Skip to main content
Configure advanced AI behavior, audio processing, language settings, and privacy controls.

​
AI Behavior

​
System Prompt

​
systemPrompt
string
Custom system prompt for AI assistantThis prompt sets the behavior, personality, and response style of the assistant. It’s sent with every AI request.
Keep prompts concise and clear. The assistant is optimized for voice responses, so instruct it to be brief.
Default prompt:
You are SlasshyWispr, a helpful desktop voice assistant. 
Keep replies concise and easy to speak aloud.
Example custom prompts:
You are a coding assistant. Provide concise code examples 
and explanations. Keep responses under 3 sentences unless 
code is requested.

You are a writing coach. Help improve clarity, grammar, 
and style. Be encouraging and constructive. Keep feedback 
brief and actionable.

You are a personal assistant named Lily. Be friendly, 
efficient, and proactive. Prioritize quick, helpful 
responses for daily tasks.

​
Model Parameters

​
temperature
number
default:"0.35"
AI model temperature (creativity vs consistency)
  • Range: 0.0 to 2.0
  • 0.0 = Deterministic, consistent, focused
  • 0.35 = Balanced (default)
  • 1.0 = Creative, varied responses
  • 2.0 = Highly creative, unpredictable
Lower temperatures (0.2-0.5) are recommended for dictation and factual tasks. Higher temperatures (0.7-1.2) work better for creative writing.
​
maxTokens
number
default:"320"
Maximum tokens in AI response
  • Default: 320 (~240 words)
  • Lower values = shorter, faster responses
  • Higher values = longer, more detailed responses
Voice responses should be concise. 320 tokens is optimized for TTS playback length.

​
Style Profiles

​
styleProfile
StyleProfile
default:"adaptive"
Response style and tone preset
  • adaptive - Adjusts style based on context (default)
  • professional - Formal, business-appropriate tone
  • casual - Friendly, conversational tone
  • concise - Ultra-brief, minimal responses
  • developer - Technical, code-focused responses
Dynamically adjusts tone based on:
  • Time of day
  • Query complexity
  • Previous conversation context
Best for general-purpose use.

​
Language Settings

​
dictationLanguage
string
default:"en"
Primary dictation languageSupported languages:
  • en - English
  • es - Spanish
  • fr - French
  • de - German
  • it - Italian
  • pt - Portuguese
  • hi - Hindi
  • bn - Bengali
  • ja - Japanese
  • ko - Korean
  • zh - Chinese
  • ar - Arabic
  • ru - Russian
​
dictationLanguageMode
DictationLanguageMode
default:"single"
Language detection mode
  • single - Use one language for all dictation (default)
  • multiple - Auto-detect from allowed languages
Multiple language mode requires dictationLanguageAllowList to be configured.
​
dictationLanguageAllowList
string[]
default:"[]"
Allowed languages for auto-detectionWhen dictationLanguageMode is set to multiple, the STT model will detect and transcribe in any of these languages.Example:
["en", "es", "fr"]

​
Audio Processing

​
removeFillers
boolean
default:"false"
Remove filler words from transcriptionAutomatically removes:
  • “um”, “uh”, “er”
  • “like”, “you know”
  • Repeated words
  • Hesitations
This improves transcript readability but may occasionally remove intentional words.
​
autoPunctuation
boolean
default:"false"
Automatically add punctuation to transcriptionThe STT model infers punctuation from speech patterns:
  • Periods at natural pauses
  • Question marks for rising intonation
  • Commas for breath breaks
Works best with online models. Local models may have limited punctuation support.
​
numberedLists
boolean
default:"false"
Convert spoken lists to numbered formatWhen enabled:
  • “First, second, third” → “1. 2. 3.”
  • “Item one, item two” → “1. 2.”
Useful for dictating structured content.
​
backtrackCorrection
boolean
default:"false"
Enable voice correction commandsAllows you to say:
  • “Scratch that” - Remove last sentence
  • “Undo” - Remove last word
  • “Delete that” - Remove last phrase
Experimental feature. May occasionally trigger on similar-sounding phrases.
​
dictationSoundEffects
boolean
default:"false"
Play audio feedback during dictation
  • Start recording: chime
  • Stop recording: click
  • Processing: subtle tone
  • Error: alert sound
​
muteMusicWhileDictating
boolean
default:"false"
Automatically mute system audio during recordingPrevents background music/audio from being picked up by the microphone.
Requires system audio control permissions.

​
Privacy and Data

​
incognitoMode
boolean
default:"false"
Disable history loggingWhen enabled:
  • No conversation history saved
  • No transcripts stored locally
  • Home history remains empty
  • Usage stats not tracked
This does NOT affect data sent to online model providers. Use local mode for complete privacy.
​
contextAwareness
boolean
default:"false"
Enable multi-turn conversation contextWhen enabled, the assistant remembers previous turns in the session for context-aware responses.
Context is stored in memory only and cleared when the app restarts (unless incognito mode is disabled).

​
Workflow Settings

​
autoPasteDictation
boolean
default:"false"
Automatically paste transcription into active appWhen enabled, transcribed text is immediately pasted at cursor position. When disabled, text is copied to clipboard only.
​
copyToClipboard
boolean
default:"false"
Copy assistant responses to clipboardUseful for saving AI responses without manual selection.
​
commandMode
boolean
default:"false"
Enable command detection in dictationWhen enabled, spoken commands like “new line”, “period”, “comma” are converted to formatting.
​
wakeWordEnabled
boolean
default:"false"
Enable wake word detection for hands-free assistantWhen enabled, say the wake phrase (e.g., “Hey Lily”) to activate the assistant without pressing a hotkey.
Wake word detection runs continuously and may increase battery usage.

​
Microphone

​
microphoneDeviceId
string
default:""
Selected microphone device IDSet in Settings > General > Microphone. SlasshyWispr lists all available input devices.Leave empty to use system default microphone.

​
UI and Appearance

​
themeMode
ThemeMode
default:"system"
Application theme
  • system - Match OS theme (default)
  • light - Always light mode
  • dark - Always dark mode
​
showFlowBar
boolean
default:"false"
Show live pipeline status bar during processingDisplays real-time STT/AI/TTS latency and status.
​
showAppInDock
boolean
default:"false"
Show SlasshyWispr icon in dock/taskbarWhen disabled, app runs in system tray only.
​
launchAtLogin
boolean
default:"false"
Start SlasshyWispr automatically on system boot
May require OS permissions on first enable.

​
Best Practices

​
For Privacy

  • Enable incognito mode to prevent local history
  • Use local runtime mode to keep data on-device
  • Disable context awareness to avoid multi-turn tracking
  • Disable remember API key to avoid persisting credentials

​
For Accuracy

  • Enable auto punctuation for better readability
  • Enable remove fillers for cleaner transcripts
  • Use single language mode unless multilingual dictation is needed
  • Set temperature to 0.2-0.4 for factual/dictation tasks

​
For Performance

  • Use local STT + online AI for balanced speed
  • Lower max tokens to 200-250 for faster responses
  • Enable Piper TTS instead of Coqui for fastest playback
  • Disable wake word when not needed to save battery

​
For Workflow

  • Enable auto paste dictation for seamless text entry
  • Enable copy to clipboard to save assistant responses
  • Use concise style profile for quick lookups
  • Customize system prompt to match your use case

Build docs developers (and LLMs) love

Get started for freeTalk to us