trigger.

Auto-Stop Mode

Double-click your trigger key to enter auto-stop mode. Dictation continues after you release the key and automatically stops after 2 seconds of silence. This is ideal for longer dictation sessions where you don't want to hold a key down.

Hands-Free with Siri

Say "Hey Siri, start touch" to begin dictation completely hands-free. In auto-stop mode, your text is automatically inserted after 2 seconds of silence.

Accessibility Permission

Hush Touch needs Accessibility permission to insert text into other apps and detect when you release the trigger key. Go to System Settings > Privacy & Security > Accessibility and enable Hush Touch.

Without Accessibility permission, the app falls back to auto-stop mode (text is inserted after 2 seconds of silence instead of on key release).

Trigger Keys

Choose which key activates dictation. Configure via the status bar menu under Trigger Key.

Key Behavior
Fn (Globe) Default trigger. Hold to dictate, release to stop.
Right Option Alternative trigger. Same hold-to-dictate behavior.
Double-click either key Enters auto-stop mode. Dictation continues hands-free and stops after 2 seconds of silence.

Overlay Controls

When the transcription overlay is enabled, it shows your live dictation with keyboard shortcuts for quick actions. Toggle the overlay on or off via the status bar menu.

Shortcut Action
Esc Cancel dictation and close the overlay without inserting text.
+ Space Send/submit the transcribed text into the active app.

Voice Commands

Speak these commands during dictation to control formatting and text flow.

Formatting

Say Result
new line Inserts a single line break
new paragraph Inserts a paragraph break (double line break)
all caps on Starts typing in ALL CAPS
all caps off Returns to normal case

Caps toggle also accepts: caps on/off, caps lock on/off, uppercase on/off, and upper case on/off.

Send Phrase

The send phrase stops dictation and presses Enter in the target app. The default is ok press send — customize it in the status bar menu.

Variations are automatically recognized: "okay press send", "please press send", etc. You can prefix commands with: ok, okay, please, hey, hush, or touch.

Stop Phrases

Say stop hush touch or stop hush to stop dictation without pressing Enter. Your transcribed text is still inserted.

Emoji Dictation

Insert emoji by speaking their names. There are two styles depending on how unique the name is.

Say "emoji" after the name

For common single words that might appear in normal speech, add "emoji" after the name to avoid false positives.

Say Result
heart emoji❤️
fire emoji🔥
star emoji
sun emoji☀️
moon emoji🌙
eyes emoji👀
party emoji🎉
wave emoji👋
skull emoji💀
crown emoji👑
rainbow emoji🌈
rocket emoji🚀
poop emoji💩
muscle emoji💪
sparkles emoji
100 emoji💯
ghost emoji👻
alien emoji👽
robot emoji🤖
bomb emoji💣
peace emoji✌️
kiss emoji😘

Multi-word names (no suffix needed)

These phrases are unique enough to be recognized without saying "emoji" afterward.

Say Result
smiley face😊
sad face😢
crying face😭
laughing face😂
winking face😉
thinking face🤔
thumbs up👍
thumbs down👎
clapping hands👏
folded hands🙏
praying hands🙏
waving hand👋
flexed bicep💪
party popper🎉
check mark
cross mark
broken heart💔
rolling eyes🙄
mind blown🤯
face palm🤦
hot dog🌭

The Dictation engine (macOS 26+) also has built-in emoji support, so saying phrases like "smiley face" may produce the emoji directly from the speech engine itself.

Speech Engines

Hush Touch supports multiple speech recognition engines. Select one via the status bar menu under Speech Engine.

Engine macOS Description
Dictation 26+ Apple's dictation engine with auto-punctuation, native emoji, and formatting commands. No time limit. 100% on-device.
Speech 26+ Apple's newest on-device speech-to-text model. Fast streaming output.
Classic 15+ Legacy Apple engine. Supports custom vocabulary for domain-specific terms. ~1 minute time limit per session.
Dual 26+ Runs Dictation + Classic concurrently and intelligently merges results. Best accuracy for custom vocabulary with natural punctuation.
Lightning Whisper MLX 15+ Local Whisper model powered by Python + MLX. Requires separate installation via pip.
Parakeet TDT 15+ Local Parakeet 0.6B model via Python + ONNX. Requires separate installation.

Accuracy Over Speed (Dual only)

When using the Dual engine, enable Accuracy Over Speed in the Speech Engine menu to wait longer for Classic engine results before merging. This can improve vocabulary capture at the cost of slightly slower output.

Smart Text Features

Toggle these features via the status bar menu under Smart Text.

Feature Default Description
Remove Filler Words ON Strips filler words: "um", "uh", "umm", "hmm", "you know", "you know what I mean", "basically", and ", like,"
Format Numbered Lists ON Detects spoken ordinals like "number one", "first", "secondly" and formats them as numbered lists (1. 2. 3.)
Auto-Correction OFF Say "actually", "I mean", "no wait", or "sorry" to replace the previous clause with what follows
Voice Snippets ON Expands custom trigger phrases into longer replacement text (see Voice Snippets)

Stutter removal is always active — repeated words like "the the" or "to the to the" are automatically cleaned up.

Context Modes

Context modes override individual Smart Text toggles to optimize for specific workflows. Set via the status bar menu under Smart Text > Context Mode, or per-app via App Profiles.

Mode Filler Removal Auto-Correction Numbered Lists Voice Snippets
General Your setting Your setting Your setting Your setting
Email Forced ON Forced ON Forced ON Your setting
Code Forced ON Forced OFF Forced OFF Your setting
Notes Forced ON Your setting Forced ON Your setting

Audio Settings

Configure audio behavior via the status bar menu under Audio.

  • Sound Effects (default: ON) — Plays an audio cue when dictation starts and stops.
  • Lower Volume While Listening (default: ON) — Reduces system volume during recording so the microphone picks up your voice more clearly.

Custom Vocabulary

Teach Hush Touch words it doesn't recognize — technical terms, brand names, acronyms, or names specific to your work.

  • Up to 500 custom words and phrases
  • Auto-learns from corrections — when you fix a transcription error, the correct word is saved automatically (requires Auto-Learn Per App to be ON)
  • Per-app profiles — create up to 10 app profiles with app-specific vocabulary and context modes via Edit App Profiles
  • Edit vocabulary via the status bar menu: Edit Vocabulary

Vocabulary data is stored locally on your Mac at ~/Library/Application Support/Hush Touch/ and never leaves your device.

Voice Snippets

Voice snippets let you expand short trigger phrases into longer replacement text — great for email signatures, addresses, boilerplate responses, or anything you type repeatedly.

  • Up to 50 snippet mappings
  • Format: trigger phrase → replacement text
  • Example: my email → john@example.com
  • Edit via the status bar menu: Edit Voice Snippets

When you say the trigger phrase during dictation, it's automatically replaced with the full expansion.

Apple Intelligence Final Output

On macOS 26+, Hush Touch can use Apple Intelligence as a final polish pass on your transcription before inserting it. Configure via the status bar menu.

Modes

  • Off — No AI post-processing (default)
  • Proofread — Fixes grammar and spelling while preserving your voice
  • Rewrite — Rewrites the output in a chosen style

Advanced Settings

When Proofread or Rewrite is active, you can fine-tune the AI behavior:

  • Temperature — Controls randomness (0.0 = deterministic, 1.0 = creative). Default: 0.15
  • Max Tokens — Maximum output length. Default: 384
  • Rewrite Style — Light, Balanced, or Heavy. Default: Light
  • Sampling Mode — Greedy, Top-K, or Probability. Default: Greedy

Text Processing Pipeline

Your dictation passes through these processing stages in order. Understanding the pipeline helps explain why certain features interact the way they do.

  1. Voice snippets — Trigger phrases are expanded first, before any other processing
  2. Emoji conversion — Spoken emoji names are converted to Unicode characters
  3. Formatting commands — "new line" and "new paragraph" are converted to line breaks
  4. All caps — "all caps on/off" toggles uppercase mode on surrounding text
  5. Filler word removal — "um", "uh", and other filler words are stripped
  6. Stutter removal — Repeated words and bigrams are cleaned up
  7. Auto-correction — "actually" and similar triggers replace the previous clause
  8. Numbered list formatting — Spoken ordinals are formatted as numbered lists

After this pipeline, if Apple Intelligence is enabled, a final proofread or rewrite pass is applied before the text is inserted.